Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Import shocks and labour market adjustment: an analysis with longitudinal data Sargent, Timothy C. 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1995-060594.pdf [ 5.7MB ]
Metadata
JSON: 831-1.0088379.json
JSON-LD: 831-1.0088379-ld.json
RDF/XML (Pretty): 831-1.0088379-rdf.xml
RDF/JSON: 831-1.0088379-rdf.json
Turtle: 831-1.0088379-turtle.txt
N-Triples: 831-1.0088379-rdf-ntriples.txt
Original Record: 831-1.0088379-source.json
Full Text
831-1.0088379-fulltext.txt
Citation
831-1.0088379.ris

Full Text

IMPORT SHOCKS AND LABOUR MARKET ADJUSTMENT: AN ANALYSIS WITH LONGITUDINAL DATA. by TIMOTHY CHARLES SARGENT B.A.(Bcon), The Victoria University of Manchester, 1987 M.A., The University of Western Ontario, 1988 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES (Department of Economics) We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA October 1994 © Timothy Charles Sargent, 1994 0 6 ' 2 6 - 9 5 14:08 O 6 0 4 S22 5913 UBC ECONOMICS k£«03 00.: In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of Britisn Columbia, i agree that the Library shaii make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives, it is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. (Signature) Department of &&4&MLCJL The University of British Columbia Vancouver, Canada D«e yu^ XL, MIL DE-6 (2/88) ABSTRACT The objective of this thesis is to examine the labour market effects of increasing international competition. In particular I am interested in whether those who leave employment in import-affected industries find it easy to obtain equally well paid employment elsewhere. The thesis begins with a theoretical overview of the issues involved. After a review of the literature, I argue that a necessary condition for active trade policy to assist workers in an affected industry is that the individuals concerned be earning rents: they must suffer significant losses from displacement. In the next section I construct a theoretical model in which adverse selection in the labour market ensures that some workers earn rents, and that if a firm shuts down because of international competition the social costs from the displacement of its workforce can outweigh the benefits of lower import prices. This model is then adapted into a search framework in order to allow for unemployment. This enables me to derive predictions concerning the observable labour market behaviour of agents in different industries. Next I conduct an empirical analysis of worker displacement in Canada using the Labour Market Activity Survey. After categorising individuals according to the long term health of their industry and the degree of import competition it has faced in recent years, I estimate various duration models using the length of joblessness as the dependent variable, as well as models of the re-employment wage. In both cases it is found that those in import-affected declining industries suffer significant losses, both relatively and absolutely. I also estimate a model for pre-displacement earnings that attempts - ii -to compare the wages of those who leave their job with the wages of those who stay. I find that in declining industries the better paid tend to leave, as predicted by the theoretical model. Finally I find that a relatively small proportion of employees remain in the same industry, and that for the import affected declining industry group there is evidence that changing industry is positively correlated with re-employment, whilst the reverse is true for other industries. - iii -TABLE OF CONTENTS Abstract ii Table of Contents iii List of Tables vi List of Figures viii Acknowledgements x I. Introduction. 1. Outline of Thesis 2 2 . Theoretical Literature 5 3. Empirical Literature 9 II. Theoretical Model - Finite Horizon, No Unemployment. 1. One Period Model when workers are already Skilled 18 2. Two Period Model 25 3 . Conclusions 29 Appendix 32 III. Theoretical Model - Infinite Horizon, Unemployment. 1. Non-Employment duration for laid off workers 36 2. Exhaustion of Unemployment Benefits 44 3 . Conclusions 48 - iv -IV. Employment and Trade Patterns in Canadian Industry 1961-90. 1. Employment 1961-90 53 2 . Changes in Import Penetration 55 3. Declining Industries affected by Import Competition 59 4. Declining Industries not significantly affected by Import Competiton 61 5. Expanding Industries affected by Import Competition 62 6. Expanding Industries not significantly affected by Import Competition 63 Appendix 65 V. Micro Data. 1. Construction of the Data Set 66 2. Summary Statistics 69 VI. A Model for Pre-Displacement Wages. 1. Data 75 2. OLS model of Earnings 18 3 . A Duration Model of Employment Duration 81 4 . Instrumental Variables model of Earnings 84 VII. An Empirical Model of Non-Employment Duration. 1. Duration Models 87 2. Estimated Model 88 3 . Results 91 Appendix 100 - v -VIII. A Self Selection Model for Re-Employment Earnings. 1. The Model 113 2. Estimation 115 3 . Results 115 4 . Further Work 121 IX. An Empirical Model of Sectoral Mobility. 1. Summary Statistics 124 2. Bivariate Probit Model 127 X. Conclusions 132 Bibliography 140 - vi -LIST OF TABLES Employment in Forestry, Mining and Manufacturing 1961-90 54 Decomposition of Change in Gross Output between 1961 and 1985..56 Decomposition of Change in Gross Output between 1985 and 1990..57 Categories of Industry 59 Sample Construction 68 Means of Covariates used in the Calculations 70 Means of Covariates for Left Censored Jobs 1988-90 76 OLS Earnings Equation for Left Censored Jobs 1988-90 78 OLS Earnings Equation for all covariates 79 Results from the Piecewise Constant Duration Model 82 Results from the Instrumental Variables Model 85 Covariate Estimates from the Piecewise Constant Model 94 Covariate Estimates from the Piecewise Constant/Gamma Mixture..55 Expected Non-Employment Duration 96 Estimated Hazards from the the Kaplan Meier Model 100 Estimated Baseline Hazards from the Piecewise Constant Model..103 - vii -Estimated Baseline Hazards from the Piecewise Constant/Gamma Mixture Model 104 Estimated Baseline Hazards from the Piecewise Constant Model: Over 24 years of Age Sample 205 Covariate Estimates from the Piecewise Constant Model: Over 24 years of Age Sample 106 Instrumental Variables/Type 2 Tobit Model of Re-Employment Earnings 117 Wage in Former Job and Expected Re-Employment Wage 118 Characteristics of Former Job o£ those observed to be Re-employed 125 Characteristics of New Job of those observed to be Re-employed.127 Covariate Estimates for Probit Model for Remaining in the Same Industry 129 - viii -LIST OF FIGURES Labour Supply to the lower wage Skilled Industry 21 Payoffs associated with possible strategies of each worker 25 Hazard Functions for Agents laid off in Industry B 50 Behaviour of VN over the Non-Employment spell 51 Simulated Hazard Functions 52 Kaplan Meier Model of Re-Employment: Baseline Survivor Function 107 Piecewise Constant Model of Re-Employment: Baseline Hazard Function 108 Piecewise Constant Model of Re-Employment: Baseline Survivor Function 109 Piecewise Constant/Gamma Model of Re-Employment: Baseline Survivor Function 110 Piecewise Constant Model of Re-Employment (Over 24 years): Baseline Hazard Function Ill Piecewise Constant Model of Re-Employment (Over 24 years): Baseline Survivor Function 112 Relative Effects of Age in Instrumental Variables/Tobit Model...123 - ix -ACKNOWLEDGEMENTS I would like to thank Garry Barrett, Brian Copeland, Denise Doiron, David Green, Ashok Kotwal, Angela Redish and Craig Riddell for their helpful comments and suggestions. This work is dedicated to my wife, Anik, for her patience and understanding during the writing of this thesis. - x -I. INTRODUCTION. The objective of this thesis is to examine the impact of import shocks upon the employment patterns of Canadian workers during the late 19 8 0s. Increased foreign competition is often blamed for non-cyclical plant closures and layoffs, leading many in affected industries to call for assistance to either protect threatened industries or to retrain those permanently laid off. Governments have in the past been quick to respond: in the early 1980s the Canadian government set up the Canadian Industrial Renewal Board, which offered assistance to workers and businesses affected by foreign competition in the clothing and textiles industries. More recently the Canadian Steel Trades Employment Congress (CSTEC) was set up by business, labour and government to assist those displaced from the steel industry. A key question for policy is the subsequent labour market experience of these workers. If displaced workers from affected industries quickly obtain equally rewarding jobs elsewhere then there would appear to be little cause for concern. Similarly, if the labour market experience of those affected by import shocks is similar to that of those displaced for other reasons, then there seems no reason to single out the former group for special assistance. This thesis will examine these questions from both a theoretical and an empirical perspective, in the first place asking when might structural change be detrimental to workers and society as a whole; and in the second place asking what has been the actual experience of Canadian workers in trade affected industries during the late 1980s. - 1 -1. Outline of the Thesis. The thesis is organised as follows. The remainder of this chapter reviews the theoretical and empirical literature on worker displacement, with special reference to effects of import competition. In chapter two we present a theoretical model of worker displacement, in which import price shocks may cause workers to be laid off from one industry and re-employed elsewhere. In this model not only may workers be harmed by restructuring, but society as a whole may suffer if lower import prices lead to large layoffs. The model is dynamic, with three time periods and forward looking agents who must form expectations about the future of their industry. The model is expanded upon in chapter three, where we allow for the fact that workers may not be immediately successful in locating new jobs, leading to the possibility of unemployment. This allows us to generate predictions concerning the unemployment behaviour of those laid off in different industries. We also examine employment duration, comparing the behaviour of those who quit and those who are laid off. The emphasis in this chapter is upon adapting the theoretical model in order to provide observable predictions that can be compared to actual data. Chapter four marks the beginning of the empirical portion of the thesis. In this chapter we examine the trade and employment performance of Canadian mining, manufacturing and forestry industries in recent years, seeking to categorise the individual 2-digit industries according to the extent of import competition and by whether employment seems to be on the decline. To decide whether a worker was affected by a import shock we will use a relatively simple technique: analysis of changes in (weighted) import penetration ratios. This avoids the task of estimating - 2 -industry level structural models for employment: our intent is not to estimate the precise magnitude of import shocks but rather to identify those workers most likely to have been adversely affected by foreign competition. This exercise yields four different groups of industries. The empirical work contrasts the behaviour of those in the different groups. The next chapter describes the micro data set to be used. We use a longitudinal data set - Statistics Canada's Labour Market Activity Survey (LMAS). The time frame of the LMAS is 1986-90, which is useful for our purposes because it covers a period of general expansion in manufacturing employment (excepting 1990), and thus simplifies the task of discriminating between cyclical and structural unemployment. Individuals are assigned to one of the four industry groups based upon the industry in which they were formerly employed. We then discuss the characteristics of these four samples as well as the variables used in subsequent analyses. The sixth chapter is an empirical analysis of employment duration and flows out of employment. We examine the relationship between wage and whether or not a worker quits or is laid off. This requires both an empirical model of earnings and an empirical model of job tenure. Chapter seven also uses duration models; however, in this case we examine non-employment duration and flows out of non-employment. In this chapter we are concerned with not only testing the predictions of the theoretical model, but also in generating unbiased estimates of how long workers from the different industries can expect to be out of work. Chapter eight presents a model of the re-employment wage. The empirical specification uses a Tobit formulation which allows for the effects of non-employment duration upon the re-employment wage. - 3 -Estimating this model allows us to construct unbiased estimates of the wage a displaced worker can expect in a new job. Chapter nine examines the extent to which workers change industry. A probit model is estimated in order to determine what factors influence the decision to move from one sector to another. Finally in the last chapter we offer conclusions and directions for further research. - 4 -2. Theoretical Literature. In any discussion of displaced workers the central issue is: how big are the losses suffered by those who lose their jobs? In simple neo-classical models with homogeneous workers and perfect information these losses would be zero: the operation of the labour market would ensure that workers were indifferent between their current job and their next best alternative, and market clearing and perfect information about job opportunities would eliminate the possibility of unemployment. In more sophisticated models of the labour market however, employees can incur significant costs when displaced from their job. One scenario is that portrayed in standard search theory models, in which locating job offers is a time consuming process - all the more so if the worker has specialised skills or seeks certain characteristics (e.g. location or conditions) in the new job. Indeed if a worker had acquired firm specific human capital in their previous job then even after lengthy search their earnings would likely be lower when hired by another firm for which they do not have specific capital. Another situation when one would expect large displacement losses would be when workers are earning wage premia due to the influence of their unions or because of "insider" pressure1. Displacement raises the possibility of being forced to accept employment that does not offer these rents, or at least a long stretch of "queuing" or "wait" unemployment. 1 See Lindbeck and Snower (1988) for an elaboration of the latter scenario. - 5 -However the aforementioned scenarios have a common weakness: they beg the question of why, if displacement is expected to involve significant costs, do workers not offer to take a pay cut in order to retain their existing jobs? This question is crucially important for policy. For example consider an industry employing only labour that is in a zero profit equilibrium. Suppose that there is a small fall in the price of competing imports, so that the firm cannot remain viable without reducing wage costs. If workers in the affected industry accept a small pay cut in order to retain their jobs, which would be rational if their new wage is still higher than their best alternative, then employment and thus production is unchanged. Hence GNP cannot be lower and indeed will be higher since more imports can now be consumed. Consequently from a social perspective there can be no case for policies such as tariffs which try to restrain the fall in import prices. If however employment does fall, and if the workers so displaced do not immediately find equally productive employment elsewhere then real GNP may fall and there is a possible case for active policy measures. Thus the question is: under what circumstances would one observe workers who are earning rents (i.e. their wage is better than their next best alternative) losing their jobs rather than offering (or-being offered) the option of a pay cut? One way of answering this question is to allow for the possibility of moral hazard in the operation of the labour market. The assumption that workers' effort is not costlessly observable by the employer leads to the well known efficiency wage models. In such models the employees in a particular industry are permitted to earn rents in order that dismissal be an effective threat against possible shirking. Therefore an adverse shift in the labour - 6 -demand curve will lead to layoffs even though the workers laid off would be willing to see their wages fall rather than be forced into unemployment or a lower paying job. Thus, as demonstrated in Bulow and Summers (1986), there is the possibility that a drop in import prices can reduce real GNP. Another approach is to assume the existence of adverse selection in the labour market due to asymmetric information. In this paradigm, pioneered by Weiss (1980), employees have abilities that are heterogeneous and only partially observable. This implies that reservation wages will differ amongst workers according to their productivity, and thus if wages fall the best workers may leave. Since ability is partially hidden the firm cannot impose the pay cut upon only those workers who are earning significant rents. In the next section we discuss a model similar to that of Weiss, adapted into the international trade context. A different asymmetric information model of layoffs is that of Gibbons and Katz (1991). In this paper it is assumed that a worker's ability is known to the worker, and after a period of time the firm with which they are employed. However other firms who may wish to bid for the worker's services do not know the worker's ability, and so base-their offers upon whether or not the worker's original firm is willing to retain the worker. The main implication of the paper is that workers laid off due to a plant closure will be expected to be of higher quality than those that are otherwise laid off or quit. Empirical work is presented which tends to confirm the hypotheses of the model. Lawrence and Lawrence (1985) examine the question, of why relative wages rose for union members in those industries in the U.S. facing - 7 -severe employment decline (e.g. steel, autos). In this case the cause of rents is clear, but the puzzle is similar: why did unions in these industries not try to protect the jobs of their members at the price of lower wages? The explanation advanced by the authors is that with low substitutability of labour and capital, an exogenous reduction in demand increases capital intensity and therefore reduces the employment cost of a given wage. - 8 -3. Empirical Literature. There are numerous studies that examine worker displacement in the general case. Podgursky and Swain (1987) analysed the U.S. displaced workers survey (D.W.S.), and found that those workers who appeared to have large specific human capital investments suffered "large and enduring" job losses. The length of post displacement joblessness was shown by Addison and Portugal (1989) to be positively correlated with earnings losses: however they also found that the losses from obsolescence of firm specific human capital were overstated without joint estimation of unemployment duration and earnings changes. Hamermesh (1987) studied the costs of displacement using the Panel Study of Income Dynamics (PSID) . He found that displacement cost to the average worker was about US$7000 (1980 dollars) in lost firm specific capital. Another recent study on displacement is by Jacobson, LaLonde and Sullivan (1993), who found that laid off workers with long job tenure underwent an earnings drop of 25 percent, on average, and that typically the fall in relative earnings began even before the worker's separation from the company. This pre-displacement effect on earnings is echoed in a U.K. study by Blanchflower (1991) and in de la Rica (1992) who uses the 1986 U.S. D.W.S. Ruhm (1991) finds that after four years the U.S. workers in his sample (the PSID) were earning 10-13% less than their employed counterparts. Carrington and Zaman (1994) also use the U.S. displaced worker surveys, in order to determine to what extent the wage loss from displacement is influenced by the demographic characteristics of the workers involved, and to what extent industry specific factors are - 9 -responsible. They find considerable variation across industries in post-displacement behaviour that cannot be explained by demographic variables. Not only the degree of wage loss but also the relationship between this variable and an agent's labour market experience and job tenure varies across industries, and the authors are unable to explain these latter variations as being due to firm size, union density or other industry characteristics. Farber (1993) examines the incidence and severity of displacement over the period 1982-91, again using the U.S. D.W.S. He finds that there is a tendency for more educated workers and those in the service industries to be more likely to be displaced over time. However in general the costs of displacement display no significant secular trend. Never the less, these costs were found to be substantial with wages falling by more than ten percent on average. Another study that uses the American D.W.S. is Kletzer (1992). She examines both the incidence and duration of post-displacement joblessness, and finds that industry affiliation was an important determinant of joblessness. One explanation that she examined was the possibility that high wages are the underlying factor behind the result: high wage workers are perhaps more likely to wait for jobs. However workers displaced from some high wage industries such as communications had shorter jobless durations. Canadian studies of displacement have typically used Statistics Canada's displaced workers survey, which covers the period 1981-85. Gray and Grenier (1992) analyse unemployment duration using a Weibull distribution model. They find that older, less educated workers tended to - 10 -have longer expected unemployment durations: results similar to those of the U.S. surveys. Another study that uses the Canadian D.W.S. is Doiron (1993). The author examines Gibbons and Katz's model of layoff costs and finds a signalling effect of layoffs, but only for white collar occupations. The advantage of the LMAS over the D.W.S. are several. Firstly the LMAS is more recent, and covers a period when the relative importance of trade pressures was perhaps greater. Second the LMAS was conducted annually whereas the D.W.S. was a five year retrospective survey. The means that respondents to the LMAS are less to prone to recall bias. Thirdly the LMAS has more detailed information about industry. Finally, in the case of more than one displacement the D.W.S. provides information only on the job with the longest tenure. The main advantage of the D.W.S. however is the longer time frame, which results in much less right censoring of the data due to incomplete unemployment spells. Jones and Kuhn (1992) examine the effects of mandatory notice of closures upon the subsequent unemployment experience of workers, using a survey conducted by the Ontario Ministry of Labour. They find that there is considerable variation in the experience of workers, and that their "results are consistent with a scenario in which a minority of workers in all firms are capable of finding new jobs very quickly after starting to search, while the rest experience quite long durations." A paper that uses micro data to estimate the impact of import shocks is by Kruse (1988), who uses the U.S. displaced workers survey to estimate a Weibull duration model for post-displacement joblessness. - 11 -Included in the list of covariates are the changes in import share and exchange rates in the period of the displacement, along with the long term changes in these variables between 1972-4 and 1982-4. He finds that increases in long term import share lead to a significant increase in unemployment duration, although exchange rate changes had no discernible effects. This result held even when long term employment change was included as a regressor. About half of the negative effect of import change was found to be due to the demographic characteristics of the workforce in the affected industries. Our study differs from Kruse's in three ways. First we use a more flexible specification for the duration model. Secondly, instead of including import change as a regressor we categorise industries according to whether they were trade affected and by whether they are declining or not. The advantage of this approach is that it permits the parameter values to differ across industry groups; the disadvantage is that we do not obtain elasticity estimates for the effect of import share changes upon employment. Our focus is upon the behaviour of workers given that they were affected by trade, rather than on the extent to which industries have suffered or gained from changes in their trading environment. The third major difference with Kruse's analysis is that we examine not just the duration of unemployment but also the wage of a worker once re-employed. This is not easy to model using the U.S. displaced worker survey, since one has information on the current rather than the re-employment wage^. 2 Other drawbacks to the U.S. displaced workers survey include the lack of information on union status and the fact that total weeks of displacement are top-coded at 99 weeks. - 12 -Another perspective on the effects of international competition upon employment and wages in the context of unionised industries is provided by Abowd and Lemieux (1991). Using aggregated data the authors estimate an empirical model for Canada and the United States of America relating real wages and employment to export and import penetration for various industries. The estimation period for Canada is 1968-1983, and attention is restricted to unionised workers. The authors find that "For bargaining units in both countries, increased import penetration is associated with very large [negative] employment effects." (p.360) . In contrast to our paper, the emphasis in Abowd and Lemieux's article is upon quantifying the effects of import changes on industry employment, rather than on the subsequent labour market experience of those workers laid off. As noted above, the implications for policy depend heavily upon what happens to workers after they leave an industry. Since the authors are not using data upon individuals, it is difficult for them to address issues of this kind. Glenday, Jenkins and Evans (1982) provide a Canadian perspective upon worker displacement due to trade pressures. The authors examine the town of Sherbrooke, which is heavily dependent on the clothing and textile industries. The period of the survey is 1974-6, which covers a period of increasing import penetration in these sectors. Estimates of mean unemployment duration varied from six months for a young single male to 17.5 months for a married forty year old female. 3 The authors of this study appear to be using data on completed spells of unemployment only: they do not however make clear what steps they have taken to avoid the well known problems of bias that occur when - 13 -To finish this section mention should be made of a paper by Osberg (1988). As we stated in the introduction (p.3) an advantage of our data compared to the D.W.S. is that it avoids the recession of 1982. We claim that the effects of restructuring are likely to be relatively more important during cyclical upturns. This view is supported by Osberg's paper, which finds a "chilling" effect of recessions upon inter-industry labour mobility. uncompleted spells are not included in the analysis. - 14 -II. THEORETICAL MODEL - FINITE TIME, NO UNEMPLOYMENT. In this chapter we wish to examine the following questions: (1) When a factory is forced to close, will the employees be able to find equally good jobs elsewhere, or will they be consigned permanently to less desirable occupations? (2) If workers suffer from being relocated when a plant closes down, why do they not offer to take a pay cut in order to save their jobs? (3) Even if employees in affected industries suffer, does this pose a problem for society as a whole? (4) Do the losses that employees might undergo result merely from poor foresight, or would society be worse off even if workers had perfect foresight? (5) Should the government respond to economic restructuring by merely providing adjustment assistance to affected workers, or should it attempt to keep embattled industries in operation? Prom the perspective of theoretical economic modelling, -these questions raise additional issues. For example: what is meant by a "good" job? How does one model foreign competition? How can the process of retraining be modelled? The models presented below assume that the economy is small and open. Thus prices are exogenously determined. Intensified foreign competition is therefore proxied by a fall in one of these prices, which will in turn will put pressure on wages. Workers in that industry will then have to decide whether to accept a fall in the - 15 -earnings or move to another industry. They may have no choice in the matter: if enough of their colleagues leave in response to a pay cut the plant may become too inefficient and they will have to leave anyway. The decision to enter another sector is based upon a training cost which varies across industries and (in manner unobservable to other agents) across workers. A good job in this context is a job in which an employee earns more than their next best alternative: if they are laid off their utility will fall by a significant amount as they are forced to retrain or they become unemployed. Thus a good job is not necessarily one of the economy's higher paid jobs: it is one that is high paid relative to the other options of the worker concerned. There are a number of sub-issues to be settled: a) Sunk vs. Recurrent Costs of Training: Clearly whether or not training costs must be paid in the current period will affect employee's decisions: we shall examine both possibilities. b) Fixed Costs vs. Minimum Efficient Scale: In order to model plant closures when some workers are earning rents one requires some form of increasing returns to scale, otherwise workers will just set up their own firms and continue as before1. One way of doing this would be to assume fixed costs: as employment falls average product drops until the f-irm is forced to close down. However we shall assume that production function exhibits constant returns to scale for levels of employment greater than some minimum efficient scale: below this the firm must shut down. The advantage of this is that it ensures constant average product above the shutdown point ensuring that standard gains from trade theorems will 1The importance of increasing returns to scale in explaining unemployment is demonstrated in Weitzman (1982) 16 apply for this region: this allows us to isolate the effect of plant closures from the effects of trade on productivity. c) Division of the Surplus: The models below assume free entry. Thus firms earn no rents, and the employees absorb the entire value of the firm. This is a simplifying assumption: this paper is concerned with employee surplus rather than producer surplus. We doubt that that our results would change qualitatively if entry were restricted and firms were able to share in worker's rents (if human capital were firm specific for example): with unobserved worker heterogeneity employees would still be able to extract a rent from their private information. In the rest of this chapter we develop a theoretical model of plant closure and worker retraining. First we deal with a model in which workers are already endowed with skills, and secondly we examine the case where all workers are initially unskilled. It is shown that: a plant may close even if almost all workers would be willing to take a pay cut that would keep it open; that the losses to workers from a shutdown can outweigh the gains to consumers of lower import prices; that if the price fall is foreseen it can actually make matters worse; and that the government might increase welfare by protecting domestic industry but reduce welfare by providing retraining assistance. - 17 -1. One Period Model when Workers are already Skilled. In this section it will be assumed that all workers have been trained in some particular skill: however if they are forced to move jobs they will have to pay a retraining cost. Discussion of how these skills were obtained will be postponed until section III. 1. Initial Conditions. Workers are indexed by their aptitude for training X and are uniformly distributed on [0,1/]. Note that a worker's type is private information. There are two technologies available: each produces good i according to r OCLi ; Li > Li ai = -j i = A,B. L 0 ; o.w. Thus there is a minimum efficient scale for the production of each good. Above this minimum the technology exhibits constant returns to scale. In order to be employed in an industry a worker must be trained: the cost of this training depends both on the industry and the individual and is given by the function fci (X) , i = A,B. Each worker is assumed to be already endowed with the training for one of the two industries: we denote the already skilled labour force in each industry as Li. It is assumed that agents are ordered in each set so that the higher is X the lower is the cost of training for a job in the other sector. Note that training is assumed to be industry and not firm specific. One way of thinking about the training cost is to assume that when a worker switches sectors they must spend a certain proportion of - 18 -time learning the job, during which time they receive no wages. Once they are trained they then earn the full wage. The training cost ts (X) is equal to the percentage of time spent training multiplied by the wage. The workers take-home wage is then wi — fcj (X) . The economy is assumed to be small and open: thus there is a fixed world price p± for each good. Free entry is assumed: hence profits will be driven to zero. Note that since prices are fixed the elimination of profits will occur through competition for workers. Finally, consumers have homothetic preferences over the two goods, and there is an outside option of unemployment for those who do not wish to work in either sector. This option pays a wage of w. 2. Equilibrium. If a firm operates at all the free entry assumption will drive wages up to the point where w% = apt, i = A,B. Incumbent skilled workers with firms in industry A will choose to remain so long as their wage exceeds the wage in sector B (less the cost of training) and the value of the outside option. wk > wB - tB(X) , w^ > w If the firm wishes to hire workers who are not skilled in industry A then it will have to ensure that wk - tA(x) > wB. - 19 -Assume without loss of generality that initially wA > wB. We will further restrict our attention to the case where wB > w and JOj > L-x. Clearly all those workers who are already trained in the skills demanded by sector A will wish to remain there. The behaviour of workers endowed with sector B skills will depend upon how much it will cost them to switch sectors. It may be that fcA(X) is large enough that w3 > wA - tA(X) V X G LB, in which case no workers will want to shift. If however some workers find it worth their while to switch, employment will fall until the marginal worker is indifferent to switching (recall tA is continuous) : thus employment in sector B will fall until WB = wk ' tA(LB) . These two possibilities are illustrated in figure 1 below. In the first case (t = t') the cost of retraining for sector A is low enough that some workers (LB - Lg') find it profitable to move sectors and employment in B falls. In the other case retraining is so expensive that no workers endowed with sector B skills wish to move. The crucial difference for our purposes is that in the second case all sector B employees are earning rents, that is more than their opportunity cost, whilst in the first case the marginal worker is earning the same wage as their next best alternative. - 20 -Figure 1: Labour Supply to the lower wage Skilled Industry. wk - t l c n ^ — - ~ ~ — ^ ~ — ~ ^ ^ — — ^ Z ^ f ^ ) 'B LB ' L B Another way of highlighting the differences between the two cases is to observe what happens when there is a small drop in wB. If all workers are earning rents then nobody will wish to leave if the fall in wages is small enough. However, if one worker is indifferent to leaving the industry then the smallest fall in wB will prompt that agent to quit, thus reducing the workforce in industry B. Note the importance of the free entry assumption: if there was one monopsonistic firm in this sector one would never observe the case where all workers were earning rents, since the firm would reduce its workers' pay until some employees were on the verge of leaving. However without knowledge of its employees' types the firm would not be able to extract all the workers' rents. 3. Comparative Statics. The variable of interest is P\. We wish to gauge the effect on welfare of a small drop in P;. Social welfare is identified with real GNP V = [wALk + wBLB + w(L - LA - LB) ] IT - 21 -where P is the exact price index2. An increase in V denotes a potential Pareto improvement. Proposition One: If 3 X e LB s.t. wB = wA — tA(T), then a small enough fall in PB will, ceteris paribus, reduce welfare for LB close enough to Lg. To understand why this proposition holds, consider the case where L-Q = L. Here the departure of a single worker will force the firm to close. Let PB fall by a small amount. Then the firm must either persuade its workers to take a pay cut or go out of business. However if one worker is indifferent to remaining at the firm then this employee will have to be exempted from the wage reduction or else they will leave. But since X is unobservable a concession offered to one must be offered to all, which is impossible if the industry is to remain viable. Thus the firm exits, and although the marginal worker does not lose from having to relocate sectors the infra-marginal workers will suffer a discrete drop in their utility. Since L is positive the total utility loss to workers in this sector is bounded away from zero - unlike the total gain to consumers from lower prices, which can be made arbitrarily small"for a small enough fall in PB. A key point to note is that it is relative wage movements that are important, as the next proposition makes clear: 2Recall that preference were assumed to be homothetic, thus such a index exists. - 22 -Proposition Two: If PA and P# fall so that the wage difference between the two sectors is unchanged, then the conclusions of proposition one will not hold. The intuition behind this result is that with the wage differential unchanged both firms can pass through the price cut to wages without any of their workers switching to the other sector. How important is the minimum efficient scale assumption? This is answered as follows: Proposition Three: If LB = 0, then a fall in the price of a good improves welfare if it improves the terms of trade. What does the above model say about government retraining programmes designed to cushion the effect of plant closures? We consider a scheme that provides free training in sector A skills to sector S workers: Proposition Four: Suppose sector B closes down and the government pays for all the affected workers to be retrained. If w > w& - t^ix) for some laid off workers then the programme reduces social welfare. Otherwise the policy has no effect on welfare. This reason for this result is simple. If w < wk - t^iT) for all the laid off employees, they would all retrain for sector A anyway, and thus the programme is merely redistributive. If however some agents would - 23 -choose unemployment without the subsidy, then the programme generates a social loss since the training costs are greater than net wage increase. Since sector A displays constant returns to scale for LA > LA, there is no social benefit from artificially expanding this sector if it is already in operation. - 24 -III. A Two Period Model. In this section we consider a model with two periods. In the first period, agents choose which sector to enter based upon their aptitude for acquiring the relevant skills. Then in the second period agents can choose if they wish to retrain and change sectors, as in section II. Let the two periods be labelled 1 and 2. The utility function of each agent is now 1 2 ,1,1. .2,2. U = u + u = u(a ,b ) + u(a ,b ) , where u(&,&) is homothetic as before. Agents are initially unskilled: at the start of each period they may train for either industry. In doing so a worker must make a prediction about future wage levels and employment opportunities. Figure two below illustrates the payoffs to each possible strategy: Figure 2: Payoffs associated with the possible strategies of each worker. 1 1 2 ^A " t A + WA 1 1 2 2 wB - tB + wA - _tA 1 1 2 wB - t B + WB 1 1 — WB ' tB + W w + w To simplify the analysis ' it will be assumed that prices are such that both sectors are initially viable and that wA > wB > w in both - 25 -periods. We are interested in the situation where a drop in PB forces this sector to close down. The equilibrium will- differ according to whether the price change is anticipated at the start of period one or unanticipated. The first case is the "baseline" equilibrium: 1. Equilibrium with no Price Changes. Since prices and thus wages are the same in both periods, no agent would wish to switch sectors at the start of period two. The reason for this is simple: if it is worthwhile to switch at this point then it was certainly worthwhile to enter the sector at the start of the first period. 2. Equilibrium with an Unanticipated Price Change. We now assume that there is a drop in PB at the start of period two, and that this fall is not anticipated by agents at the start of period one when the initial training decision is made. Then the change in social welfare will depend upon the benefit to consumers of lower prices and the loss to workers in sector B who must either take a wage cut or lose their jobs. The first result is that if the fall in PB is small enough all workers will consent to a small pay cut. Proposition III.2.1: If APg is small enough then A L B = 0 and thus industry B remains viable. Thus the fall in PB will increase welfare if it improves the terms of trade. - 26 -Proof: Let X e LB be the worker who earns the least rents. Since X chose - to enter sector B in period one it must be that — 1 1 ~ 1 — 2 wB s wB - fcB(T) > wA = wA - tA(T) . Thus wB > wA which implies that wB > wk for wB close enough to wB. The intuition behind this result is simple. If a worker is indifferent between sectors before paying the training cost then then ceteris paribus he cannot be indifferent once the training cost for one of the sectors is sunk and thus effectively zero. Thus in order for a plant shut down to occur the fall in PB must be large enough. The change in welfare then depends upon weighing up the (now non-trivial) positive terms of trade effect against the negative employee surplus effect. It is still quite possible for welfare to fall: Proposition III.2.2: 2 If PB falls by a large enough increment then industry B will no longer be viable. This shutdown can reduce welfare even if the terms of trade improve. Proof: By Example. See the Appendix. 3. Equilibrium with a Perfectly Anticipated Price Change. Suppose now that agents know that sector B, although viable in the first period, will not be operating in period 2 because of a fall in PB. Clearly this will result in fewer workers choosing to enter sector B in the first period, resulting in fewer workers needing to relocate in the second period. However one also has to take into account the possibility - 27 -that so few workers will now enter sector B that this industry will not be viable even in period one. The first question to be answered is whether an anticipated price change can reduce welfare: Proposition III.3.1: 2 A fall in PB can reduce welfare even if it is perfectly anticipated and it increases the terms of trade. This is because the fact that a price change was foreseen does not change the fact that some workers are denied the possibility of putting their skills to the best possible use. Note that unlike the unanticipated case the losses relative to the baseline case are spread out over the two periods. The next proposition deals with the importance of being able to foresee price changes: Proposition III.3.2: It is possible for an anticipated price fall to reduce welfare more than an unanticipated fall if it results in sector B not being viaBle in either period. The logic behind this result is that some workers may have such a low aptitude for sector A work that they would rather enter unemployment. Thus their welfare will fall if they are denied the opportunity of entering sector B even for one period. - 28 -IV. Conclusions. The results of this chapter are in one sense inconclusive: we show that acquiescence in the loss of industries to trade pressures may not be optimal, even when anticipated. However one cannot make blanket statements either way. Policy makers need to consider the costs of retraining, the relative attractiveness of unemployment or unskilled jobs and the welfare benefits of lower import prices before definite pronouncements can be made. Much depends upon the social rate of time preference: retraining costs are short term whereas the benefits from permanently reduced import costs are long term. In another sense however we can come to a conclusion: one cannot simply assume that all labour market adjustment problems can be taken care of by retraining programmes3. It may not be socially optimal to push workers into occupations for which they have little aptitude, nor can one assume that costs of readjustment to trade pressures are only short term. It is not necessarily the case that lower import prices can make up for the loss of skilled jobs. An interesting feature of the model exposited in this chapter is what is implied about the dynamic of structural change. The first "people to leave a declining industry are those whose next best alternative - be it either alternative employment or withdrawal from the labour market -is relatively attractive. These people will lose little from restructuring. However as relative wages in the industry fall further behind those paid elsewhere, even those with fairly high mobility costs 3 For an exposition of the view that trade policy towards threatened industries should be limited to retraining programmes, see Economic Council of Canada (1988). - 29 -may wish to leave, until firms are no longer viable. This will result in the layoff of a hard core of workers whose alternative value of time is very low. We would expect these workers to undergo long periods of unemployment, and to suffer a considerable drop in their wage once re-employed. Were this not the case, these workers would have left the industry beforehand. Thus there are likely to be considerable differences between those who quit early on and those who are pushed out through layoffs. We explore this idea empirically in chapter VI. Note however that even those who quit from an industry voluntarily may be thought of as being harmed by trade, because their decision to leave is forced upon them by the pattern of declining relative wages. The model outlined above is strictly only applicable to the case where there is only one firm left in the skilled industry that is affected by import competition. Otherwise workers laid off from one firm would find their existing skills valuable elsewhere. However the model can easily be generalised to a situation where firms producing skilled labour goods operate upon a continuum of goods, with each specific type of good requiring a specific type of labour. The distinction between firm-specific and industry-specific human capital would then vanish: each firm is in its own industry, and the important question then becomes, how far is the firm on the skills continuum from other firms? Of course the answer may be different for each worker, and if this is partially unobservable then it seems probable that our results carry through largely unchanged. A weakness of the model is the assumption workers -are able to find new jobs instantaneously, which precludes the possibility of unemployed - 30 -job search. This means that the model has little to say about the non-employment behaviour of workers in different industries. This omission is rectified in the next chapter, when we adapt the model in order to generate richer predictions about the labour market behaviour of displaced workers. - 31 -Appendix. In this appendix we simulate the two period model outlined in section III in order to prove the various propositions outlined in that section. The size of the labour force is normalised to 1. The productivity parameter a is also set to 1, thus the wage is equal to the price of the good produced. Preferences are assumed throughout to be Cobb-Douglas with parameter t. Proof of Proposition III.2.2; We need to show that an unexpected reduction in price of good B can reduce welfare even though it increases the terms of trade. The training cost functions are: t (X) = X2 t (T) = ( T ' ° - 5 ) 2 C A I U l ' LBVI-' 0.175 these expressions are illustrated in figure 3 below. Case One: pi - 0.69, industry B shuts down. Time 1 2 sum H 0.50 0.85 ^B 0.50 0.00 h) 0.00 0.15 ^min 0.32 0.32 tA tB WA WB E ( W B ) W y 0.14 0.09 1.00 1.00 1.00 0.50.76 0.14 0.00 1.00 0.69 1.00 0.50.94 1.70 Time 1 2 sum Case 0.50 0.67 Two: p| ^B 0.50 0.33 = 0.72, I'll 0.00 0.00 industry B stays viable. Anin tA tB wA wB E(wB) 0.32 0.14 0.09 1.00 1.00 1.00 0.32 0.03 0.00 1.00 0.72 1.00 w y 0.5 0.76 0.5 1.03 1.79 - 32 -This example proves the proposition: in case one the price of B falls- so far that the industry must shut down, forcing some workers to engage in costly retraining and others to choose unemployment. In case two employment falls in industry B, but not enough to precipitate a shutdown: thus although the affected workers are worse off, real GNP for the entire economy is higher compared to case one. This is despite the fact that when industry B is viable the economy is a net importer of good B. Proof of Proposition III.3.1: In this instance the fall in pB is known before agents make their decisions about which industry to enter. The training cost functions are assumed to be tA(X) = 3X2, t B ( T ) = (T-Q0-34)2, Upon simulating the model one obtains: Case Three: pi 0.64, industry B is never viable. T i m e 1 2 sum LA 0 . 5 1 0 . 5 1 LB 0 . 0 0 0 . 0 0 ^u 0 . 4 9 0 . 4 9 An in 0 . 4 2 0 . 4 2 tA t B wA wB B(wB) 0 . 1 4 0 . 0 0 1 . 0 0 1 . 0 0 1 . 0 0 0 . 0 0 0 . 0 0 1 . 0 0 0 . 6 4 0 . 6 4 w y ore 0.67 0 . 6 1 . 0 1 1 . 6 7 Case Four: p | = 0.65, industry B is viable in both periods. Time LA LB Lu L m i n fcA fcB wA wB E ( w B ) w V 0 . 0 4 0 . 0 5 1 . 0 0 1 . 0 0 1 . 0 0 0 . 6 0 . 8 1 0 . 0 0 0 . 0 0 1 . 0 0 0 . 6 5 0 . 6 5 0 . 6 0 . 9 4 1.75 1 2 0 . 3 4 0 . 3 4 0 . 4 2 0 . 4 2 0 . 2 4 0 . 2 4 0 . 4 2 0 . 4 2 sum - 33 -Proof of Proposition III.3.2: The objective now is to show that an anticipated price decrease can reduce welfare more than an unanticipated price decrease. Let all variables save E(wB) be the same as for Case Four above, then one has: Case Five: p§ = 0.65, industry B shuts down. Time LA LB Lu L m i n fcA tB wA wg E(wB) w V 1 0.50 0.50 0.00 0.40 0.14 0.09 1.00 1.00 1.00 0.50.76 2 0.85 0.00 0.15 0.40 0.10 0.00 1.00 0.76 1.00 0.50.94 sum 1.75 - 34 -Ill: THEORETICAL MODEL - INFINITE HORIZON, UNEMPLOYMENT. The purpose of this chapter is to develop the theory in the previous chapter into an adequate empirical model of the labour market. Despite the presence of adverse selection in the previous model, there was no unemployment per se: upon being laid off agents immediately found work elsewhere, albeit in possibly less desirable jobs. However in reality agents are not fully informed of job prospects, and a period of search will be required before a laid off worker will find what they consider to be suitable employment. This form of unemployment is a cost to society, and for that reason alone it is important to estimate its duration. However another important motivation is that unemployment duration is an observable quantity that is directly related to some unobservable features of the model (such as unobserved ability), and thus incorporating it into our empirical work may help us evaluate the validity of our theory. Section one outlines the basic model, a stationary search model in which agents are assumed to have been laid off from a particular industry and are assumed to face different costs of mobility when moving to another industry for which they are not skilled. As in the previous chapter we shall assume that there are three industries: two skilled (A and B) and one unskilled (C). Our focus is upon contrasting the behaviour of workers laid off in skilled industries, one of which (A) is assumed to be higher paying than the other. Note that we assume that the arrival rate of job offers is independent of labour market status. This has the effect of simplifying considerably the model. An empirical justification - 35 -for this assumption is provided by Blau (1986, quoted in Mortensen 1990; see also Blau and Robins 1990). Section two introduces a non-stationarity into the model: unemployment benefits. We examine the relative effects of benefit exhaustion upon workers laid off in the different industries. 1. Effect of former industry upon Non-Employment duration of laid off workers. In this section we shall examine the hazard rate out of non-employment for those who are laid off in industry A or industry B. Subsequently we shall wish to examine the non-employment behaviour of those who quit their jobs: this however will involve modelling the quit/layoff process, which we postpone for the moment. Individuals are assumed to be infinitely lived and to maximise expected lifetime wealth subject to the discount rate p. Each period, whether employed or not, the agent may receive a wage offer. The arrival rate of these wage offers is modelled as a Poisson process with arrival rate X. There is assumed to be a single, constant wage in each of the two other sectors, and thus the density of the wage distribution is prob(W = w*) = yA, prob (W = w8) = yB, prob(JV = w0) = yc = (1 - yA - YB> • One has w^ > w8 > w0 by assumption. In order for an agent from one sector to take a job from the other skilled sector they must pay a moving cost M e {0,/n}, where prob (M = m) = [X. - 36 -The utility received whilst non-employed is B e {jbH,bM,bL}, and conditional on being laid off in sector j the density of B is prob (B = foH) = %J, prob (B = hM) = TCMJ, prob(B = bL) = 7lLJ = 1 - 7THJ " ^ Mj-We also assume w A > j b H > w B > J b M > w c > j b L . In view of the fact that agents were involuntarily laid off it seems reasonable to assume u^ > bH: that is, all agents would prefer to be re-employed in their own sector rather than be non-employed for a given period. However it seems reasonable that some though not all agents may prefer non-employment to jobs lower paying than those previously held: thus we assume w0 > bL. It is important to bear in mind that agents may accept an unskilled job whilst continuing to search for skilled jobs. This eliminates the possibility of "wait" unemployment, whereby workers turn down jobs which have higher utility than non-employment in the hope of something better turning up. The involuntary layoff condition in conjunction with the time invariance of wages and non-employment utilities also means that 7lHB = 0. This follows because if sector B workers had high values of leisure they would not have accepted jobs in that sector in the first place (recall ±>n > w8) . We shall however assume that relative proportions of B = b« and B = bL workers are the same in both industries. This implies that ^ = i[W > ^  ' (1) - 37 -(2) Finally we simplify matters somewhat by postulating a zero probability of layoffs in the new job. Note that the assumptions we have made thus far are sufficient to render the model stationary. The present value of employment (net of moving costs) at a wage w for an arbitrarily small period of time h is VE(w,mA,mB) = - i w + (1-Xh) VE(w) + Xh Ew(max{ VE (w) , VE (W) }) \ + o (h) , where mi i s the cos t to the agent of moving to s ec to r j . In the l i m i t as (p+X) VE(w,mA,mB) = w+X Ew(max{VE(iv) , VE(W) }) . (3) The e x p e c t e d v a l u e i n t h i s e x p r e s s i o n i s g i v e n by Ew(max{VE(w) ,VE(W) }) = VE(w) + Ev,(vE(W)-VE(w) | VE{W) > VE(w)j P(vE(&0> VE(w)) E l - (wi-w) yJ [wi - p mi > w] , P (4) j=A,B,C where the term in square b r a c k e t s takes on the va lue of one if the express ion i s t r u e and zero o the rwise . Combining (3) and (4) y i e l d s VE(w,mA,ma) = — + X E l - yJ max{ (wJ - p mi - w) ,0} P j=A,B,C The p r e s e n t va lue of non-employment i s de r ived s imilar ly" , and i s given by Vn(b,mA,mB) A mB\ _ X + — Z l . - yJ max{ (wJ - p mi - b) ,0} P j = A , B , C A non-employed agent will accept an offer of employment if VE(wJ) - p mi > VM - 38 -=>wi-pmi-b+X - |Ew(max{VE(w) - p mi,VE(W)}) - Ew (max{ VN (b) , VE (W) }) 1 > 0 => wi - p mi > b. The hazard rate out of non-employment is the product of the arrival rate and the probability that the agent will accept the offer: 0(t,mA,mB,b) = X V yj [w* - p mi > b] . j=A,B,C The aggregate hazard function for those laid off in a given sector will depend upon the proportion of low mobility agents and the proportion of agents with high or low values of leisure. For agents of a given type the hazard function will be constant, however the aggregate hazard function will appear to exhibit duration dependence as the pool of non-employed becomes dominated over time by those with lower re-employment probabilities. The first result in this section is as follows: Proposition One. The aggregate hazard rate for those laid off from sector A will initially be lower than the aggregate hazard rate for those laid off from sector B if 7lHA > 0 an<^ YB i-s close enough to yA. Proof: To show that this result holds we examine the aggregate hazard function for both sectors in the first period after being laid off, 9-i(0), which is equal to the density function for t = 0. For those laid off in sector A this is 6A(0) =Pr(B=hH) A- yA + Pr(B=bM) (pr (M=m) X yA + Pr(M=0) X (yA+yB) j + Pr (B=bL) |Pr(M=m) X (yA+yc) + Pr(M=0) X (yA+yB+yc)J - 39 -= 7tHA A y A + ^ M A [ n A YA + (1-M-) A, (y A +y B ) l + reLA f n X (yA+y c) + ( I - J J . ) X (yA+YB+Yc)) = X fyA + ( l - j j . ) (7i„A+7iLA) yB + rcLA Y C ] • The corresponding express ion for those l a i d off in s e c t o r B i s 9 B ( 0 ) = KHB f|u A- Y B + (1-M-) A (YA+yB)l + TCLB m X (yB+Yc) + (l-fj.) A (yA+yB+yc) 1 = A f ( l - n ) yA + yB + TTLB YC1 . By a s s u m p t i o n , yA — YB- T n e d i f f e r e n c e be tween t h e two e x p r e s s i o n s i s 6A(0) - 9 B (0) = A f [ l - ( l - ( l ) ] YA + t( i- |J .) ( ^ M A + ^ L A ) - I ] YB + [TCLA-7tLB] Yc] • We c l a i m t h a t t h i s d i f f e r e n c e becomes n e g a t i v e a s yA K yB+- S e t t i n g yA yB , we h a v e 0 A ( O ) - 0 B ( O ) yA-r^yB A f-(i-n) KCA yB + [7iLA-7rLB] ycl < 0, because 7lLA-7CLB < 0 from (4) above. If the result holds at yA = yB then by continuity it also holds for a region around this point. Thus 0A(O) < 0B(O) for yB close enough to yA. •. The intuition behind the result is that because sector A contains some workers with high values of leisure, they will be unwilling to take some jobs that those laid off from sector B would be willing to- take. Thus as long as the frequency of job offers from sector B is not too low, the initial hazard rate out of non-employment may be higher for sector B agents. Proposition one does not hold true for the hazard functions at all durations. To show this we require the following lemma: - 40 -Lemma One. Suppose the aggregate hazard function for some population of N types at t = 0 is Q(0) = YKi ei-i = l where the proportion of each type in the population is K-ir and where the hazard function for each type, Qlf is constant. Then lim 0 ( t ) = min {Q^ , t_>00 i i.e. the aggregate hazard function tends to that of the type(s) with lowest exit probability. Proof: The aggregate hazard function at time t is the weighted sum of the individual hazard functions, 8(t) = [Pi 0i i=l where the weight (pj) is the proportion of that type in the original population multiplied by the probability of still being left in the sample: p. = _ i 1 . jTrij Fj(fc) j=i The survivor function is t - i F-x ( t ) = e x p • 5> exp{t&0 i}. L e t then 0 = min {0i>, - 41 -F*(t) = exp{-t&9*} . We can rewrite pi as %i Fi(t)/F*(t) Pi = £ %j Fj(t)/F*(t) j=l We will show that the numerator vanishes for types such that Gj > 9*: lim exp{-(0i-9*) t} t_»oo = 0 when (9j - 9*) > 0. = 1 when (Gj - 9*) = 0 . Consequently lim 9(t) = 9*. •. t^oo This result is illustrated in figure 3 below for industry B.1 Note how the average hazard rate for the four different types (the solid line) tends towards the lowest hazard rate. We are now in a position to prove the following Proposition Two. If yB is less than yA, the aggregate hazard rate for those laid off from sector A will eventually be greater than the aggregate hazard rate for those laid off from sector B. Proof: From Lemma One we know that aggregate hazard tends over time to that of the group with the lowest hazard. In the present case the lowest hazard for sector A workers is that for those who will only accept sector A jobs: X yA. For those laid off from sector B, the lowest hazard is that attached to workers who will only accept sector B employment: X yB. Thus 1The particular parameter values were A, = 0.25, m = 2, (l = KM = KL = 0.5, wA = 12, wB = 6, wc = 4, i>M = 5 and hL = 0. - 42 -n . Fj (t) lim =^ t_»co F*(fc) 9A —> ^ YA a n d 9B —> ^ YB- Because yA > JB, in the limit it must be that 9A* > 9B*. •. What these two propositions imply is that although the re-employment probability for those laid off in the depressed skilled industry may initially be greater than that of those laid off in the other skilled industry, this re-employment probability will be lower at longer durations of non-employment. Intuitively the latter situation is due to the fact that at longer durations the people we observe tend to be those who are reluctant to move industry, and so the hazard depends upon the frequency of job offers in that industry, which will be lower for those in the depressed sector. In the model above we have assumed a constant search intensity. In Fallick (1992) this assumption is relaxed. He shows that if workers can search in two sectors simultaneously, search intensity will be lower in the sector with the higher layoff rate. In our model we assume a degenerate wage distribution, and therefore search is irrelevant, which makes the model very much more tractable. Nevertheless, we think that incorporating the features of Fallick's model would strengthen our conclusions, because workers with low mobility in depressed industries would tend to search less intensively and would therefore have lower hazard rates out of joblessness. - 43 -2. Exhaustion of Unemployment Benefits. This section examines the relative effects on non-employment duration of the exhaustion of unemployment benefits. Typically (e.g. Mortensen 1990 or Lancaster 1990) the per week benefit is assumed to be constant across all eligible workers. However in reality unemployment insurance benefits generally depend upon the wage in the previous job. For example in Canada in the late 1980s the replacement rate was 60%, up to a certain maximum2. This point is important in the present context because it means that an unemployed worker's previous wage and thus industry affiliation may have an important bearing upon their subsequent labour market behaviour. A worker from a high paying industry will receive a much higher benefit in absolute terms than a worker from a low paying industry. Indeed for a worker earning close to the minimum wage there may be little difference in benefits from the UI scheme and benefits received after UI expires (i.e. welfare), whereas for a formerly high paid worker there may be a significant difference. In this case there will be a very great variation in the response of the agents' hazard rates to the exhaustion of UI benefits. Let the replacement ratio of UI benefits to the previous wage be denoted by (3 e (0,1), and let the welfare payment made to those who have exhausted their UI entitlement be b < wJ V j . We shall assume that P wi > b, j = A,B,C. i.e. that UI benefits are at least as high as welfare benefits for all workers. The value of time spent non-employed U varies over agents and industries, and is distributed according to 2 $339 per week in 1988. - 44 -u ~ FHU) . The variable is assumed to be non-negative. We adopt the following shorthand notation: 7lHJ =Pr{b+U>wB i previously employed in industry j } , 7lMJ = Pr{wB>b+U>wc\ previously employed in industry j }, "K\j =Pr{ vf > b + U I previously employed in industry j } . As in the previous section, we argue that KHB and 7lHc can be assumed to be zero, because if agents from these lower paying industries preferred leisure to work in their previous sector, they would not have accepted these jobs in the first place. Denote the point at which benefits run out as T. Then the value of time spent non-employed is a time varying function given by b(t) = P vA + U = b'i, t < T, wu + U = b, t > T. After T the solution to the agent's problem is as described in section one above, and is therefore stationary with N N N V = V = V V i > T. T+i T — The agent's value function before T will in general be time varying. One has N N (p+X) V Tti(U,mA,mB) = b + X Ew(max{V ,^(W)}), T — T N N (p+X.) V (U,mA,mB) = i>J + X Ew(max{V , V^W)}), T-l T N N (p+X) V (U,mA,mB) = JbJ + X Ew(max{V , V^iW)}) T-2 T-l ... etc. By induction one can show that if b(T) < bit) where t < T, then N N V > V V i > 0. T-i T-i-1 If £>J = b then Vto is constant. From section I we have - 45 -N N _ , 1 E w (max{V , Vs (W) }) = v + ) - yj max{ (wJ - p mi - b) , 0} T T " P ' — j = A , B , C and s o V" P P V - yj max{(wJ - p JIJJ - b) ,0} j=A,B,C V hJ T-i P+^ + X V yj max{ (wJ - p jnJ - i>J) , 0} j =A,B,C i-lf ^k A. p+X \. J I k=0 A. /• "1 X p+X V. J • 1-1 N V T . As the point at which benefits terminate T recedes into the future, one has -N N i.J X V = lim V = — + -T-i P P T-» 00 V yj max{ (wJ - p mJ - i>J) , 0} j =A,B,C In figure four below we illustrate the behaviour of VN over time. How will the non-stationarity of V affect the hazard function out of non-employment? The answer to this question depends upon the extent to N which the higher value of V before benefits run out induces agents to reject offers which they would be willing to accept after T. Those agents for whom N E V > V > V t — for some t > 0 will exhibit discontinuous offer acceptance behaviour over the length of their non-employment spell. The hazard function for a particular agent will be an upward step function, with the step points E N being the periods for which V (j) = V (j) for different industries j . The effect upon the observed hazard function for all agents separated from a 46 particular industry will depend upon relative proportions o£ each different type. Because of this it is difficult to make generalisations about the shape of the hazards before T, however we do expect that the response of agents separated from industry A to the change in benefits at T will be much more pronounced. In figure five we presents results from a simulation of the model.3 The average hazard functions for agents from both industries slope downwards over time due to self-selection out of non-employment, as predicted in the last section. This is due to the fact that over time those agents with the lowest costs of moving and lowest value of leisure will tend to become under-represented in the pool of non-employed. However for industry A there is a pronounced spike around 25 weeks, which is the point at which benefits terminate. Indeed the hazard for industry A starts to rise at about the 17 week mark, as workers start accepting lower wage jobs in the knowledge that benefits will soon expire. After 25 weeks the hazard falls again, due to the aforementioned selection effect. 3The parameters of the simulated model were as follows. T = 25, X = 0.1, wA = 12, wB = 6, wc = 4, P = 0.6, wu = 3, m = 2, |X = 0.5, Y A = 0 . 3 and yB = 0.1, The distribution of U was assumed to be uniform on [0,Wj-wu), the assumption being that it must be that an agent preferred the wage on their old job to not working. - 47 -III. Conclusions. The principal conclusions of this chapter are that at longer non-employment durations we can expect the hazard function of those separated from a lower paid depressed industry to be below that of from a higher paid expanding industry, despite the fact that workers in lower paid industries will on average have lower valuation of leisure. This is because the average hazard for any group ultimately tends to that of its most immobile group, which in the present case will be those with high valuations of leisure and high mobility costs. Such persons will have a better chance of finding work if they come from an industry which is expanding (i.e. for which the arrival rate of jobs is higher) than if they are from an industry with a lower arrival rate. In the short run however a number of different effects may dominate, making the hazard for those separated from the expanding industry lower than those from the depressed industry. One such influence is the existence of an unemployment insurance scheme which ties benefits to wages. In section two we show that this may cause a "spike" in the hazard, as agents refuse jobs early on in their unemployment spells because they can earn more from the insurance scheme. However as the expiration point approaches, relative value of unemployment declines, since there is more chance that the agent will not be able to find a job in time to avoid a fall in welfare once benefits are terminated. The size of this effect will tend to be most pronounced in better paying industries: in lower paid industries there may be no effect at all, because their unemployment benefits (which are tied to the previous wage) - 48 -may differ little from the amount received in public assistance (which is not a function of the previous wage) when UI benefits expire. - 49 -Figure 3: Hazard Funct ions for agents laid off in indus t ry B 15 20 25 30 35 Weeks Unemployed 45 50 o rN Figure 4: Behaviour of V over t he Non—employment spell. I I 1 1 1 1 1 1 I • ~^x : \ : \ vf — — Upper Bound Lower Bound ---i i i _i — J — i , J i i T Figure 5: Simulated Hazard Funct ions CD a >^ o i CD P^ ti—i O o OH 0 d ao o d CD o d < * o d CX2 o r~5 *—' O o - \ • \ " \ - \ ----\ 1 \ 1 , , , \ \ \ \ / - ^ ^ \ / ""- -I I 1 '"1 "1 • ' " " " I " I industry A — — Industry B --" ---~ N. ^ \ ^ "^"-^ ^^  1 1 1 ~~\ l i n 10 15 20 25 30 35 Weeks Unemployed 40 45 50 IV. EMPLOYMENT AND TRADE PATTERNS IN CANADIAN INDUSTRY 1961-90. This section attempts to classify industries according to two criteria. Firstly we seek to identify those industries which can be viewed as in long term decline. By this we mean an industry in which employment growth has been depressed over a such long period of time that workers laid off in that industry believe that their chances of obtaining another job in the same industry are relatively bleak. One would expect employees in such an industry to behave differently from those in an industry which is undergoing only a temporary downturn: workers in the latter case might feel more tempted to wait in the hope of obtaining a new job without attempting to retrain for another industry. The second dimension of interest is the extent to which job losses in an industry are due to international trade factors such as exchange rate fluctuations or the improving quality of foreign goods. We are interested in whether or not workers in these industries experience longer periods of joblessness as well as other differences in their post separation behaviour. 1. Employment 1961-90. Table 1 below illustrates the employment history of 25 primary and secondary industries for the period 1961-901. Eight industries saw employment grow by less than average in both the period 19 61-85 and the 1 The excluded industries are agriculture and fishing. These industries are not covered by the SEPH due to the high number of self employed. - 53 -period 1985-90. These industries were typically in either the traditional manufacturing or primary sectors. Seven other industries, all in the high technology or "other" sectors, saw employment increase by more than average. Table 1: Employment in Forestry, Mining and Manufacturing 1961-90. Industry All Mining & Forestry Traditional Mfg Resource-based Mfg High Tech Mfg Other Mfg Forestry Metal Mining Mineral Fuels Non Metal Mines Quarries & Sand Pits Mining Services Food & Beverages Tobacco Rubber & Plastic Leather Textiles Clothing Wood Products Furniture Paper & Allied Printing & Publ. Primary Metals Metal Fabricated Machinery Transport Equipment Electrical products Non Metal Mineral Petroleum & Coal Chemicals Misc. Manuf Prods Employment in 1985 1952.7 % Change 1961-85 18.4 % Change 1985-90 5.0 221 .0 443 .9 3 8 9 . 1 558 .6 340 .0 59 .4 4 6 . 8 4 4 . 0 2 2 . 2 4 . 7 4 4 . 0 2 1 6 . 7 6 . 6 58 .9 2 9 . 4 2 4 . 6 113 .3 99 .4 53 .5 114 .5 124 .0 103 .8 142 .5 7 8 . 3 192 .2 145 .2 4 8 . 2 23 .2 84 .0 7 3 . 5 - 4 . 2 0 . 3 1 3 . 8 4 8 . 6 31 .6 - 4 5 . 3 - 1 3 . 7 131.4 19 .2 - 4 8 . 6 107 .4 8 . 2 - 3 4 . 4 2 9 . 2 - 3 4 . 6 - 2 2 . 1 4 . 5 6 . 6 12 .6 2 1 . 2 4 8 . 2 1 6 . 7 15 .5 6 0 . 8 71 .2 3 4 . 5 3 . 9 2 3 . 2 36 .9 4 3 . 3 - 5 . 3 - 1 . 5 2 . 6 12.4 10 .9 - 4 . 3 - 6 . 4 - 0 . 1 - 2 . 4 8 9 . 7 - 2 2 . 3 4 . 1 - 2 6 . 6 3 4 . 4 - 3 5 . 9 - 1 7 . 1 - 2 . 7 15 .3 7 . 9 6 . 4 1 5 . 1 - 1 2 . 1 9 . 9 2 . 2 13 .9 7 . 6 8 . 3 - 1 6 . 3 11 .4 5 . 8 Source: Employment, Earnings and Hours (cat.no. 72-001) - 54 -2. Changes in Import Penetration. An industry will be designated as adversely affected by import competition if there has been a significant increase in imports in that industry over the relevant time period. Following Abowd and Lemieux (1991) , one can decompose the rate of growth in gross output (Y) of industry i into the proportions due to domestic market growth, export growth and import growth. The decomposition is given by Amr i t = Ii±zlitAlnDit + | i iAmx i t - ^±AiPRit, M t rit yit where X is real exports, M is real imports, D is real domestic consumption and IPR is the import penetration ratio (M/D). The third term in the expression indicates the extent to which the growth in an industry's sales can be thought of as being due to a reduction in import competition. Table two above provides this decomposition for the years 1961 and 1985, and table three gives the decomposition between 1985 and 199 0. See appendix one for details of the data. - 55 -Table 2: Decomposition of Change in Gross Output between 1961 and 1985. Industry %A Output %A in Output due to: IPR %A Domestic %A Exports %A Imports in 1985 All Mining &Forestry Traditional Mfg Resource-based High Tech Mfg Other Mfg Forestry Iron Ore Mining Mineral Fuels Non Metal Mines Food & Beverages Tobacco Rubber & Plastics Leather Textiles Clothing Wood Products Furniture Paper & Allied Printing & Publ. Primary Metals Metal Fabricated Machinery Transport Equipment Electrical Non Metal Mineral Petroleum & Coal Chemicals Other Manuf Products 98.1 103.4 58.7 82.5 161.3 91.9 63.4 97.5 129.2 64.4 55.7 9.9 223.5 4.1 100.0 59.9 104.1 98.0 72.3 90.5 73.6 82.6 161.9 182.9 126.2 74.1 78.2 133.4 117.0 5.7 36.8 58.3 61.0 93.5 83.2 65.9 29.4 30.2 23.0 52.0 10.4 176.9 45.7 90.3 75.6 51.5 81.4 40.6 80.5 64.3 69.7 128.7 53.7 116.8 62.9 52.3 133.4 104.9 92.0 55.3 8.5 22.8 129.8 18.8 0.0 58.8 86.4 27.4 7.5 -0.8 43.0 2.9 11.1 13.4 51.4 21.6 34.8 6.9 17.9 16.2 73.2 242.8 41.9 13.7 37.2 5.6 44.8 0.4 14.5 -8.1 -1.3 -32.3 -10.2 -1.0 7.7 23.7 4.9 -3.7 -0.0 3.3 -43.2 -2.3 -23.1 0.0 -6.8 -3.3 3.0 -9.2 -3.2 -33.1 -46.3 -31.7 -2.5 4.6 -5.5 -35.6 92.9 18.1 17.2 15.1 59.3 31.7 3.0 48.6 21.8 26.9 10.8 3.0 30.9 46.9 37.1 28.6 9.8 14.9 13.5 13.8 20.9 24.5 72.8 79.1 50.2 20.3 9.7 24.6 60.0 Source: The Input-Output Structure of the Canadian Economy (cat.no._ 15-20 In order to construct four samples for use in subsequent analysis we classify industries according to whether they seem to be in long term decline, and by whether they have suffered from import competition in the period 1985-90. An industry is judged to have depressed employment prospects if employment grew by less than average over the period 1961-85, and if employment in the industry declined in the period 1985-90. The object is to - 56 -Table 3: Decomposition of Change in Gross Output between 1985 and 1990. Industry %A Output %A in Output due to: IPR %A Domestic %A Exports %A Imports in 199 0 All Mining &Forestry Traditional Mfg Resource-based High Tech Mfg Other Mfg Forestry Iron Ore Mining Mineral Fuels Non Metal Mines Food & Beverages Tobacco Rubber & Plastics Leather Textiles Clothing Wood Products Furniture Paper & Allied Printing & Publ. Primary Metals Metal Fabricated Machinery Transport Equipment Electrical Non Metal Mineral Petroleum & Coal Chemicals Other Manuf Products 14.1 8.2 3.8 8.3 15.8 10.0 9.0 -16.7 10.1 2.6 3.4 -9.3 6.1 -24.8 7.0 6.6 13.9 12.9 10.4 12.1 3.5 10.6 35.0 11.5 27.6 5.9 8.6 9.0 6.3 0.9 5.9 5.8 6.9 10.3 14.1 11.1 -11.3 4.0 18.7 1.6 -5.6 17.7 11.8 6.7 12.0 15.2 29.7 -6.4 17.8 -6.6 14.4 23.9 0.9 29.0 10.5 10.2 1.4 9.9 13.9 7.9 4.0 5.4 12.5 2.1 -3.1 -5.9 14.9 -9.1 4.4 -0.9 -2.2 0.5 5.6 -0.3 0.6 0.8 30.8 -1.9 13.6 1.3 29.4 7.5 13.6 0.5 2.4 8.2 7.8 -0.7 -5.6 -6.4 -4.6 -7.6 -5.9 -0.7 -0.5 -8.0 -4.9 -4.1 -2.8 -9.9 -36.5 -5.6 -5.4 -1.3 -16.4 -5.9 -4.1 -9.1 -5.0 -18.7 3.4 -14.0 -4.9 -4.4 -15.5 -11.1 93.6 24.7 22.9 20.7 65.1 38.7 3.8 49.6 31.7 33.9 14.7 6.0 39.4 64.8 41.1 32.6 11.8 31.6 20.7 21.7 30.3 28.9 81.6 76.0 60.2 24.8 14.0 36.6 66.6 Source: The Input-Output Structure of the Canadian Economy (15-201) identify industries in which laid off workers are likely to have gloomy employment prospects if they do not seek jobs in other industries. A prime example is the textile industry where employment has halved since 1961: we would argue that unemployed workers in this industry have much 2 Figures for the whole metal mining industry are unobtainable before 19 88 - 57 -more incentive to retrain than those laid off in the transportation equipment industry, where employment has almost doubled since 1961. However some industries are hard to classify using this criterion: for instance the petroleum and coal products industry grew very fast between 1973 and 1981, only to shrink again in the next decade so that employment in 1990 was the same as the early 1960s. It might not be unreasonable for a worker in this industry to expect another boom in employment and thus to delay switching industries: consequently these sectors are excluded from the sample of declining industries. Expanding industries are defined as those where employment grew by more than average in both periods. Our criterion for whether or not an industry was negatively affected by competition from imports over the period 1986-90 is based upon the predicted impact of changes in import penetration. An industry is viewed as trade affected if the import penetration rate in 19 9 0 was greater than twenty five percent, and if the output decomposition described above indicates that sales would have been more than five percent higher were it not for increases in import penetration. Thus for example table three indicates that output growth in the rubber and plastics industry would have been almost double had it not been for the increase in "import penetration. Of the twenty-five industries examined, we were able to assign fourteen to one of the four categories described above. The breakdown is given in table four below, following which we provide a brief description of each of the groups. - 58 -Table 4: Categories of Industry. Employment Prospects long term decline long term increase (1) (3) Trade Affected Leather Textiles Clothing Primary Metals Rubber & Plastics Machinery Electrical Products Chemicals (2) (4) Not Trade Affected Forestry Metal Mining Tobacco Transportation Equipment Paper & Allied Printing & Publishing 3. Declining industries affected by import competition. The first group of industries - those classified as in long term employment decline due in large part to import competition - consists of the leather, clothing, textiles, and primary metals industries. The first three are traditional industries in which employment has been stagnant or declining since 1961. These are low paying industries: average Veekly earnings (AWE) including overtime in 1985 were respectively $342, $421 and $385. This low pay reflects the relatively low productivity of these industries. Since the 1960s the leather, clothing and textile industries have faced severe competition from Newly Industrialising Countries (NICs) in Asia and Latin America. In particular the four Far Eastern NICs3 3 South Korea, Singapore, Taiwan and Hong Kong. - 59 -increased significantly their share of Canada's imports. In the case of leather goods these countries' share of imports increased from 7.6% in 1971 to 32.9% in 1985; for textile products the increase was from 22.2% to 52.1%; and for clothing the share rose from 32.8% to 47.6%4. Not surprisingly, the affected industries have received much attention from government. Policy has taken two forms: aid to affected firms for restructuring and to workers for retraining; and tariff protection. An example of the former was the formation of the Canadian Industrial Renewal Board in 1981; an example of the latter was Canada's participation in the Multi-Fibre Agreement of 1974 which placed limitations on imports. Canada also maintains high tariffs in this area: the rates of nominal protection in the three industries in 1985 were higher (11.2% to 22%) than in all but two of the other sixteen manufacturing industries5. This special treatment is also reflected in the Canada-U.S. free trade agreement (FTA): tariffs on textiles and apparel will not be completely eliminated until 1999, whereas for most products tariffs were either eliminated immediately or were totally phased out by the end of 19936. In contrast to the three traditional industries just discussed, the primary metals industry is high paying (AWE were $617 in 1985), and was expanding until the mid seventies. However the industry has been declining steadily since then, even during the expansionary period of the late eighties. In the steel industry especially, these difficulties can be traced in large part to global over-capacity: many NICs have invested 4 These figures are taken from Economic Council of Canada (1988), Table A-3. 5 See Economic Council of Canada (1988), Table A-l. 6 Department of Regional Industrial Expansion (1988), p.32. - 60 -heavily in large modern plants, whilst in the developed world many governments have provided heavy protection for their domestic industries. Indeed in 1983 it was estimated that as much as 75% of the world's steel industries were government-controlled7. In contrast Canada has fairly moderate protection - average tariffs on primary metals were 4.1% in 1985 - although steel is another industry that was granted a ten year phase -period for tariffs under the FTA. The steel industry has also been the focus of remedial government attention. Policy has taken two forms: subsidies to maintain the operation of existing plants; and an adjustment programme for displaced steelworkers known as the Canadian Steel Trade and Employment Congress. Set up in 1988, this latter initiative offers retraining and job search assistance, and emphasizes close cooperation between management and unions. It appears to have been relatively successful8, although in our time period only a minority of workers would have had access to it. 4. Declining industries not significantly affected by import competition. The second group consists of declining industries that do not seem to have suffered from import competition in the period 1985-90: Forestry, Metal Mining and Tobacco. These are all high paying industries (AWE in 1985 were respectively $501, $659 and $675) in which employment has been steadily declining. 7 Government of Canada (1983), p.80. 8 See Canadian Labour Market and Productivity Review (1992), p.29. - 61 -Canada's forestry industry faces negligible import competition: its problems in recent years have stemmed instead from shortages of easily accessible timber, especially softwoods. Similar problems have affected the metal mining industry: unused mineral deposits tend to be more inaccessible and thus more costly to develop. The tobacco products industry is heavily protected in Canada: the nominal tariff rate was 43.1% in 1985. In consequence import penetration is almost non-existent. However the industry rivals the leather industry in its rate of decline: this is no doubt caused by falling demand due to the obvious health concerns and concomitant heavy taxes. 5. Expanding industries affected by import competition. Sample three is composed of high tech industries: rubber and plastics, machinery, electrical goods and chemicals. These industries have seen employment grow in the long term; however they have all lost market share to imports in recent years. In general these industries pay reasonably well: average weekly earnings in 1985 were $511. Import penetration is high, as is the degree of dependence on exports: with increasing returns efficiency requires production for the world market. Thus these industries are very sensitive to trade pressures, especially since tariffs are low - generally in the 3-6% range. In consequence one might expect the four industries to have been adversely affected by the significant appreciation in the Canadian dollar in the period 1985-90 (see figure 1 below). However unlike more traditional industries, the products produced by theses high-tech industries seem likely to be in - 62 -strong demand over the long term: these sectors also have shown strong export performance. In consequence government policy towards this sector has been more concerned with "upside adjustment" - retraining workers for jobs within the industry - rather than "downside adjustment" which involves relocating workers elsewhere. An example of this is Sectoral Skills Council of the electrical products industry, set up in 1987 to promote the training of new entrants as well as existing workers9. 6. Expanding industries not significantly affected by import competition. The fourth group consists of expanding sectors that did not appear to suffer from an increase in foreign competition over the time period. The sample consists of the transportation equipment industry, the paper and allied products industry, and the printing and publishing industry. The transportation equipment industry has very high import penetration: however unlike any other industry but iron ore mining, this rate actually fell between 1985 and 1990. Employment growth has been strong over both the long and short terms. An important factor in this growth has been the 1965 auto pact with the United States." This eliminated tariffs between the two countries while at the same time requiring American producers to locate a certain proportion of their production in Canada. In recent years the Canadian government has responded to increased competition for Par Eastern producers by negotiating Voluntary Export Restraints: these and equivalent U.S. 9 See Canadian Labour Market and Productivity Review (1992), p.27. - 63 -restrictions may have been a factor in the decision of several Asian car manufacturers to locate plants in Canada. The paper and allied products industry and the printing and publishing industries in Canada have fairly low import penetration rates and relatively low tariffs. The former industry has seen considerable restructuring with companies investing in more modern plant - often large scale mills that are integrated with lumber production. Some of this capital investment was stimulated by the Federal/Provincial Pulp and Paper Modernisation Programme initiated in 1979. This measure provided subsidies for modernisation. Along with the specific industry adjustment programmes mentioned above, another active labour market policy initiative during the 1986-90 period was the Canadian Jobs Strategy (CJS). Announced in 1985, the CJS incorporated a number of initiatives targeted at different sections of labour market. The most relevant of these in the case of laid off workers was the Job Development programme, which provided wage and other subsidies in order to induce employers to hire and train workers who had been unemployed a significant period of time. The actual effect of the CJS on workers is somewhat in doubt however: federal expenditures on training and job creation actually fell from 1984-5 to 1987-8 (see McBride 1992, pp.149-50). - 64 -Appendix. Statistics Canada does not publish comprehensive data on imports and exports by industry: the data is only available by commodity. Our task is further complicated by the fact that there are no less than four different commodity classifications used in published data. Three of these: those for imports, for exports, and for the input-output tables are based on a standard classification system and are in theory convertible: however this would require considerable work. In any event these three systems are being replaced by the new harmonised system: unfortunately this process is incomplete, and the new system is not fully compatible with the old. In view of these complications the simplest method of constructing industry import and export statistics is to use a single data source which is designed to be consistent: the input-output tables. The disadvantage is that data is only available annually. On the other hand all business sector industries are covered. In order to construct the decomposition of output given above one requires real imports (M) , exports (X) and gross output (Y) for each industry. Since the input-output definition of a commodity is* very closely tied to the definition of an industry it is a simple matter to match commodities to industries. - 65 -V. MICRO DATA. This chapter introduces the data that is the main focus of our analysis: the two Labour Market Activity Surveys (LMAS), produced by Statistics Canada. These are two complementary longitudinal data sets covering the periods 1986-87 and 1988-90. For each survey a representative sample was questioned each year about their labour market activities during the previous year. Information was obtained about up to five jobs that the respondent may have held, as well as demographic characteristics and information about the respondent's activities whilst unemployed or out of the labour force. Perhaps the main disadvantage of the LMAS is that one only has two years (for the 1986-87 study) or three years (for the 1988-90 study) of data, although one does know when the respondents began their jobs. However these two or three year windows are relatively long compared to the length of an average unemployment spell, and so the LMAS is well suited to the analysis of this sort of behaviour. 1. Construction of the Data Set. For the period 1986-87 the LMAS file comprises 89,947 records, and for 1988-90 the file comprises 97,081 records: however because some records detail the activities of those who held no jobs at all, the number of actual jobs is less than this: 164,642 when both surveys are combined. We follow the LMAS definition of a job by not counting employment after a temporary absence as a new job unless there is a change of - 66 -employer or position. With regard to joblessness, we make no distinction between the various degrees of attachment to the labour market that are possible for those without jobs. This is due in part to the difficulty of identifying different non-employment spells from the LMAS data, discussed at length by Jones and Riddell (1991). In addition, since our interest is in post-displacement non-employment, we only examine those non-employment spells that commence within the sample period as the result of the termination of job. Thus we exclude left censored non-employment spells: the reason for this is that the LMAS does not provide information about jobs held prior to the survey period, thus we would be unable assign such jobs to one of the four industry groups. The period of non-employment ends when the respondent obtains another job, whether full time or part time. For many non-employment spells this event will not be observed, leading to right censoring. If the individual loses a job but is still working at another job then this event is not counted as non-employment. The main variable of interest is the period of joblessness following job loss1. This is of course only available if the employment spell ended within the survey period 1986-87 or 1988-90. We excluded spells in which the lost jobs were unpaid or where the worker was self-employed. This left 50,956 jobless spells. Table 5 presents the industry breakdowns. From these spells we constructed four samples according to the taxonomy illustrated in table 4 in the previous chapter. These are the data sets 1 The LMAS also provides information on stretches of joblessness associated that began before the survey period. However these are not useful for our purposes since the industry from which the respondent was displaced is not identified. - 67 -Table 5: Sample Construction. Records: 187,028 Jobs: 164,642 Paid Jobs held by those aged between 20 and 64: 120,739 Non Employment Spells following a Job: Agriculture Other Primary Manufacturing Construction TCOU Trade FIRE Services Public Administration 2,078 2,091 6,656 5,541 2,532 6,737 1,293 13,351 3,014 43,293 Non Employment Spells in Other Primary and Manufacturing: Sample (1) 790 Sample (2) 1,444 Sample (3) 834 Sample (4) 1,176 Other 4,503 8,747 - 68 -used in chapters VII, VIII and IX. In all calculations we used the sample weights in order to ensure that our data sets are representative of the entire Canadian labour market. 2. Summary Statistics. Table 6 below gives summary statistics for the four samples as well for all industries and for mining, manufacturing and forestry (MMF). All variables are for the job prior to the non-employment spell except where specified2. It is important to note that the statistic given for length of joblessness in the table above includes censored spells: thus the average completed spell of joblessness will be greater than figure given in the table. A similar caveat applies to the average wage upon re-employment: since not all spells are observed to finish we do not observe the re-employment wage for all persons and thus the mean value in the table is a biased estimator. If for example the wage exhibits negative duration dependence then the average re-employment wage for all spells of joblessness will be lower than that for uncensored observations. 2 The models proved to be unstable for covariate values much greater than one: thus all numerical covariates were scaled to standard normal variables before estimation. - 69 -Table 6: Means of Covariates used in the Calculations. All MMF (1) (2) (3) (4) Duration of Non-Employment: (in weeks) 2 2 . 2 8 23 .27 2 7 . 5 1 20 .44 2 3 . 4 2 2 6 . 3 2 Job Tenure: (in weeks) Job Tenure: 0-6 months 7-12 months 1-5 years 5-10 years 10+ years Region: Atlantic Quebec Ontario Prairies BC Gender: Male Female Age: Aged 20-24 Aged 25-34 Aged 35-44 Aged 45-54 Aged 55-64 Marital Status: Married Single (reference) (reference) (reference) (reference) (reference) Position in Family: Head Not Head Education: Elementary High School (reference) (reference) Post Secondary University 155.11 0.40 0.15 0.29 0.07 0.09 0.11 0.26 0.35 0.17 0.12 0.53 0.47 0.31 0.34 0.18 0.10 0.08 0.60 0.40 0.52 0.48 0.09 0.31 0.36 0.24 200.53 0.37 0.14 0.29 0.08 0.12 0.12 0.29 0.37 0.12 0.10 0.64 0.36 0.28 0.35 0.18 0.10 0.09 0.64 0.36 0.55 0.45 0.14 0.35 0.33 0.18 225.14 0.32 0.16 0.30 0.08 0.15 0.04 0.48 0.37 0.07 0.04 0.42 0.58 0.26 0.34 0.21 0.11 0.08 0.67 0.33 0.42 0.58 0.18 0.36 0.34 0.12 132.74 0.56 0.11 0.21 0.05 0.07 0.20 0.28 0.19 0.06 0.27 0.87 0.13 0.27 0.34 0.20 0.10 0.09 0.70 0.30 0.72 0.28 0.24 0.39 0.23 0.14 237.21 0.25 0.14 0.35 0.11 0.15 0.03 0.29 0.55 0.09 0.03 0.59 0.41 0.28 0.39 0.17 0.08 0.08 -0.59 0.41 0.51 0.49 0.07 0.31 0.30 0.32 265.52 0.31 0.13 0.29 0.10 0.17 0.08 0.24 0.50 0.10 0.08 0.62 0.38 0.27 0.33 0.18 0.10 0.11 0.60 0.40 0.58 0.42 0.10 0.29 0.37 0.25 Occupation: - 70 -0 . 7 3 0 . 2 7 0 . 2 1 0 . 2 1 0 . 1 2 0 . 2 5 0 . 2 1 0 . 6 8 0 . 3 2 0 . 1 8 0 . 1 8 0 . 1 3 0 . 3 2 0 . 1 9 0 . 6 3 0 . 3 7 0 . 1 9 0 . 2 0 0 . 0 5 0 . 3 7 0 . 1 8 0 . 7 3 0 . 2 7 0 . 0 9 0 . 1 3 0 . 3 7 0 . 2 2 0 . 1 8 0 . 6 9 0 . 3 1 0 . 2 0 0 . 2 0 0 . 0 4 0 . 3 4 0 . 2 2 0 . 6 2 0 . 3 8 0 . 2 3 0 . 2 3 0 . 0 5 0 . 2 8 0 . 2 1 Managerial/Professional 0.21 0.12 0.08 0.08 0.22 0.16 Other White Collar 0.26 0.14 0.14 0.03 0.19 0.18 Blue Collar (reference) 0.53 0.74 0.78 0.89 0.60 0.66 Weekly Hours 38.34 41.24 39.10 46.26 39.45 39.43 Average Hourly Earnings 9.24 9.83 8.68 10.78 10.37 10.43 (deflated by the CPI) If unionised or covered by a collective agreement: No (reference) Yes Reason for Leaving Job: Quit to Unemployment Quit out of Labour Force Seasonal Layoff Other Layoff (reference) Other Average Hourly Earnings in 9.73 10.09 8.86 10.84 10.86 10.71 New Job (if observed) Number of Observations 43,293 8,747 790 1444 834 1176 Weighted Total 12.6mil 2.6mil 341,384 289,062 415,727 431,902 The covariates used are for the most part standard in studies of this kind. A variable of particular interest is the dummy for those who left their jobs for seasonal reasons. One might think that such workers were beyond the scope of this study: however those laid off for seasonal reasons may still experience difficulties finding another job for non-seasonal reasons, such as declining industry employment due to foreign competition - the focus of this paper. For many variables there are marked differences across the four industry groups. Jobs lost in the import affected declining industries have a higher proportion of females, have lower average hourly earnings and were heavily concentrated in the province of Quebec in comparison with the other three industries studied. Jobs from the non import - 71 -affected declining industry sample were on the other hand more likely to be held by males, were highly paid, and were more concentrated in the Atlantic provinces and British Columbia. Average job tenure was much lower: this is probably due to the highly seasonal nature of employment in the forest industry. The expanding industries were fairly similar in the characteristics of jobs lost: positions were more likely to be white collar, job tenure was higher, the workers were more highly educated, and the jobs tended to be located in Ontario. The foregoing is exactly what one would expect from the nature of the industries studied. It is interesting to note just how low paid are the jobs in group one. Despite the fact that workers had similar education to other workers in the MMF sector and greater job tenure, pay is not only lower for jobs lost in these industries than for the MMF sector as a whole, it is also lower than the average for all jobs. - 72 -VI. A MODEL FOR PRE-DISPLACEMENT WAGES. In this section we examine flows into joblessness. In the theoretical model presented in chapter II, a central prediction was that workers in a declining industry who could easily move out to other sectors would do so, leaving behind those who would find it costly to switch industry. This latter group would tend to stay in their jobs despite declining relative wages, until the firm was forced to close down. Thus over time one would expect that those who quit would be earning more than those who laid off. This prediction concerning quitters is somewhat at variance with some other adverse selection models. In Greenwald (1986) for example those workers who quit are regarded by other employers as "lemons" because the firm which they have just left does not appear to be concerned about keeping them. The difference between his model and ours is that we examine the effects of adverse selection upon the workforce of a particular firm, whereas he is interested in the the effects upon the entire pool of unemployed workers. There are other theoretical reasons why those who quit iffay be earning lower wages than other workers. For instance in job matching models workers who quit are those who find that their productivity is unexpectedly low at a particular firm, or who simply find their job disagreeable. Another possibility is that an employee simply receives a better offer elsewhere: for a given wage distribution this is more likely to be so for those earning comparatively less. Also lower paid workers might be more likely to engage in on-the-job search in first place. A - 73 -third possibility is that employees accumulate firm specific human capital through on-the-job training or experience: in this case the quitters will tend to be those with less firm specific capital to lose, which means that they will generally be lower paid1. What empirical predictions are provided by our theoretical insight about the relative quality of those who quit in a depressed industry and those who are laid off? Suppose we examine a firm's employment at time t. For a given set of characteristics (age, experience, job tenure, etc.) we would expect all employees to be paid the same wage, wt. However some of these employees will be more mobile than others, and so the difference between their wage and their next best alternative will be less. Suppose now that at fc+1 the wage the firm can afford to pay falls to wt+1. Those who can find better paid work will leave rather than accept the pay cut, whereas the less mobile have no option but to accept the reduction. Now suppose that at t+2 the firm closes down and the remaining workforce is laid off. What connection do we observe between wages and the nature of the separation from employment? Those who quit rather than accept were earning wt, whereas those who were laid off were earning wt+1 < wt. This suggests that a test of our hypothesis might be to regress the wage at some point in time upon subsequent separation behaviour: this procedure is implemented in the rest of this chapter. 1 See Mortensen (1988) for a theoretical analysis of a general model that nests the matching, on-the-job search and on-the-job training models. - 74 -1. Data. The construction of the sample is similar to that described in chapter IV. However instead of selecting those employment spells that were observed to end in a period of joblessness, we use those jobs that were already held at the beginning of 1988. The idea is to compare individuals who are employed at the same point in time; those who join the industry after this point are excluded from the sample. The reason for this is that if relative wages are falling, the quality of those entering the industry will fall, so that if those with shorter job tenures are more likely to quit, then quitting will be associated by the regression model with lower wages. In table 7 we present the characteristics of this particular sample of jobs. It is instructive to compare this table to table 1 in chapter V, which is concerned with jobs that are observed to be terminated. Not surprisingly, average uncompleted job tenure is much higher for the right censored jobs described in table 7. This is due to the well known problem of length biased sampling which often arises in the context of censored duration data. In essence the problem is that by choosing for inclusion in our sample all those jobs in progress at the start of 1988, we are more likely to include a long spell than a short spell. A ten week job beginning in January 1987 would not be included in our sample, whereas a 53 week job that began at that time would be included. Thus the average job tenure of our sample is biased as a measure of the average job tenure of all jobs, since longer jobs are over-represented. - 75 -Table 7: Means of Covariates for Jobs in Progress 1988-90. All Job observed to end: 0.46 Job in Progress start of 1988 0.44 Job Tenure: (in weeks) Region: Atlantic Quebec Ontario (reference) Prairies BC Gender: Male (reference) Female Age: Marital Status: Married (reference) Single Position in Family: Head (reference) Not Head Education: Elementary High School (reference) Post Secondary University Occupation: Managerial/Professional Other White Collar Blue Collar (reference) Weekly Hours Average Hourly Earnings (deflated by the CPI) If unionised or covered by a collective agreement: No (reference) Yes 230.84 32.04 Industry Sample IAD NIAD IAE 0.55 1.00 558.60 0.52 1.00 588.80 0.50 1.00 556.98 37.51 38.98 36.75 NIAE 0.60 1.00 653.29 0 . 0 9 0 . 2 4 0 . 3 9 0 . 1 7 0 . 1 2 0 . 5 3 0 . 4 7 0 . 0 3 0 . 4 4 0 . 4 2 0 . 0 7 0 . 0 3 0 . 6 1 0 . 3 9 0 . 0 9 0 . 2 1 0 . 3 7 0 . 1 0 0 . 2 3 0 . 9 1 0 . 0 9 0 . 0 3 0 . 2 5 0 . 6 0 0 . 0 9 0 . 0 2 0 . 7 2 0 . 2 8 0 . 0 6 0 . 2 6 0 . 5 2 0 . 0 8 0 . 0 9 0 . 7 5 0 . 2 5 38.31 0 . 5 6 0 . 4 4 0 . 5 0 0 . 5 0 0 . 0 6 0 . 3 3 0 . 3 5 0 . 2 6 0 . 2 6 0 . 2 8 0 . 4 7 3 6 . 0 6 9 . 8 0 0 . 7 9 0 . 2 1 0 . 6 3 0 . 3 7 0 . 1 5 0 . 3 0 0 . 3 5 0 . 2 0 0 . 1 3 0 . 1 0 0 . 7 7 3 8 . 9 0 1 0 . 6 9 0 . 8 3 0 . 1 7 0 . 8 7 0 . 1 3 0 . 1 9 0 . 2 8 0 . 2 8 0 . 2 5 0 . 1 4 0 . 0 5 0 . 8 0 4 3 . 2 1 1 3 . 8 7 0 . 7 4 0 . 2 6 0 . 6 6 0 . 3 4 0 . 0 5 0 . 1 9 0 . 3 8 0 . 3 9 0 . 2 6 0 . 1 4 0 . 6 0 3 9 . 8 6 1 2 . 6 7 0 . 7 5 0 . 2 5 0 . 7 3 0 . 2 7 0 . 0 8 0 . 2 0 0 . 4 1 . 0 . 3 2 0 . 1 7 0 . 1 7 0 . 6 6 3 9 . 1 3 1 2 . 8 6 0 . 6 9 0 . 3 1 0 . 4 8 0 . 5 2 0 . 4 8 0 . 5 2 0 . 6 3 0 . 3 7 0 . 4 5 0 . 5 5 - 76 -Reason for Leaving Job: Quit to Employment Quit to Unemployment Quit to Out of Lab. Force Non Seasonal Layoff (Ref) Seasonal Layoff Other 0.09 0.09_ 0.11 0.56 0.09 0.06 0, 0. 0. 0. 0. 0. .10 .07 .07 .62 .12 .02 0.10 0.06 0.06 0.60 0.10 0.07 0.14 0.08 0.06 0.59 0.10 0.02 0.10 0.08 0.05 0.68 0.08 0.01 Real Average Hourly Wages 6.63 11.15 14.26 13.28 13.43 in 1988 Note that in chapter IV and in subsequent chapters we exclude those who make direct job-to-job transitions. This because our primary interest in these chapters is in non-employment rather than employment duration. However the data used in this chapter does include direct job-to-job transitions. This bias towards longer jobs is also reflected in the average hourly wages paid to workers in jobs in progress at the start of 19 88, since higher job tenure is usually associated with increased earnings. A caveat to this inference is table 1 refers to jobs held only in the 1988-9 0 period, whereas as table 1 in chapter V includes jobs held in the 1986-90 period, and thus part of the discrepancy can be ascribed to growth in real wages over the period. - 77 -2. An OLS model of Earnings. Initially we shall estimate some simple Ordinary Least Squares models of earnings. In all case the dependent variable is the logarithm of average hourly wages in 1988. The results of the first regression are detailed in table eight below. This calculation includes as regressors only the week in which employment ended (or was right censored) and the reason for the termination of the job, if indeed the job ended. The idea is not to model earnings per se, but simply to get an idea of the correlation between the reason for leaving and earnings. The most interesting result from our perspective is the coefficient upon the layoff variable. For samples one, two and four, the wages of those permanently laid off were 10-15% lower than those who remained in employment, whereas for sample nine wages were 12% higher. Except for sample four, this is consistent with our theoretical model: in declining industries those laid off are those who were willing to accept wage cuts rather than move elsewhere to better paying jobs. Table 8: OLS Earnings Equation for Jobs in progress at start of 1988. IAD NIAD IAE NIAE Parameter value s.e. value s.e. value s.e. value s.e. constant stopemp quit2emp quit2un quit2olf layoff seasonal other -7.369 .002 .165 .003 -.023 -.147 - .320 .171 (2.665)* (.001)* (.080)* (.097) (.092) (.070)* (.124)* (.091)+ -.370 .001 - .011 -.131 .003 -.109 -.074 - .044 (2.299) (.000) (.069) (.079) (.083) (.073) (.083) (.082) -8.235 .002 .130 .043 - .015 .124 .111 .168 (2.218)* (.000)* (.063)* (.070) (.074) (.063)+-(.114) (.073)* -3.545 .001 -.081 - .122 -.075 -.180 -.160 .090 (1.89)+ (.000)* (.050) (.054)* (.062) (.052)* (.097) (.057) - 78 -Table 9: OLS Earnings Equation for all covariates. Parameter constant stopemp jten jten2 atlantic quebec prairies be female age age2 single nothead elem postsec univ mgrlprof owcllr whrs collagre guit2emp quit2un quit2ol£ layoff seasonal other IAD value -2.190 .001 .000 -.000 < -.261 -.073 -.100 .068 -.340 .033 -.041 -.092 -.046 -.155 .090 .229 .188 .063 .001 .057 .134 .068 .174 -.005 .056 .137 s.e. 2.16) .000)+ .000)* .000)* .077)* .028)* .051)+ .071) .034)* .009)* .010)* .031)* .033) .040)* .032)* .039)* .041)* .042) .002) .026)* .059)* .072) .071)* .054) .093) .068)+ NIAD value 1.614 .000 .000 -.000 -.178 -.100 - .071 .100 -.073 .012 -.011 -.009 -.167 -.021 -.002 .066 .098 -.089 .001 .108 .017 -.039 -.061 -.069 .011 -.049 s.e. 2.23) .000) .000) .000) .051)* .038)* .048) .038)* .060) .009) .011) .039) .050)* .042) .038) .040) .045)* .068) .001) .031)* .063) .075) .079) '.067) '.077) (.078) value -4.479 .001 .000 -.000 -.113 -.024 .008 .039 -.242 .031 -.034 -.069 -.054 -.075 .044 .129 .240 .097 .006 .021 .070 .070 .072 .130 .105 .224 IAE s.e. (1.92)* (.000)* (.000)* (.000) (.064)+ (.028) (.041) (.075) (.030)* (.008)* (.010)* (.029)* (.029)+ (.060) (.033) (.035)* (.032)* (.036)* (.002)* (.027) (.052) (.058) (.062) (.053)* (.097) (.060)* value .684 .000 .000 -.000 -.109 -.035 - .096 .039 -.222 .034 -.041 -.043 -.101 .005 .042 .110 .129 .086 .005 .069 - .071 -.027 -.021 -.104 -.190 .101 NIAE s.e. (1.58) (.000) (.000)* (.000)* (.037)* (.021) (.034)* (.031) (.024)* (.006)* (.007)* (.021) + (.023)* (.037) (.024)+ (.026)* (.026)* (.027)* (.001)* (.021)* (."042) + (.046) (.053) (.043)* (.082)* (.048)* - 79 -The next regression uses all the available covariates. The results are presented in table nine above. In" this case we attempt to ascertain the extent to which the results of the first regression are due to the observable characteristics of those laid off. Including the other covariates does indeed change the coefficient estimates markedly. For sample 1 the value of the layoff coefficient is now effectively zero, indicating that those laid off were lower paid because they had characteristics associated with lower paying jobs. However, these results must be viewed with caution, since it seems plausible that job tenure may be a function of earnings. This will be so if those earning less are more likely to leave their jobs and thus to have shorter job tenure. One method of remedying this problem is to take an instrumental variables approach and replace job tenure by its expected value. - 80 -3. A Model for Employment Duration. The key quantity of interest in this section is the hazard function. Let T be the length of an individual's employment spell: then the hazard function at time t is given by Q/i> . . P(t+dt > T > t\T > t) U(fc) = lim — 1 , dt-»o dfc which is the probability that the spell ends at t, given that the spell has lasted at least as long as t. Typically one would expect this quantity to vary over time and over individuals. It is convenient to factorise the hazard into a "baseline" hazard 8Q which is a direct function of time and constant over individuals, and a second term which is a time independent function of individual specific covariates (x) . (These latter variables may however be time varying). Thus one has 6(t) = 90(t) exptf(x)] . The likelihood of a spell lasting T periods is given by L(T) = 0(T) F(T) where F(T) is the Survivor function which is defined as the probability that the spell lasts at least T periods. One can then estimate the model by maximum likelihood. For right censored observations the likelihood is just the value of the survivor function. The advantage of this class of models is that the estimation process incorporates explicitly the value of the censored variable. This makes it more efficient than limited dependent variable models of the Tobit variety, since in these models the dependent variable is only incorporated into the likelihood if the observation is uncensored. For an individual whose duration is censored at T, one does not, make use of the information that the agent managed to survive up to the period T-l. - 81 -Table 10: Results from the Piecewise Constant Duration Model. IAD NIAD IAE NIAE Parameter Beta 01 Beta 02 Beta 03 Beta 04 Beta 05 Beta 06 Beta 07 Beta 08 Beta 09 Beta 10 Beta 11 stopemp atlantic quebec prairie be female age age2 single nothead el em postsec univ mgrlprf owcllr whrs cavrd Mean InL E(Dur) value -1.457 -1.11 -1.008 -1.123 -.345 -1.345 -1.100 -1.335 .082 .293 .092 -1.023 -.202 -.027 .120 .198 -.029 -.054 .183 .067 .135 - .130 .561 .734 -.128 -.147 -.005 -.054 -0.71? 13.87E * : indicates that @ : indicates that s.e. .179)* .181)* .232)* .198)* .186)* .297)* .276)* .351)* .278) .278) .560) .079)* .491) .123) .321) .540) .145) .008)* .055)* .134) .173) .144) .132)* .298)* '.223) (.155) (.022) M 8 7 ) i ; the es value -.498 -.450 -1.234 -1.89 -.315 -1.235 -1.218 -1.78 -.027 ( .128 < .093 ( -1.01 ( .321 ( .678 < .218 ( .821 .411 1 - .274 .083 .198 .151 .261 .036 .095 -.686 -.675 .002 -.452 -0.354 4.4E timate is ; the estimate is s.e. .334) .341)+ .566)* .572)* .344) .504)* .799)+ .947)+ .534) .640) .874) .134)* .345) .287)* .589) .300)* .314) .091)* .082) .282) .301) .289) .382) .345) .502) .455) '.007) '.246) + I >6 value -1.345 ( -1.678 ( -1.751 ( -2.13 1 -.863 ( -.991 ( -1.348 1 -1.613 .665 ( .289 ( .203 ( -1.456 ( -.162 -.064 ( .086 .077 .003 -.163 .099 .194 .231 .219 .178 .606 - .145 - .174 -.003 -.143 -0.6r 17.: 3 significant at 3 signif icant at s.e. .242)* .278)* .251)* .298)* .249)* .227)* .291)* .429)* .223)* .334) .445) .075)* .607) .131) .271) .437) .131) .013)* .036)* .071)* .085)* .297) .168) .743) '.235) '.191) '.006) (.174) ?8 L20 the 5% the 10% value -1.351 -1.79 -1.356 -1.735 -.391 -1.012 -1.099 -1.489 .121 .123 .441 -1.421 -.304 .019 .145 .143 -.042 -.098 .101 .098 .134 -.123 .396 .797~ -.143 - .177 .009 -.187 -0. 17 level, level. S.( (.21: (.23' (.21: (.39-(.19 (.14 (.26 (.53 (.11 (.45 (.78 (.08 (.40 (.15 (.18 (.25 (.13 (.01 (.05 (.19 (.18 (.26 (.J 6 (.26 (.22 (.14 (.01 (.23 786 .643 - 82 -The empirical specification used in this chapter is the Piecewise Constant Spline specification, outlined "in Lancaster (1990) . We postpone until chapter VII a full discussion of this model. The results from estimation are given in table 10 above2. The estimated completed duration was highest for workers in the two expanding industry groups, and lowest in the declining, non-trade affected industries. In all industries those workers with higher earnings conditional upon observed characteristics tended to have lower hazard rates out of employment and thus longer employment durations. This result also held for older workers. However as we have noted in the introduction the results above, especially the estimate of employment duration, are biased because longer job spells are over-represented in our sample. This is because our sample consists of those jos in progress at the beginning of 1988. Thus our results are to be viewed with caution. Indeed the main purpose of estimating the model is to provide an estimator of employment duration for use in the following section. Note that in estimating the duration model we use data on all jobs observed within the period 1988-90. Whilst this does not eliminate the "bandwidth" bias outlined above, it does ameliorate the problem since it means that jobs which started aft-er the beginning of 1988 are included, and these jobs will tend to be of shorter duration. An ideal way of dealing with the problem would be to use only those jobs observed to start in the period 1988-90 (this is the approach taken in the examination of non-employment data), however our sample in this case would not include jobs of longer than three years duration. It 2The intervals for which the hazards were calculated are given in years by c = 0.25,0.50,0.75,1,2,3,4,5,10,15,20. For jobs of longer than twenty years duration no hazard was calculated. - 83 -would thus be very hard for us to make reliable predictions of completed job duration, given that a significant proportion of jobs do indeed last for longer than three years. 4. Instrumental Variables model of Earnings. The next task is to use the duration estimates of the previous model in a least squares regression in order to eliminate dependence between earnings and job tenure due to unobserved variables. As with the OLS model, the dependent variable is the logarithm of average hourly earnings in 1988, and the sample consists of those workers who were employed at the start of 1988 in one of the four industry groups. The results are to be found in table eleven below. Once again we focus upon the variables dealing with the reason for leaving. In comparison with the OLS model (see table nine), the instrumental variable model predicts somewhat lower relative wages for those who remained in the industry. This is what one would expect. Suppose there is some unobservable factor that influences the decision to stay or go. Then this variable will be positively correlated with job tenure and negatively correlated with the quit decision. If one does not include this variable in the regression, its effects will show up as a higher than otherwise coefficient on job tenure and a lower than otherwise coefficient on the quit or layoff variables. Once the effects of this variable are controlled for, which is the purpose of the instrumental variables (IV) procedure, we should observe a rise in the estimated coefficient on quitting, which is what has happened. - 84 -Table 11: Instrumental Variables Earnings Equation. Parameter constant stopemp j ten j ten2 atlantic quebec prairie be female age age2 single nothead el em postsec univ mgrlprof owcllr whrs collagre quit2emp quit2un quit2olf layoff seasonal other value -2.61 .001 .000 -.000 - .225 - .067 - .087 .114 - .187 .046 - .036 - .080 -.047 -.178 .141 .298 .135 .067 .004 .089 .243 .174 .330 .098 .161 .312 IAD s.e. (2.28) (.000)* (.000)* (.000)* (.043)* (.033)* (.059) (.098) (.044)* (.011)* (.014)* (.045)* (.037) (.051)* (.034)* (.047)* (.043)* (.065) (.007) (.028)* (.071)* (.069)* (.091)* (.052)+ (.093)+ (.079)* value 2.129 -.000 .000 -.000 -.176 -.032 - .038 .169 - .042 .010 -.007 - .014 -.186 .011 .018 .098 .054 -.056 .003 .043 .127 .056 .113 .015 .085 .143 NIAD s.e. (1.929) (.000) (.000)* (.000) (.078)* (.065) (.065) (.067)* (.068) (.008) (.014) (.041) (.055)* (.046) (.035) (.061)+ (.099) (.065) (.002) (.013)* (.049)* (.067) (.091) (.051) (.124) (.091) value -2.381 .001 .000 -.000 -.,034 .032 .054 .021 - .145 .024 -.078 -.056 -.071 -.039 .087 .165 .201 .034 .006 .045 .196 .278 .265 .301 .256 .371 IAE s.e. (1.356) (.000)* (.000)* (.000)+ (.024) (.018) (.053) (.071) (.034)* (.011)* (.078) (.023)* (.031)* (.051) (.067) (.038)* (.034)* (.043) (.004)+ (.065) (.055)* (.067)* (.071)* (.088)* (.104)* (.085)* value 1.086 .000 .000 -.000 -.168 - .071 - .094 .022 - .243 .034 - .046 -.054 - .073 -.004 .135 .187 .111 .068 .005 .032 .041 .063 .179 .023 -.078 .134 NIAE s.e. (1.404) (.000) (.000)* (.000)* (.056)* (.032)+ (.034)* (.025) (.089)* (.007)* (.008)* (.042) (.024)* (.038) (.037)* (.031)* (.045)* (.078)* (.001)* (.037) (.035) (.-041) (.067)* (.033) (.075) (.042)* : indicates that the estimate is significant at the 5% level, : indicates that the estimate is significant at the 10% level. - 85 -Turning our attention to the quit/layoff distinction, the results of the IV model suggest that for the declining industry, those who quitted to employment were earning more given their demographic characteristics than those who were laid off, whereas for the expanding industries there was no significant difference. This tends to confirm our hypothesis from chapter II that the "best" (in some sense) workers are the most likely to quit in industries with a bleak future, leaving behind the least mobile. The rationale for this behaviour is that firms are unsure about exactly who their most mobile workers are, and will be therefore be unable to selectively raise wages for these marginal employees to the extent required. Note that we are implicitly assuming that the unobserved heterogeneity is correlated with some observed characteristics, or else those who quit would on average earn the same as everyone else. Our assumption is the unobserved (to the firm) characteristics are positively correlated with the characteristics that are unobserved to us but not to the firm. - 86 -VII. A DURATION MODEL FOR NON-EMPLOYMENT. This chapter is devoted to the estimation of a duration model for joblessness for those separated from resource or manufacturing industries in the 1986-90 period. The model is estimated separately for each of the four samples. In light of the theoretical models in chapters two and three, we shall be paying particular attention to the hazard function at longer durations. Also of interest are differences between the four samples with respect to the unemployment insurance system, and the relation between non-employment duration and the reason for leaving one's job. Note that in this chapter and those that follow we use the data set described in chapter IV which excludes direct job-to-job transitions. This is because our focus is on joblessness, rather than employment (as in chapter VI). 1. Estimated Model. In specifying the duration model there are two main issues: how flexible to make the functional form for the baseline hazard; and whether or not to allow for unobserved heterogeneity. With jobless spell data one is likely to have spikes in the hazard rate at the point at which benefits run out. This suggests that the piecewise constant specification of Meyer (1990), discussed briefly in the previous chapter, may be preferable to assuming a smooth parametric form for the hazard such as that produced by the Weibull distribution, which is the form most popular in the - 87 -literature. One disadvantage of the piecewise constant specification is that the model may become unwieldy if each time period is allocated a separate parameter. If individuals' hazard functions differ according to some random variable that is omitted from the estimation then the resulting estimates will be biased. To understand why this so, consider the case where the true hazard function is constant over time, but that a variable y that is positively correlated with the hazard is omitted. Then individuals with high realisations of y are likely to exit earlier, which implies that after several periods the sample of surviving individuals will consist of agents with generally lower values of y. Thus the estimated conditional probability of exiting at time t will be lower than at time 0 not because the underlying hazard rate is lower but merely because the pool of those who might exit at t is made up of those less likely to exit in the first place because of their lower value of y. In the light of the foregoing we estimate three different duration models. These are as follows: Kaplan Meier Hazard Model. (M-l parameters) In this specification we ignore the covariates and estimate a piecewise constant spline for the baseline hazard. Let fcj denote the spell length observed of respondent i, and let the time axis be divided into M intervals of the form [cm_1,cm), where c0 = 0 and cM = oo. The likelihood function for each individual i is given by1 1 This and subsequent expressions are drawn from Lancaster (1990). - 88 -t e icm-i,cm), m = 1,...,M—1. cm~cm-l M - l M m - 1 2(j3) = 5j [ d i m ln[ l -exp{-exp(p m )}] - [ d i m [ e x p < P 1 ) m=l "m=2 1=1 where ( 1 ; t i e [cm_!-cm) 0 ; o . w . and Sj = 1 indicates that the observation is not right censored. The hazard rate for terminating employment in the rath period is given by 0 ( t ) = expjP^ t cm cm-l The purpose of estimating this model is to provide a benchmark for evaluating the ability of the following models to explain the variation in the hazard rate. The size of the intervals is given in weeks by c = 0,1 ,90,157 ,-where 157 weeks is the longest possible spell length. The reason for combining all spell lengths over 9 0 weeks is that there are very few uncensored spells after this point in time. Piecewise Constant Spline Model. This model combines the piecewise spline formulation of the Kaplan Meier model with a covariate vector to give the following log likelihojpd: M-l M m-l 2i(P,y) = Si V dimlntl-expC-exptPro+XiY)}] - ]T dimexp (xjy) £ exp ((3j) m=l m=2 1=1 The hazard function is 9i(t) = &exp((3ro) + exp (XiY) , t e [cm_1,cm), m = 1,...,M-1. In order to limit the number of parameters we set c 0,4,8,...,52,65,78,91,00 : thus M= 17. The expected duration of a completed jobless spell (T) is - 89 -E ( T ) J T &F(T) 0 M - l J C= £ f ( m - l ) Cm~2Cm-1 + F(cM_i) E ( T | T > C M _ X ) . m=l The complication in this procedure arises from the last term. In the estimation of the duration model we were not able to calculate the probability associated with the last segment of the distribution, since the survivor function .F(cM) would be zero. This is because the model assumes that all individuals change their state eventually. In consequence we have no information upon the rate at which people leave. Even if this were not so the last segment would still need to be treated differently since [cM_l7<x>) is of infinite length. To deal with this problem we adopt an approach similar to that of Green and Riddell (1993), and assume that T > CM-I are distributed according to the exponential distribution F(T\T > cM-i) = exp{ (GM-i+xy) T } , where the parameters 0M-I and y are taken from the estimation of the PWC model. Thus we are assuming that the value of the hazard is the same for the last two periods. The expected value of T conditional on being in this last segment is therefore: E(T T > cM_i) = cM_! + m —nr-exp (UM-a+xy) Piecewise Constant / Gamma Mixture Model. We assume that hazard function takes the form 0i (t) = Vi0o(t)exp(xiy) where V is a random variable with the unit mean Gamma distribution 2 9(1,0 ). The log likelihood in this case is ii(P,y,a2) = S i i n m + c r 2 ^ ] ^ 2 - [i+G2sii]'°2) + - 90 -(i-5i)in([i+a2s^i.1]"CT ), where Smi = exp (xjV) 1=1 The expected length of non-employment duration is calculated in the same manner as that for the PWC model above. 2. Results. In describing the results from estimating the three models we initially concentrate upon the baseline estimates, which describe re-employment probabilities over time of an individual who is in the reference group for each of the dummy variables, and who possesses mean values for the numerical covariates2. The hazard function of an individual with different characteristics will have the same shape as that for the baseline respondent: the covariates merely serve to move up or down the whole hazard function equally for each point in time. The relevant parameter values are listed in tables fifteen to seventeen in the appendix to this chapter, and the survivor functions in figures six, eight and nine. Figure seven displays the hazard function for the "PWC model. An examination of the baseline results indicates firstly that individuals in sample one have the lowest re-employment probabilities3. 2In this case it turned out that there was little difference between using the mean for all covarites or the mean for each different sample. 3 This difference was statistically significant at the 1% level. The test statistics for the null of no difference between the baseline hazards in the PWC model were 295.95, 159.48 and 243.85 for each of the three other models. These statistics are distributed as X2(16). - 91 -In the Kaplan Meier model the difference becomes apparent after twenty weeks: for the other two models the difference between sample one and samples two and three is obvious from the start; and the divergence from sample four begins after fifteen weeks. Note how the survivor functions for samples two and three are much lower than that for sample four in the PWC and PWC/Gamma models: this contrasts with the results of the Kaplan-Meier model, where the three survivor functions are very similar. This difference is due to the effects of the covariates: clearly the parameter estimates associated with the baseline group serve to increase the hazard and therefore lower the survival rate. Turning to figure seven, which illustrates the baseline hazard functions, one notes that in general the hazard is fairly constant after an initial steep decline: however for samples two and three there is a large spike between thirty and fifty weeks. This might possibly be explained by the exhaustion of UI benefits, which would occur around this time for most workers in our sample. In explaining the lower hazard rates for sample one, it is difficult to ascribe the differences to the different characteristics of the four samples, because the baseline estimates hold constant these attributes. Instead we appeal to the analysis in chapter three. It is important to bear in mind that the observed hazard functions are merely the aggregate of re-employment probabilities that are different for each agent. From lemma one we know that at longer durations the observed hazard will be more influenced by those with the longest durations, because they are more likely to be observed at longer spells. These enables us to explain why the hazard rate for sample one is similar to the other sample at very short durations, but much lower at longer durations. Suppose agents in all samples are divided into those with high moving costs and those with - 92 -low moving costs. Compare first agents in the latter group. Although the arrival rate of employment opportunities in the depressed sector is lower, this will not pose a serious problem for those with low moving costs, because they are open to offers from all industries. Their employment probabilities will be similar to other "footloose" agents in the other industries. However, for those workers who have strong links to their previous industry, re-employment probabilities will be tied much more closely to the arrival rate of job offers in that industry. In this case a wide variation in arrival rates will produce a large variation in hazard rates: for footloose agents this effect will be muted. Also in chapter three we gave a possible explanation for the differences in agents' responses to the expiration of unemployment insurance benefits. Recall from table six that a key feature of sample one is that average pre-displacement wages were lower for these workers than those in the other three industry groups. This in turn means that unemployment benefits will be lower for these workers, reducing the incentive not to work. In contrast, someone laid off from a full time job in an industry in sample three that paid the average wage of $10.50 per hour, would receive the equivalent of $6.30 per hour in UI benefits until their eligibility expired. This is comfortably in excess of "any provincial minimum wage. One would not expect these workers to consider lower paying jobs until the expiration point approached. However, after this point an agent's behaviour would no longer be a function of the previous wage, and relative hazard rates will be determined by mobility costs and the arrival rate of new job offers.4 4A more rigorous test of this hypothesis would be to interact a dummy variable for the expiration point with the pre-displacement wage. The main obstacle to such a test would be that the precise length of unemployment benefits depends heavily upon which of the (at that time) 48 UI regions one resided in. This information is unfortunately not available on the public use version of the LMAS used in this thesis. - 93 -Table 12: Covariate Estimates from the Piecewise Constant Model. Industry Sample Parameter ten7tol2 tenlto5 ten5tol0 tengtlO atlantic quebec prairies be female age2024 age3544 age4554 age5564 single nothead el em postsec univ mgrlprf owcllr whrs ahe collagre guit2un quit2olf seasonal other Mean LnL Sample Size (1) value -.063 ( -.254 ( -.400 ( -.503 ( -.363 ( -.602 ( .449 ( -.136 ( .111 ( .401 ( .188 ( .011 ( -2.81 ( -.132 ( -.575 ( .098 ( .479 ( .455 ( .186 ( .109 ( .385 ( 1.757 (1 .176 ( .604 ( -.607 ( -.005 ( .329 ( -1.493 790 s.e. .130) .113)* .201)+ .166)* .428) .091)* .202)* .212) .119) .117)* .115) .161) .366)* .107) .112)* .146) .109)* .160)* .202) .120) .632) .282) .098)+ .120)* .143)* .278) .111)* (2) value .060 -.226 -.264 -.454 -.214 .025 -.046 .066 -.186 .093 -.175 - .476 -1.523 -.148 -.294 -.085 -.010 -.153 .264 -.405 .402 2.933 .027 .517 -.651 .038 .428 -1.740 1,444 s.e. .088) .085)* .131)+ .124)* .133) .087) .158) .077) .096)+ .086) .077)* .103)* .151)* .079)+ .079)* .084) .080) .088)+ .101)* .176)* .200)+ .578)* (.067) (.102)* (.119)* (.082) (.087)* (3) value -.109 -.346 .019 -.474 -.254 -.396 -.001 - .343 .031 .391 - .147 -.740 -3.564 -.117 - .112 .116 .069 .204 - .154 .166 .928 6.319 -.036 .446 -.818 -.358 .104 -1.558 874 s.e. .131) .107)* .154) .159)* .416) .089)* .206) .183)+ .099) .101)* .123) .196)* '.695)* '.091) .083) .194) (.106) (.117)+ (.104) (.109) (.453)* (.927)* (.089) (.109)* (.134)* (.281) (.104) • (4) value -.116 -.213 -.503 -.713 - .225 - .155 .008 - .142 - .190 .155 - .282 -.227 -1.285 .015 -.261 .121 .074 .043 .369 -.096 - .087 .19-9 - .181 .300 -.866 - .261 .008 -1.750 1,235 s.e. (.092) (.079)* (.100)* (.120)* (.228) (.070)* (.125) (.125) (.080)* (.086)+ (.086)* (.118)+ (.167)* (.064) (.077)* (.144) (.082) (.089) (.091)* (.096) J.381) (.716) (.080)* (.085)* (.100)* (.180) (.085) * : indicates that the parameter estimate is significant at the 5% level, @ : indicates that the parameter estimate is significant at the 10% level, - 94 -Table 13: Covariate Estimates from the Piecewise Constant/Gamma Mixture Model. Industry Sample Parameter ten7tol2 ten2to5 ten6tol0 tengtlO atlantic quebec prairies be female age2534 age3544 age4554 age5564 single nothead elem postsec univ mgrlprf owcllr whrs ahe cavrd guit2un guit2olf seasonal other ln(CT2) Mean LnL ( value -.020 -.279 -.510 -.449 -.459 -.838 .605 -.172 .134 .465 .061 -.184 -2.715 -.269 -.744 .208 .601 .564 .123 .154 .296 1.859 .194 .717 -.811 -.066 .548 .413 -1.487 1) s.e. (.169) (.144)+ (.260)® (.217)* (.551) (.149)* (.305)® (.282) (.156) (.156)* (.148) (.217) (.478)* (.146)® (.168)* (.181) (.150)* (.215)* (.264) (.158) (.788) (1.695) (.127) (.162)* (.198)* (.349) (.163)* (.484) (2) value .046 -.323 ( -.457 ( -.650 ( -.349 ( .009 -.078 .164 -.139 .097 ( -.129 -.515 ( -1.976 ( -.268 ( -.394 -.054 .040 -.049 .346 -.667 .620 3.118 .037 .795 -.951 -.007 .600 .462 -1.735 s.e. .124) .113)* .178)* .164)* .175)® .113) .202) .100) .128) .114) .104) .153)* .210)* .108)* .115)* .117) .109) .117) .138)* .220)* .285)* .846)* '.087) (.148)* (.155)* (.106) (.126)* (.257)® (3) value -.032 -.490 -.015 -.664 -.524 - .668 .051 -.265 .228 .500 - .583 -1.746 -5.78 -.061 - .279 .319 .217 .255 .124 .691 .706 11.914 - .028 .866 -1.667 -.459 .464 -.402 -1.541 s.e. (.637) (.594) (.686) (.613) (.431) (.591) (.527) (.869) (.563) (.586) (.612) (.585)* (.250)* (.568) (.525) (.362) (.520) (.486) (.465) (.489) (.258)* (.134)* (.516) (.498)® (.535)* (.528) (.507) (.349) • (4) value -.102 -.234 -.532 -.820 -.265 -.164 .125 - .256 - .130 .155 - .433 - .404 -1.712 -.028 -.353 .117 .096 .014 .532 - .223 .036 .25-5 - .316 .432 -1.214 - .429 -.011 .472 -1.739 s.e. (.129) (.105)* (.130)* (.154)* (.293) (.091)® (.183) (.162) (.105) (.117) (.120)* (.159)* (.221)* (.085) (.102)* (.175) (.108) (.114) (.129)* (.121)® J.477) (.949) (.106)* (.120)* (.147)* (.231)® (.112) (.328) indicates that the parameter estimate is significant at the 5% level, indicates that the parameter estimate is significant at the 10% level. 95 Table 14: Expected Non-Employment Duration. Piecewise Constant Model Piecewise Constant/Gamma Model Sample : (1) (2) (3) (4) (1) (2) (3) (4) All Male Female >= 35 years < 35 years Tenure > 2yrs Tenure <=2yrs 54.9 49.1 59.0 76.1 41.7 71.9 44.7 33.9 32.4 44.1 38.8 30.2 51.8 30.4 46.2 44.5 49.0 73.3 31.7 63.3 34.7 42.3 39.5 47.9 58.9 30.6 61.9 30.2 71.7 64.0 77.3 103.7 51.9 95.5 57.6 45.4 43.8 57.1 48.8 42.9 • 66.2 41.5 17.5 18.0 16.8 36.8 7.2 28.2 10.4 58.8 56.2 63.8 81.6 42.7 85.1 42.4 Another statistic of interest is the expected length of non-employment, displayed in table fourteen above. The statistics are for the average rather than the baseline respondent: thus in calculation it was assumed that the dummy variables took on their their mean rather than their modal value. These statistics indicate again that sample one individuals had the worst re-employment record. However in contrast to the baseline estimates it is sample two which has the second longest expected jobless duration: this discrepancy can be explained by the fact that the expected jobless duration reflects the characteristics of all individuals, not just those with the "baseline" characteristics. Breaking these expected durations down by age, gender and job tenure we find that for both the PWC and the PWC/Gamma model women, older workers and those with longer job tenure had longer average durations of non-employment. It is important to recall that the numbers in table fourteen incorporate all the characteristics of the sample and thus do not - 96 -necessarily imply that gender is necessarily important in explaining duration of non-employment. This table simply gives the average duration for the women in sample, given all their other demographic characteristics. All that been done is to correct for right censoring. The parameter estimates for the covariates are presented for the PWC model in table twelve and for the PWC/Gamma mixture model in table thirteen. The parameter estimates are quite similar for all samples. The age dummies are an exception to this rule: the import affected declining industry sample shows much greater activity at the extremes, with those aged 20-24 relatively more likely to be re-employed than other groups, and those aged 55-64 relatively less likely to be re-employed than other groups. Another covariate of interest is the reason for leaving the job. It is unsuprising to see that the probability of re-employment is lower for those who quit to leave the labour market. It is interesting to note that those who leave in order to search for another job are significantly more likely to be re-employed than those who are laid off. This contradicts the Gibbons and Katz (1993) argument discussed in chapter one that workers who quit may acquire a stigma. In our data it is those permenantly laid off who appear to take longest to find a job, although whether this is due to demand or supply side effects is impossible to answer with the reduced form models we employ. Note that the negative effects of being laid off appear strongest in the import affected declining industry group. Finally, an employee's gender is typically assumed to be an important factor in explaining non-employment. Because female labour participation is lower than that for males (although it was rising fast - 97 -during this period), we might expect women to be more likely to leave the labour force for a period, and thus to be more likely to have long non-employment spells. This view is partially supported by the parameter estimates. The coefficient for being female is negative for samples two and four, and insignificant for samples one and three. Thus despite the fact that average unemployment durations for women are longer than those for men (from table fourteen) , this is not, in the case of the first industry, due specifically to their gender, but rather to other characteristics of the women in the sample (e.g. age). In tables 18 and 19, and in figures 10 and 11 we present estimates for the four samples when those aged less than 25 are excluded. The motivation for this exercise is that younger workers may have very different behaviour from that of their older counterparts, and that these differences may not be adequately captured by a single paramter. However the diferences between the two samples was slight. In summary then, the general thrust of results in this section is that there are important differences in the labour market behaviour of workers in the four different samples. In particular those in imp"brt affected declining industries appear to undergo significantly longer spells of joblessness, and this is true even when the samples are broken down by various categories. One might claim from these results that workers separated from industries in sample one are comparatively worse off than those in other industries. However this conclusion would be premature without an examination of re-employment earnings-: workers may - 98 -simply be trading longer non-employment for higher wages. This possibility is exami-ned in the next chapter. - 99 -Appendix. Table 15: Estimated Hazards from the the Kaplan Meier Model. Industry Sample Weeks I 0 , [ 1 , [ 2 , [ 3 , [ 4 , [ 5 , [ 6 , [ 7 , [ 8 , [ 9 , [ 1 0 , [ 1 1 , [ 1 2 , [ 1 3 , [ 1 4 , [ 1 5 , [ 1 6 , [ 1 7 , [ 1 8 , [ 1 9 , [ 2 0 , [ 2 1 , [ 2 2 , [ 2 3 , [ 2 4 , [ 2 5 , [ 2 6 , 1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12) 13) 14) 15) 16) 17) 18) 19) 2 0 ) 21 ) 22 ) 23 ) 24 ) 2 5 ) 26 ) 27 ) (1) value 0 . 0 0 0 0 . 0 7 2 0 . 0 4 0 0 . 0 3 9 0 . 0 2 2 0 . 0 3 0 0 . 0 1 3 0 . 0 1 1 0 . 0 1 8 0 . 0 1 5 0 . 0 0 8 0 . 0 2 0 0 . 0 1 3 0 . 0 2 3 0 . 0 2 0 0 . 0 1 2 0 . 0 1 2 0 . 0 0 9 0 . 0 1 2 0 . 0 2 2 • 0 . 0 0 9 0 . 0 1 5 0 . 0 0 9 0 . 0 1 3 0 . 0 1 0 0 . 0 0 2 s . e . 0 . 0 0 0 0 . 0 0 9 0 . 0 0 7 0 . 0 0 7 0 . 0 0 6 0 . 0 0 7 0 . 0 0 5 0 . 0 0 4 0 . 0 0 6 0 . 0 0 5 0 . 0 0 4 0 . 0 0 6 0 . 0 0 5 0 . 0 0 7 0 . 0 0 6 0 . 0 0 5 0 . 0 0 5 0 . 0 0 5 0 . 0 0 5 0 . 0 0 7 • 0 . 0 0 5 0 . 0 0 6 0 . 0 0 5 0 . 0 0 6 0 . 0 0 5 0 . 0 0 3 (2) value 0 . 0 0 6 0 . 0 6 9 0 . 0 5 0 0 . 0 2 7 0 . 0 2 2 0 . 0 2 8 0 . 0 2 8 0 . 0 2 6 0 . 0 1 8 0 . 0 3 2 0 . 0 1 2 0 . 0 1 5 0 . 0 2 1 0 . 0 1 6 0 . 0 1 2 0 . 0 1 0 0 . 0 1 1 0 . 0 2 2 0 . 0 2 4 0 . 0 1 4 0 . 0 1 6 0 . 0 1 2 0 . 0 3 0 0 . 0 1 8 0 . 0 2 2 0 . 0 1 6 0 . 0 2 9 s.e. 0 . 0 0 2 0 . 0 0 6 0 . 0 0 6 0 . 0 0 5 0 . 0 0 4 0 . 0 0 5 0 . 0 0 5 0 . 0 0 5 0 . 0 0 4 0 . 0 0 6 0 . 0 0 4 0 . 0 0 4 0 . 0 0 5 0 . 0 0 4 0 . 0 0 4 0 . 0 0 3 0 . 0 0 4 0 . 0 0 5 0 . 0 0 6 0 . 0 0 5 0 . 0 0 5 0 . 0 0 4 0 . 0 0 7 0 . 0 0 6 0 . 0 0 6 0 . 0 0 5 0 . 0 0 7 (3) value s.e. 0 . 0 1 0 0 . 0 9 3 0 . 0 4 5 0 . 0 2 7 0 . 0 2 1 0 . 0 2 1 0 . 0 1 7 0 . 0 1 0 0 . 0 0 7 0 . 0 2 5 0 . 0 1 1 0 . 0 1 7 0 . 0 1 1 0 . 0 2 1 0 . 0 1 5 0 . 0 1 9 0 . 0 0 4 0 . 0 1 2 0 . 0 2 2 0 . 0 0 9 0 . 0 2 1 0 . 0 2 7 0 . 0 1 6 0 . 0 1 4 0 . 0 1 7 0 . 0 1 0 0 . 0 1 3 0 . 0 0 3 0 . 0 1 0 0 . 0 0 7 0 . 0 0 6 0 . 0 0 6 0 . 0 0 6 0 . 0 0 5 0 . 0 0 4 0 . 0 0 3 0 . 0 0 6 0 . 0 0 4 0 . 0 0 5 0 . 0 0 4 0 . 0 0 6 0 . 0 0 5 0 . 0 0 6 0 . 0 0 3 0 . 0 0 5 0 . 0 0 7 0 . 0 0 5 0 . 0 0 7 0 . 0 0 8 0 . 0 0 6 0 . 0 0 6 0 . 0 0 7 0 . 0 0 5 0 . 0 0 6 (4) value s.e 0 . 0 0 2 0 . 0 8 7 0 . 0 4 9 0 . 0 3 3 0 . 0 2 2 0 . 0 2 6 0 . 0 1 6 0 . 0 2 0 0 . 0 1 3 0 . 0 1 7 0 . 0 0 4 0 . 0 1 7 0 . 0 1 2 0 . 0 1 1 0 . 0 1 9 0 . 0 1 0 0.0*14 0 . 0 0 9 0 . 0 3 4 0 . 0 2 7 0 . 0 4 3 0 . 0 1 4 0 . 0 1 2 0 . 0 0 5 0 . 0 1 8 0 . 0 1 5 0 . 0 0 8 0 . 0 0 1 0 . 0 0 8 0 . 0 0 7 0 . 0 0 6 0 . 0 0 5 0 . 0 0 5 0 . 0 0 4 0 . 0 0 5 0 . 0 0 4 0 . 0 0 5 0 . 0 0 2 0 . 0 0 5 0 . 0 0 4 0 . 0 0 4 0 . 0 0 5 0 . 0 0 4 0 . 0 0 5 0 . 0 0 4 0 . 0 0 7 0 . 0 0 7 0 . 0 0 8 0 . 0 0 5 0 . 0 0 5 0 . 0 0 3 0 . 0 0 6 0 . 0 0 6 0 . 0 0 4 - 100 -[ 2 7 , [ 2 8 , [ 2 9 , [ 3 0 , [ 3 1 , [ 3 2 , [ 3 3 , [ 3 4 , [ 3 5 , [ 3 6 , [ 3 7 , [ 3 8 , [ 3 9 , [ 4 0 , [ 4 1 , [ 4 2 , [ 4 3 , [ 4 4 , [ 4 5 , [ 4 6 , [ 4 7 , [ 4 8 , [ 4 9 , [ 5 0 , [ 5 1 , [ 5 2 , [ 5 3 , [ 5 4 , [ 5 5 , [ 5 6 , [ 5 7 , [ 5 8 , [ 5 9 , [ 6 0 , [ 6 1 , [ 6 2 , 28 ) 29) 30) 31) 32) 33) 34) 35) 36) 37) 38) 39) 40) 41 ) 42) 43) 44 ) 45 ) 46 ) 47) 48 ) 49 ) 50) 51) 52) 53) 54) 55) 56) 57) 58) 59) 60) 61) 62) 63) 0 . 0 2 1 0 . 0 1 9 0 . 0 0 9 0 . 0 0 7 0 . 0 1 7 0 . 0 1 4 0 . 0 1 2 0 . 0 2 0 0 . 0 1 5 0 . 0 1 0 0 . 0 1 4 0 . 0 0 5 0 . 0 1 4 0 . 0 0 5 0 . 0 3 6 0 . 0 2 2 0 . 0 1 9 0 . 0 2 1 0 . 0 1 3 0 . 0 1 5 0 . 0 0 7 0 . 0 2 4 0 . 0 0 5 0 . 0 3 0 0 . 0 1 4 0 . 0 2 6 0 . 0 4 0 0 . 0 1 8 0 . 0 0 8 0 . 0 3 2 0 . 0 4 1 --0 . 0 3 5 -. 0 . 0 0 8 0 . 0 0 8 0 . 0 0 6 0 . 0 0 5 0 . 0 0 8 0 . 0 0 7 0 . 0 0 7 0 . 0 0 9 0 . 0 0 8 0 . 0 0 6 0 . 0 0 8 0 . 0 0 5 0 . 0 0 8 0 . 0 0 5 0 . 0 1 3 0 . 0 1 0 0 . 0 1 0 0 . 0 1 0 0 . 0 0 8 0 . 0 0 9 0 . 0 0 6 0 . 0 1 2 0 . 0 0 6 0 . 0 1 4 0 . 0 1 0 0 . 0 1 4 0 . 0 1 8 0 . 0 1 2 0 . 0 0 9 0 . 0 1 7 0 . 0 2 0 • • 0 . 0 2 0 -0 . 0 2 6 0 . 0 0 6 0 . 0 2 0 0 . 0 3 1 0 . 0 2 1 0 . 0 2 1 0 . 0 2 1 0 . 0 4 2 0 . 0 3 3 0 . 0 4 9 0 . 0 3 9 0 . 0 3 2 0 . 0 2 5 0 . 0 2 8 0 . 0 2 3 0 . 0 2 0 0 . 0 3 2 0 . 0 2 9 0 . 0 4 3 0 . 0 2 2 0 . 0 2 8 0 . 0 2 5 0 . 0 1 3 0 . 0 1 0 0 . 0 1 6 0 . 0 1 8 0 . 0 2 9 0 . 0 1 7 0 . 0 1 9 0 . 0 3 7 0 . 0 2 1 0 . 0 0 6 0 . 0 0 3 0 . 0 0 8 0 . 0 3 5 . 0 . 0 0 7 0 . 0 0 3 0 . 0 0 6 0 . 0 0 8 0 . 0 0 7 0 . 0 0 7 0 . 0 0 7 0 . 0 1 0 0 . 0 0 9 0 . 0 1 1 0 . 0 1 0 0 . 0 1 0 0 . 0 0 9 0 . 0 0 9 0 . 0 0 9 0 . 0 0 8 0 . 0 1 0 0 . 0 1 0 0 . 0 1 3 0 . 0 1 0 0 . 0 1 1 0 . 0 1 1 0 . 0 0 8 0 . 0 0 7 0 . 0 0 9 0 . 0 1 0 0 . 0 1 3 0 . 0 1 0 0 . 0 1 1 0 . 0 1 5 0 . 0 1 1 0 . 0 0 6 0 . 0 0 5 0 . 0 0 8 0 . 0 1 5 . 0 . 0 0 7 0 . 0 4 1 -0 . 0 1 9 0 . 0 1 9 . 0 . 0 1 7 0 . 0 0 5 0 . 0 4 3 0 . 0 0 7 0 . 0 5 4 0 . 0 1 4 0 . 0 1 7 0 . 0 7 7 0 . 0 1 7 0 . 0 3 6 0 . 0 6 1 0 . 0 2 5 0 . 0 2 7 0 . 0 2 5 0 . 0 0 5 0 . 0 3 0 0 . 0 1 5 0 . 0 1 4 0 . 0 0 5 0 . 0 0 9 0 . 0 3 7 0 . 0 0 6 0 . 0 0 2 0 . 0 0 2 0 . 0 4 8 -0 . 0 1 9 0 . 0 0 2 0 . 0 0 2 0 . 0 2 2 0 . 0 0 5 0 . 0 1 1 . 0 . 0 0 8 0 . 0 0 8 -0 . 0 0 8 0 . 0 0 4 0 . 0 1 2 0 . 0 0 5 0 . 0 1 4 0 . 0 0 8 0 . 0 0 9 0 . 0 1 7 0 . 0 0 9 0 . 0 1 3 0 . 0 1 7 0 . 0 1 2 0 . 0 1 3 0 . 0 1 3 0 . 0 0 6 0 . 0 1 4 0 . 0 1 1 0 . 0 1 0 0 . 0 0 6 0 . 0 0 9 0 . 0 1 9 0 . 0 0 8 0 . 0 0 4 0 . 0 0 4 0 . 0 2 2 -0 . 0 1 5 0 . 0 0 5 0 . 0 0 5 0 . 0 1 6 0 . 0 1 0 0 . 0 1 2 0 . 0 0 2 0 . 0 1 8 0 . 0 2 2 0 . 0 2 5 0 . 0 1 8 0 . 0 2 0 0 . 0 3 8 0 . 0 3 3 0 . 0 2 8 0 . 0 2 3 0 . 0 2 1 0 . 0 2 0 0 . 0 1 6 0 . 0 2 0 0 . 0 2 8 0 . 0 2 4 0 . 0 2 9 0 . 0 2 4 0 . 0 1 5 0 . 0 3 8 0 . 0 2 3 0 . 0 0 9 0 . 0 3 9 0 .0_12 0 . 0 1 1 0 . 0 1 2 0 . 0 5 2 0 . 0 5 3 0 . 0 1 5 0 . 0 2 6 0 . 0 1 4 0 . 0 1 4 0 . 0 3 1 0 . 0 0 7 0 . 0 0 5 0 . 0 0 5 0 . 0 0 2 0 . 0 0 6 0 . 0 0 7 0 . 0 0 8 0 . 0 0 7 0 . 0 0 7 0 . 0 1 0 0 . 0 0 9 0 . 0 0 9 0 . 0 0 8 0 . 0 0 8 0 . 0 0 8 0 . 0 0 7 0 . 0 0 8 0 . 0 1 0 0 . 0 0 9 0 . 0 1 1 0 . 0 1 0 0 . 0 0 8 0 . 0 1 2 0 . 0 1 0 0 . 0 0 7 0 . 0 1 4 0 . 0 0 8 0 . 0 0 8 0 . 0 0 9 0 . 0 1 8 0 . 0 1 9 0 . 0 1 1 0 . 0 1 4 0 . 0 1 1 0 . 0 1 1 0 . 0 1 6 0 . 0 0 8 1 0 1 [ 6 3 , [ 6 4 , [ 6 5 , [ 6 6 , [ 6 7 , [ 6 8 , [ 6 9 , [ 7 0 , [ 7 1 , [ 7 2 , [ 7 3 , [ 7 4 , [ 7 5 , [ 7 6 , [ 7 7 , [ 7 8 , [ 7 9 , [ 8 0 , [ 8 1 , [ 8 2 , [ 8 3 , [ 8 4 , [ 8 5 , [ 8 6 , [ 8 7 , [ 8 8 , [ 8 9 , [ 9 0 , 64) 65) 66) 67) 68) 69) 70) 71 ) 72 ) 73) 74) 75) 76) 77) 78 ) 79) 80) 81 ) 82) 83) 84 ) 85) 86) 87) 88) 89) 90) 91 ) 0 . 0 0 7 0 . 0 0 6 • 0 . 0 0 5 0 . 0 5 1 -0 . 0 1 1 0 . 0 1 0 0 . 0 1 7 0 . 0 2 3 0 . 0 1 0 • 0 . 0 8 0 -0 . 0 8 5 -0 . 0 3 9 • 0 . 0 5 9 0 . 0 7 2 0 . 0 1 3 • 0 . 0 3 0 • • 0 . 2 0 6 . 0 . 0 0 9 0 . 0 0 9 • 0 . 0 0 9 0 . 0 2 9 • 0 . 0 1 4 0 . 0 1 5 0 . 0 2 0 0 . 0 2 4 0 . 0 1 7 • 0 . 0 4 9 -0 . 0 5 4 -0 . 0 4 3 • 0 . 0 5 8 0 . 0 7 2 0 . 0 3 8 • 0 . 0 7 4 . -0 . 1 6 5 . 0 . 0 0 7 0 . 0 1 4 0 . 0 2 1 0 . 0 0 9 0 . 0 0 8 0 . 0 0 3 0 . 0 0 4 0 . 0 4 3 0 . 0 6 6 0 . 0 2 6 0 . 0 1 7 0 . 0 1 1 0 . 0 2 7 0 . 0 8 4 0 . 0 3 5 0 . 0 5 6 0 . 0 1 2 0 . 0 2 1 0 . 0 5 0 0 . 0 4 8 -0 . 0 7 2 • 0 . 0 2 3 0 . 3 2 4 0 . 0 4 8 0 . 5 0 0 . 0 . 0 0 7 0 . 0 1 1 0 . 0 1 3 0 . 0 0 9 0 . 0 0 8 0 . 0 0 5 0 . 0 0 6 0 . 0 2 0 0 . 0 2 5 0 . 0 1 8 0 . 0 1 6 0 . 0 1 3 0 . 0 2 1 0 . 0 3 5 0 . 0 2 6 0 . 0 3 3 0 . 0 1 6 0 . 0 2 3 0 . 0 3 5 0 . 0 3 5 • 0 . 0 4 5 • 0 . 0 2 9 0 . 0 8 4 0 . 0 7 2 0 . 2 0 0 . • 0 . 0 1 7 0 . 0 8 6 0 . 0 2 3 0 . 0 1 6 0 . 0 0 4 0 . 0 3 5 -0 . 0 4 4 0 . 0 1 9 -0 . 0 1 3 0 . 0 0 7 0 . 0 1 1 -0 . 2 0 3 0 . 0 9 3 0 . 5 0 0 . • 0 . 0 1 5 0 . 0 3 1 0 . 0 1 8 0 . 0 1 6 0 . 0 0 8 0 . 0 2 4 -0 . 0 3 0 0 . 0 2 1 -0 . 0 1 8 0 . 0 1 7 0 . 0 2 7 -0 . 1 1 3 0 . 3 0 9 0 . 5 0 3 ' . 0 . 0 8 5 0 . 0 1 0 0 . 0 6 0 0 . 0 1 4 . 0 . 1 0 0 . 0 . 0 0 6 0 . 0 1 9 0 . 0 1 0 . 0 . 0 6 1 • -0 . 0 1 5 0 . 0 0 6 0 . 0 9 3 • 0 . 0 3 2 0 . 0 7 3 • • --0 . 1 2 4 0 . 1 1 2 ' 0 . 1 5 6 0 . 0 3 8 0 . 0 2 7 0 . 0 1 0 0 . 0 2 4 0 . 0 1 3 -0 . 0 3 3 • 0 . 0 1 0 0 . 0 1 8 0 . 0 1 3 • 0 . 0 3 3 . • 0 . 0 1 9 0 . 0 1 2 0 . 0 4 5 • 0 . 0 3 0 0 . 0 4 5 -• • • 0 . 0 7 0 0 . 0 7 7 0 . 1 2 0 0 . 0 7 4 Note: A dot indicates that there were no exits in that sample for that period: consequently the hazard function was not calculated. - 102 -Table 16: Estimated Baseline Hazards from the Piecewise Constant Model. Industry Sample (1) (2) (3) (4) Parameter value s.e. value s.e. value s.e. value s.e. beta beta beta beta beta beta beta beta beta beta beta beta beta beta beta beta 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 -1.873 -2.589 -2.800 -2.691 -2.878 -3.388 -3.075 -2.973 -2.796 -3.144 -2.524 -2.901 -2.682 -1.549 -1.715 -2.364 0.162 0.192 0.205 0.202 0.242 0.281 0.259 0.255 0.249 0.373 0.222 0.312 0.265 0.208 0.260 0.514 1. 1. 2, 2, 2. 2, 2. 2. 1. 1, 2. 1. 2 1. 1 0 .564 .959 .249 .512 .290 .293 .098 .288 .870 .664 .011 .880 .573 .348 .024 .475 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0, 0. 0. 0. .097 .108 .118 .132 ,128 .139 .145 .166 .128 .136 .186 .192 .275 .186 .162 .184 1.345 2.291 2.414 2.305 2.570 2.098 2.621 2.068 2.291 1.918 1.190 2.056 2.267 1.356 0.827 1.358 0.147 0.170 0.192 0.179 0.220 0.195 0.224 0.202 0.212 0.214 0.167 0.236 0.340 0.263 0.229 0.443 -1.768 -2.512 -2.997 -2.967 -2.456 -2.590 -2.996 -2.931 -2.275 -2.243 -2.485 -2.414 -2.242 -1.146 -1.342 -1.167 0.108 0.129 0.152 0.141 0.135 0.142 0.181 0.185 0.146 0.152 0.191 0.177 0.190 0.144 0.213 0.219 - 103 -Table 17: Estimated Baseline Hazards from the Piecewise Constant/Gamma Mixture Model. (1) (2) (3) (4) Parameter value s.e. value s.e. value s.e. value s.e. Beta 01 Beta 02 Beta 03 Beta 04 Beta 05 Beta 06 Beta 07 Beta 08 Beta 09 Beta 10 Beta 11 Beta 12 Beta 13 Beta 14 Beta 15 Beta 16 ln(CT2) .052 .015 .016 .014 .019 .015 .018 .017 .016 .015 .044 .024 .043 .033 .039 .017 .413 (1.72) (1.169) (1.21) (1.198) (1.253) (1.245) (1.282) (1.326) (1.328) (1.368) (1.329) (1.374) (1.361) (1.462) (1.52) (1.62) (.484) .049 .024 .022 .019 .026 .036 .026 .035 .087 .074 .064 .078 .051 .042 .044 .155 .462 (.567) (.613) (.636) (.648) (.675) (.679) (.700) (.726) (.738) (.771) (.806) (.839) (.843) (.862) (.884) (.885) (.257)+ .465 .143 .283 .203 .376 .250 .338 .181 .562 .637 .878 .419 .225 1.242 1.598 2.938 -.402 (.721) (.741) (.525) (.613) (.739) (.936) (.937) (.915) (.763) (.995) (.773) (.575) (.790) (.570)* (.523)* (.281)* (.349) .084 .031 .025 .026 .061 .023 .016 .031 .051 .047 .051 .057 .067 .093 .036 .029 .472 (.745) (.808) (.827) (.853) (.865) (.907) (.927) (.928) (.915) (.955) (.954) (.972) (1.04) (1.40) (1.127) (1.134) (.328) 104 -: Estimated Baseline Hazards from the Piecewise Constant Model Over 24 years of Age Sample. Parameter Beta 01 Beta 02 Beta 03 Beta 04 Beta 05 Beta 06 Beta 07 Beta 08 Beta 09 Beta 10 Beta 11 Beta 12 Beta 13 Beta 14 Beta 15 Beta 16 (1) value s. e. Industry Sample (2) value s.e. ,099 ,019 ,023 017 ,027 ,019 ,024 ,021 .019 ,011 .038 .026 .024 .025 .033 .017 (.201) (.266) (.262) (.290) (.276) (.316) (.296) (.323) (.408) (.652) (.290) (.350) (.422) (.297) (.339) (.852) 104 038 031 026 033 046 033 043 058 059 060 052 031 024 019 043 (. (. (. (. (. (. (. (. (, (, ( ( ( ( ( ( .112) .133) .155) .173) .172) .164) .186) .185) .194) .197) .220) .284) .331) .311) .438) .281) C alue 151 032 054 024 032 023 042 022 039 047 023 030 015 056 026 034 B) s. e. (.202) (.251) (.240) (.324) (.309) (.339) (.283) (.352) (.324) (.317) (.438) (.492) (.689) (.293) (.450) (.519) (4) value .179 .048 .038 .039 .067 .029 .018 .043 .039 .032 .037 .058 .068 .080 .023 .024 s.e. (.135) (.170) (.195) (.188) (.172) (.251) (.320) (.228) (.240) (.284) (.322) (.246) (.237) (.191) (.361) (.466) Mean LnL 1.437 1.701 -1.468 •1.6.68 - 105 -Table 19: Covariate Estimates from the Piecewise Constant Model: Over 24 years of Age Sample. Industry Sample Parameter ten7tol2 tenlto5 ten5tol0 tengtlO atlantic quebec prairies be female age3544 age4554 age5564 single nothead el em postsec univ mgrlprf owcllr whrs ahe collagre quit2un quit2olf seasonal other ( value .168 - .197 -.336 - .317 -.331 -.523 .307 - .229 .161 .168 -.024 -2.91 -.250 -.674 .070 .473 .501 .065 .151 .782 1.522 -.053 .666 -.716 .160 .387 1) s. e. (.182) (.152) (.235) (.196) (.487) (.115)* (.242) (.263) (.180) (.124) (.169) (.373)* (.144)+ (.167)* (.165) (.134)* (.183)* (.280) (.170) (.834) (1.516) (.118) (.146)* (.209)* (.331) (.139)* (2) value -.137 < -.213 ( -.256 -.408 - .114 .194 .176 .348 -.186 - .199 - .459 -1.455 -.213 -.329 -.055 .064 -.035 .271 - .504 .464 2.617 .045 .366 -.708 .008 .363 s. e. .107) .090)* .141)+ .127)* .161) .100)+ .183) .094)* .125) .078)* .108)* .152)* .098)* .101)* .091) .094) .124) .127)* .217)* .244)+ .626)* f.076) (.110)* (.153)* (.090) (.096)* (3) value .071 ( -.134 ( .151 ( -.453 ( -.271 ( -.474 ( .006 ( -1.19 ( -.082 ( -.164 ( -.823 ( -3.636 ( -.037 ( -.010 ( .260 ( .296 ( .095 ( -.290 ( .281 1 -.061 < 8.48 (1 -.099 .283 -1.62 -.350 .224 s.e. .175) .138) .178) .179)* .475) .115)* .224) .332)* .129) .135) .208)* .671)* .109) .104) .215) .141) * .161) .140)* .135)* .572) .672)* .135) .131)* .194)* .307) .136) ' (4) value .075 -.323 - .596 - .715 - .189 - .234 .073 -.220 - .141 -.332 -.253 -1.27 - .205 -.283 .111 .103 .195 .391 .064 .225 .775 -.190 .180 -1.46 -.283 .126 s.e. (.126) (.108)* (.113)* (.132)* (.263) (.089)* (.144) (.142) (.105) (.092)* (.127)+ (.167)* (.088)* (.101)* (.176) (.102) (.107)+ "(.111) * (.114) (.518) (.880) (.104)+ (.106)+ (.131)* (.198) (.106) - 106 -Figure 6: Kaplan Meier Model of Re—Employment Basel ine Surv ivor Func t ion en 6 oo O O CO O m o r^ o O CM o o o CD 0 -1— 1 •• ' 1 1 1 1 1 \ \V - vV - ' « \ \V 1 1" T" I • I r | - - r - — | r ^ 1 ^ I t l i p O I L A l l c C L c U U c C l l l i l l i y l l l U U S L t y — — (2) Non-Import Affected Declining Industry (3) Import Affected Expanding Industry (4) Non —Import Affected Expanding Industry • ^ N ^ v^ v ^ ^ - . \ % -" ^ ^ V 1 , 1 , 1 . 1 l 1 1 , 1 --— -— -~~ o 10 20 30 40 50 60 70 80 Weeks s ince end of prev ious j o b 90 100 o O to O CN O 00 O d q o o o Figure 7: Piecewise Cons tan t Model of Re —Employment Basel ine Hazard Func t ion —1 1 1 1 1 1 1 1 1 1 1 i 1 1 1 r (1) Trade Affected Declining Industry (2) Non-Trade Affected Declining Industry (3) Trade Affected Expanding Industry (4) Non— Trade Affected Expanding Industry \ \ / V \ \ / ° 0 10 20 30 40 50 60 70 Weeks s ince end of p rev ious j ob 80 CO O 90 Figure 8: Piecewise Cons tan t Model of Re —Employment Basel ine Surv ivor Func t ion o oo o o CD d m d d d d d o o 1 l \ \ :\K - \ \ _ \ \ \ -~ -i 1 1 T T | 1 | \ ^ v \ ^ ^ S s \ ^ ^ ^ " v \ ^ s ' • . . ^ ^ "*N. 1 , 1 , 1 i 1 I | l | l | l | I | ytj iraae MTieciec ueciinmg inausiry — — (2) Non—Trade Affected Declining Industry (3) Trade Affected Expanding Industry (4) Non—Trade Affected Expanding Industry * " * * " "" ' • * — . i , i ~ i r ^ T ss^~,'~—v ---— _ -i o 0 10 20 30 40 50 60 70 80 90 100 Weeks s ince end of p rev ious j o b Figure 9: Piecewise C o n s t a n t / G a m m a Mix ture Model of Re —Employmen t Basel ine Surv ivor Func t ion o • I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 r i • •• . . i • i i i i i i i i i i "^T - " = 1 - - = . r-0 10 20 30 40 50 60 70 80 90 100 Weeks s ince end of p rev ious j o b o CM CD d O 00 o d o 6 o o Figure 10: Piecewise Constant Model of Re-Employment (Over 24 years) — Baseline Hazard Function — i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 — (1) Import Affected Declining Industry — — (2) Non —Import Affected Declining Industry • • • • (3) Import Affected Expanding Industry (4) Non —Import Affected Expanding Industry ° 0 10 20 30 40 50 60 70 Weeks since end of previous job 80 90 Figure 1 1 : Piecewise C o n s t a n t Model of Re —Employment (Over 24 years of age) — Basel ine Surv ivor Func t ion o • I 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 r j i i ~ ~*i i ' 1 1 tz 0 10 20 30 40 50 60 70 • 80 90 100 Weeks s ince end of p rev ious j o b VIII. A SELF SELECTION MODEL FOR RE- EMPLOYMENT EARNINGS. In this chapter we estimate the relation between the wage of the old job and the wage on the next job obtained by the respondent, given that the person is observed to find another job. In doing so one must allow for self-selection into employment, since disturbances to the level of the replacement wage are likely to affect the probability of re-employment. As in the previous chapter, our objective is to examine the relative and absolute performance of those separated from the trade affected declining industry group. 1. The Model. We assume that each individual has an unobserved reservation wage Vf[ , which is a function of the covariates x2i and a random disturbance U2i-Each period the jobless person is offered a job with wage W^ . The wage is assumed to depend upon covariates xXi and a random disturbance u-n. If this offer is at least as great as the reservation wage the lob is accepted and the observed wage W-x is equal to the offered wage. The two disturbances are assumed jointly distributed with a bivariate normal distribution. The model can therefore be written as off ^i = «ii Pi + "li ^i r S S = x 2 i fr> + u2i Wi = . . off r T o f f ^ r e s Wi Wi > W-1 0 , O.W. 113 Uj ~ N(0,2) . This specification corresponds to the type 2 Tobit model described in Amemiya (1985) and elsewhere. We impose the restriction that Z be diagonal. Using the estimated parameters from this model we can calculate the unbiased expected value of W-x for all respondents, whether re-employed or not, and thus the expected value of the replacement ratio for each sample group. The covariate vector x n consists of the variables used in the previous chapter, along with the duration of non-employment. The covariates x2i consist of non-employment duration, the dummies for being female, not the head of the family, and being single, and the wage in the previous job. The log-likelihood for this model is1: 2 0 1 ^ 2 , a i , o 2 ) = [mO *2ip2 - x l i P l H •i ) 4 - l n a i - t l Wl - X i i P i + lnO> W, * 2 i P 2 ®2 } where $ and (j) a r e t h e c . d . f . and p . d . f . r e s p e c t i v e l y of t h e s t a n d a r d normal d i s t r i b u t i o n , and where cJ 2 2 CTi+a2. A key variable of interest is the expected wage upon re-employment: the expression for this quantity is off EdVil&Ti > Wv ) = x n P J O-i (^(Ai) w h e r e Aj = X l i P l - x 2 i / 3 2 a <D(Ai) "_1 a Using this formula we are able to calculate the re-employment wage for those individuals who have not yet found employment. 1 See Maddala (1983, p.176). - 114 -2. Estimation. Estimation of the model above yielded large positive coefficients (between 0.3 and 0.4) for the effect of non-employment duration upon the (log) reservation wage. This result seems counter intuitive: one might expect reservation wages to fall as unemployment benefits run out. We suspected that the culprit might be simultaneity between non-employment duration and the disturbance term in the reservation wage equation: if, ceteris paribus, the reservation wage is high, then this will increase non-employment duration. This militates in favour of a model that simultaneously estimates wage and non-employment duration, along the lines of Addison and Portugal (1989). They adopt an instrumental variable approach, where in order to eliminate the dependence between one of the covariates (non-employment duration T) and the disturbance term one replaces the variable with its expected value. This quantity will be asymptotically uncorrelated with u. To form the expected value one then requires an estimated distribution for T: in our case we used the results from the estimation of the Piecewise Constant model of the previous chapter. 3. Empirical Results. In tables eighteen and nineteen below we present the results of estimating the instrumental variables Tobit model presented in the previous sections. The first of the two tables lists the parameter estimates. As one would expect, the replacement of employment duration by - 115 -its expected value reduced significantly the coefficient attached to this variable in the reservation wage equation: indeed in general it is no longer significant for either equation. The exception to this result is the fourth industry group. For displaced workers in these sectors one observes declining reservation wages and increasing wage offers over time: this is in accordance with standard non-stationary search models. The age variables appear to have insignificant effects upon wage offers: however this is not the case for the reservation wage equation, where the age-squared variable is significant for all equations. In figure seven we plot the relative effects of age upon the log reservation wage for each industry - note that the vertical axis measures the difference between log earnings at the given age and the log reservation wage for the mean age of the sample. The graph indicates that reservation wages rise most strongly with age for sample one: recall that this is what one would expect from the theory since the relative attractiveness of training is higher for workers in these industries. However the other group of declining industries does not display the same behaviour. With regard to the other covariates, one especially strong factor in the wage offer equation is the gender of the re-employed worker. Females are apparently offered significantly lower wages than male counterparts with similar observed characteristics. As with the duration data estimates, seasonally laid off workers fared better than average, whereas those who left because of illness or to study etc. fared worse. There was no significant difference between those laid off or quitting for other reasons. - 116 -Table 20: Estimates from Instrumental Variables/Type 2 Tobit Model of Re-Employment Earnings. Industry Sample (1) (2) (3) (4) Parameter 01: constant durunemp j ten j ten2 atlantic quebec prairies be female age age2 single nothead elem postsec univ mgrlprf owcllr whrs ahe cavrd layoff seasonal other 02 = constant durunemp female value 1.147 .021 -.045 -.049 -.054 -.151 -.169 -.524 -.212 -.010 .061 .024 -.064 -.003 .032 .272 - .226 .163 .035 .365 .099 .047 .295 -.159 .515 .042 .103 s. e. (.234)* (.037) (.077) (.034) (.380) (.065)* (.161) (.120)* (.087)* (.056) (.055) (.108) (.083) (.085) (.072) (.103)* (.109)* (-110) (.031) (.089)* (.076) (.069) (.156)® (.070)* (.413) (.072) (.171) value 1.53 -.009 .304 - .112 - .108 .008 -.011 .111 -.170 -.012 -.026 -.017 -.017 -.065 -.062 -.082 - .062 - .047 -.002 .249 -.072 .041 .043 -.084 .014 -.033 .082 s. e. (.109)* (.028) (.075)* (.024)* (.093) (.059) (.092) (.059)® (.074)* (.040) (.036) (.061) (.066) (.068) (.056) (.062) (.058) (.136) (.018) (.047)* (.044) (.051) (.066) (.060) (.333) (.048) (.135) value .979 .024 .081 -.036 -.053 - .069 - .114 .172 -.218 -.013 .011 -.026 .014 .014 -.104 .223 .108 .068 .007 .460 .057 .117 .240 -.135 -.676 .008 - .069 » s.e. (.198)* (.028) (.064) (.031) (.320) (.054) (.104) (.185) (.073)* (.054) (.051) (.063) (.071) (.185) (.057)® (.075)* (.084) (.078) (.024) (.085)* (.062) (.077) (.089)* (.067)® (.410) (.052) (.122) value s. 1.73(.136)* .083(.031)* -.094 (.052)® .007 (.023) -.209 (.142) -.148 (.050)* - .144(.091) -.205(.068)* -.186(.057)* .143 (.042)* -.128 (.035)* .065(.051) -.09K.059) .019 (.084) .159 (.053)* .160(.058)* .027 (.063) .098(.0"55)@ .038 (.021)® .218 (.061)* -.028(.048) -.073 (.053) .194 (.081)* -.215(.053)* .190(.350) - .122 (.064)® .152 (.115) - 117 -age age2 single nothead ahe vi1 .048 .437 -.296 .190 .377 1.655 (.098) (.127)* (.194) (.163) (.157)* (.083)* -.036 .193 .071 .185 .534 1.399 (.060) (.081)* (.105) (.109)® (.109)* (.108)* .260 .160 .026 .094 .981 1.616 (.095)* (.085)® (.115) (.120) (.155)* (.104)* .082(.068) .180(.084)* .108(.091) -.176(.100)@ .536(.118)* 1.557(.098)* CT; .688 (.211)* Mean LnL -0.61 Sample Size 817 - 0 . 6 9 1,558 •0.60 874 - 0 . 6 1 1,235 Table 21: Wage in Former Job and Expected Re-Employment Wage. Industry Job All Male Female Age < 30 Age > 30 Tenure < Tenure > 52 52 wks wks IAD old 8.44 10.94 6.61 7.22 9.08 7.89 9.90 new 6.64 8.01 5.79 6.68 6.62 6.36 7.46 NIAD old 10.48 10.80 8.17 8.46 11.50 10.12 14.04 new 8.32 8.30 8.54 9.61 7.74 7.79 15.97 IAE old 10.13 11.07 8.22 7.91 11.50 9.18 12.35 new 8.70 8.73 8.65 6.92 10.04 7.33 13.02 NIAE old 10.56 11.90 7.96 8.19 11.91 9.83 12.49 new 8.52 8.24 9.10 6.97 9.55 7-53 11.84 In table nineteen above we compare the average wage in the pre-separation job with the predicted wage in the new job. Overall, the expected re-employment wage is about two dollars lower for each industry - 118 -than that of the previous job2. Note the contrast with the actual re-employment wage: the latter variable was about the same as the previous wage (higher for sample one). This difference indicates that the mean of the wage for those actually re-employed at the time of the survey is an upwardly biased measure of the eventual re-employment wage for the entire sample. The figures indicate that laid off workers in any industry do not immediately recoup the losses from non-employment by obtaining higher wages in their new job. It is possible that the returns to tenure profile in the new job is very steep so that over the long run workers do not suffer much from displacement: however the large drop in wages suggests that older workers especially are likely to suffer significant losses over the long term, quite apart from the opportunity costs associated with non-employment. Especially noteworthy is the maintenance of wage differentials between the four industries: it is sometimes argued that displaced workers in low paying industries suffer less from displacement because they are likely to be re-employed in higher paying industries3: this does not appear to be the case here. Another interesting feature of table 19 is the difference between pre and post displacement for those with more than 52 weeks of tenure. For those in the Import Affected Declining industries there is a significant drop, whereas for the*other three industries there is little if any decline. This would tend to 2 Replacement rates were 0.78, 0.75, 0.83 and 0.78 respectively for each sample. 3 This argument is advanced in Glenday, Jenkins and Evans (1982). - 119 -indicate that losses from obselete firm specific human capital are concentrated in the Import Affected Declining industries. The empirical model in this chapter has assumed a diagonal covariance matrix for the errors in the reservation and offered wage equations. If this assumption is false it will bias our estimates. In particular, if the off-diagonal element G12 is not zero but is in fact positive, the conditional mean of the error term in the reservation wage equation will not be zero, but will rather be 0,12/0'12u1, where ux is the residual from the wage offer equation (see Maddala 1983, p.176). In our results we found that the predicted re-employment wage was lower than the observed wage. If this is due to the predicted offered wage being less than the observed offered wage, then this would tend to indicate that u± = F1-X1P1 > 0. Thus u2\ux is being assumed to be zero when it is in fact positive, and so (at the very least) , the constant term in the reservation wage equation is likely being over-estimated. This means that the predicted reservation wage is over-estimated also, and therefore the predicted wage also4. Thus to the extent that C712 is greater than zero, our estimates may under-estimate the wage loss. The problem then seems greatest when o~12 < 0. If this is true the our empirical results overstate the wage losses undergone by those who leave their jobs. Is this situation likely? We argue that the reverse is more likely: that someone who for some reason has a skill or other charateristic which makes them more desirable to an employer is probably going to have a higher reservation wage. To the extent that this is not picked up by the wage on the previous job, one would have CT12 > 0. 4Burdett and Ondrich (1985) show that if the wage offer distribution is log-concave then increases in labour demand will imply a rise in the observed wage. - 120 -4. Further Work. Although the model presented above attempts to allow for the effects of jobless duration upon re-employment earnings, duration is not modelled per se. The next logical step would therefore be to estimate a joint model for the duration of joblessness and re-employment earnings. One approach to specifying the relevant functional form would be to develop a complete structural model of optimal job search in which one makes assumptions about preferences and the nature of the offer distribution. Such models can be quite complex, especially when one allows for non-stationarity. An alternative strategy is to concentrate on a reduced form model: this is in keeping with the methodology of the previous chapters wherein we estimated reduced form specifications for re-employment. An example of this latter approach is as follows. One specifies a form for the hazard function: 0(fc) = 90(t) exp(-xY-v) and for the re-employment wage: w = z(3 + 6v + u, u ~ N ( o , a ^ ) . The random variable v is assumed to take on one of a finite number of values: {V\, V2, . .., Vj] . The distribution is such that E(v) = 0. The parameter 8 measures the dependence between joblessness duration and the re-employment wage. If S is positive then one will observe a positive correlation between jobless duration and re-employment wages, since high values of v imply not only higher values of w, but also longer non-- 121 -employment duration since an above average v shifts down the hazard rate for leaving non- employment. The likelihood function for uncensored observations would then be given by L(P,Y,G,5) = [f(t) r " t " §V P(v=V-}). j=l *• J Estimation would then proceed in the usual manner. - 122 -Figure 12: Relat ive Ef fects of Age on In Reservat ion Wages Tobl t Type 2 / IV Model o L O CM O CM u j o o o (1) Trade Affected Declining Industry — — (2) Non—Trade Affected Declining Industry (3) Trade Affected Expanding Industry • (4) Non—Trade Affected Expanding Industry m LO o 1 0 20 30 40 Age 50 60 70 IX. M EMPIRICAL MODEL OF SECTORAL MOBILITY. In this chapter we examine agents' decisions to move sector. Firstly we present summary statistics about the nature of the new jobs obtained by those who left or were laid off from jobs in one of the four industry groups. Secondly we estimate a probit model of sectoral mobility in order to identify those factors which influence the decision to change industry. This model also allows us to determine the extent to which the decision to move industry is correlated with becoming re-employed. 1. Summary Statistics. In table 1 below we present information about the former jobs of those agents who are observed to be re-employed. It is instructive to compare these results to those of table 6 in chapter V, which provides summary statistics for all those who left employment in the time period. In comparison with the whole sample, those observed to be re-employetl were younger and had shorter job tenure on their former jobs. This is consistent with the results of chapter VII. Those observed to be re-employed were also better educated. There was little difference in hours of work or salary. The duration of (completed) non-employment was of course much shorter for those observed to find new jobs: 16-20 weeks for those in samples 1, 2 and 3. This contrasts with an average uncompleted duration of 29-32 weeks for all - 124 -Table 22. Characteristics of Former Job of those observed to be Re-employed. Job Tenure: (in weeks) Region: Atlantic Quebec Ontario Prairies BC (reference) Gender: Male (reference) Female Age: Marital Status: Married (reference) Single Position in Family: Head (reference) Not Head Education: Elementary High School (reference) Post Secondary University Occupation: Managerial/Professional Other White Collar Blue Collar (reference) Weekly Hours Average Hourly Earnings (deflated by the CPI) If unionised or covered by a collective agreement: No (reference) Yes Reason for Leaving Job: Quit to Employment Quit to Unemployment Quit to Out of Lab. Force All 105.30 28.23 (1) 147.54 29.92 (2) 87.43 (3) 162.27 30.71 28.84 (4) 166.72 0 . 0 9 0 . 2 3 0 . 3 9 0 . 1 7 0 . 1 2 0 . 5 5 0 . 4 5 0 . 0 3 0 . 4 4 0 . 4 0 0 . 0 8 0 . 0 4 0 . 5 0 0 . 5 0 0 . 1 7 0 . 2 6 0 . 1 8 0 . 0 8 0 . 3 0 0 . 8 8 0 . 1 2 0 . 0 3 0 . 2 3 0 . 6 1 0 . 0 9 0 . 0 3 0 . 6 2 0 . 3 8 0 . 0 7 0 . 2 0 0 . 5 4 0 . 1 1 0 . 0 9 0 . 6 4 0 . 3 6 30.25 0 . 4 5 0 . 5 5 0 . 4 5 0 . 5 5 0 . 0 5 0 . 3 3 0 . 3 8 0 . 2 4 0 . 2 0 0 . 2 8 0 . 5 2 6 . 8 5 8 . 5 5 0 . 5 7 0 . 4 3 0 . 4 5 0 . 5 5 0 . 0 9 0 . 3 9 0 . 3 7 0 . 1 5 0 . 0 9 0 . 1 6 0 . 7 5 3 9 . 1 9 8 . 7 9 0 . 6 2 0 . 3 8 0 . 6 7 0 . 3 3 0 . 1 9 0 . 4 0 0 . 2 4 0 . 1 7 0 . 1 0 0 . 0 2 0 . 8 9 4 6 . 0 7 1 0 . 7 1 0 . 5 1 0 . 4 9 0 . 4 9 0 . 5 1 0 . 0 4 0 . 2 8 0 . 3 6 0 . 3 3 0 . 2 1 0 . 1 9 0 . 6 0 . 3 9 . 6 2 1 0 . 1 9 0 . 5 2 0 . 4 8 0 . 5 3 0 . 4 7 0 . 0 5 0 . 2 9 0 . 4 0 0 . 2 6 0 . 1 7 J ) . 19 0 . 6 4 3 8 . 9 8 9 . 8 2 0 . 7 9 0 . 2 1 0 . 2 2 0 . 1 8 0 . 1 6 0 . 6 4 0 . 3 6 0 . 1 9 0 . 1 9 0 . 1 4 0 . 7 6 0 . 2 4 0 . 1 3 0 . 0 9 0 . 1 0 0 . 7 4 0 . 2 6 0 . 2 6 0 . 1 7 0 . 1 4 0 . 7 0 0 . 3 0 0 . 2 2 0 . 2 1 0 . 1 3 - 125 -Non Seasonal Layoff (Ref.) 0.19 0.18 0.18 0.20 0.19 Seasonal Layoff 0.17 0.27 0.17 0.21 0.20 Other 0.09 0.04 0.32 0.02 0.04 Duration of Non-Employment 34.26 16.13 2.54 21.49 20.41 (in weeks) workers who left or lost their jobs. Those seasonally laid off or quitting were more likely to be observed in a new job. Table 21 below provides information about the nature of the new job, in particular the sector of the economy that the new job is associated with. The category "Same Industry" denotes those workers who were re-employed in the same industry as their old job, where industry is defined according to the 2-digit SIC code as before. Thus an individual who was laid off in Clothing and was re-employed in Textiles is counted as having changed industries. The category "other MMF" indicates that the respondent was re-employed in another mining, manufacturing or forestry industry. For those who were not formerly employed in any such industry this category is equivalent to "any MMF". From table 21 it is evident that the mobility of workers in the different samples varies significantly, with workers from sample three the most likely to be re-employed in another industry, and workers from sample two the least likely to move. What is striking is the extent to which agents do move sectors. For samples one, two and three fully half of those re-employed were re-employed outside of manufacturing or resources, with the bulk of these people going to services or trade. The vast majority of the new jobs were full time (90-95%) and very few people went into self employment (4-5%). - 126 -Table 23: Characteristics of New Job of those observed to be Re-employed. All (1) (2) (3) (4! Industry of New Job: Same Industry Other MMF Agriculture Construction TCOU Trade FIRE Services Public Administration Class of New Job: Paid Worker Unpaid Worker Self Employed Full Time/Part Time Status Full Time (> 30 hrs/wk) Part Time 0 . 3 7 0 . 1 2 0 . 0 2 0 . 1 0 0 . 0 6 0 . 2 0 0 . 0 5 0 . 3 4 0 . 0 6 0 . 9 5 0 . 0 0 0 . 0 5 0 . 7 9 0 . 2 1 0 . 2 9 0 . 2 2 0 . 0 1 0 . 0 7 0 . 0 2 0 . 1 3 0 . 0 2 0 . 2 0 0 . 0 3 0 . 9 5 0 . 0 0 0 . 0 5 0 . 9 0 0 . 1 0 0 . 4 9 0 . 1 1 0 . 0 3 0 . 0 8 0 . 0 4 0 . 0 9 0 . 0 1 0 . 1 0 0 . 0 4 0 . 9 4 0 . 0 0 0 . 0 6 0 . 9 5 0 . 0 5 0 . 2 2 0 . 2 7 0 . 0 1 0 . 0 8 0 . 0 5 0 . 1 4 0 . 0 2 0 . 1 7 0 . 0 3 0 . 9 5 0 . 0 0 0 . 0 5 0 . 9 3 0 . 0 7 0 . 3 1 0 . 1 9 0 . 0 1 0 . 0 7 0 . 0 3 0 . 1 2 0 . 0 3 0 . 2 2 0 . 0 2 0 . 9 6 0 . 0 0 0 . 0 4 0 . 9 0 0 . 1 0 Of course it must be borne in mind that the tables presented above refer to those whom we observe to be re-employed, and thus any conclusion about the behaviour of all those laid off or leaving employment will be biased. 2. A Bivariate Probit Model of Sectoral Mobility. Whether or not an agent decides to move industries is a discrete variable, and thus the Probit model is a natural choice for an empirical model. However, this procedure is complicated by the fact that the decision to move is observed only when the agent is re-employed, which leads to a sample selection problem if those not yet' re-employed are excluded from the analysis. - 127 -Thus in order to examine the link between re-employment status and industry mobility we estimate a model in which these two-variables are determined jointly. This is the Bivariate Probit model, the specification of which is as follows: 1, Yi* > 0 Xi = where Y,* = XV + Si, 0, o.w. ' „' F2* - where Y2* = xB + £2, 0, o.w. z v z I2 = 8 ~ JV(0,E) where I = 1 P P 1 and p > 0. The latent variable Yi* measures the respondent's propensity to be re-employed, and Y2* the propensity to remain in the same industry. These variables are determined by reduced form specifications that capture demand side and supply side factors. The covariate matrix X is the same as that used in the duration model of chapter VII. We do not include the duration of non-employment in X as this variable is likely to be jointly determined with the decision to change sectors. The log likelihood function for an individual i is given by lnLj = Pr(ri=0) lnO(-xy) + Pr(Irl) (Pr(J2=0) 0 (Xy, -xp, -rho) +Pr(I 2=D <p (Xy,xB,rho)) where <P is the univariate Normal c.d.f and <p is the bivariate Normal c.d.f. The results of estimating this model are presented in table 22 below. For our purposes it is the parameter estimates from the second equation that are'of the most interest. In all industries the coefficient on age was positive, however only in the case of the import affected declining industries was the parameter significant. The coefficient was also twice as large for these industries. This implies that older workers - 128 -Table 24: Covariate Estimates for Bivariate Probit Model. (1) value Parameter s.e. Equation 1 (Whether re-employed). (2) value s.e. (3) value s.e. (4) value s.e. const j ten jten2 atlantic quebec prairies be female age age2 single nothead elem postsec univ mgrlprf owcllr whrs ahe cavrd quit2emp quit2un quit2olf seasonal other .658 -4.671 1.576 -.154 -.342 .064 - .372 -.103 -.922 -1.754 .190 -.228 - .125 .064 .264 -.094 -.005 .059 1.824 .064 1.38 .570 -.420 -.273 .316 (.215)* (2.996) (12.151 (.651) (.144)* (.374) (.329) (.159) (.893) (5.80)* (.169) (.161) (.207) (.152) (.243) (.277) (.180) (.864) (2.33) (.147) (.275)* (.216)* (.171)* (.352) (.177)+ .737 -4.711 6.427 -.236 -.101 .105 .245 -.228 - .214 -9.876 - .112 -.087 .043 -.090 .206 .170 -.436 -.027 .378 -.129 .962 .338 -.361 .074 .391 (.163)* (2.931) (1.041) (.217) (.135) (.233) (.148) (.152) (.839) (4.119)* (.143) (.128) (.173) (.135) (.164) (.201) (.397) (.405) (1.211) (.117) (.238)* (.213) (.166)* (.143) (.193)+ .839 -3.91 7.525 - .210 -.390 -.149 - .075 .043 -2.112 -12.70 - .017 -.050 .041 .151 .367 -.282 -.069 .053 1.786 -.006 .573 .314 -.616 -.582 .159 (.210)* (3.182) (13.753) (.579) (.138)* (.305) (.344) (.152) (1.40)+ (5.861)* (.157) (.137) (.366) (.144) (.179)* (.149)+ (.182) (.781) (1.91) (.138) (.183)* (.207) (.183)* (.399) (.187) .985 -4.611 8.415 -.403 -.242 -.037 -.032 -.197 .072 -11.553 .083 -.060 - .031 .143 .167 .123 -.097 .047 1.346 -.246 .806 .235 -.602 .029 .214 (.216)* (2.265) + (7.512) (.318) (.130)+ (.246) (.190) (.133) (.759) (4.99)* (-125) (.131) (.248) (.135) (.160) (.188) (.167) (.600) (1.45) (.145)+ (.193)* (.162) (.157)* (.321) (.176) Equation 2 (Whether remained in same industry) const j ten j ten2 atlantic quebec prairies be female age age2 single nothead elem postsec univ mgrlprf owcllr whrs -.399 •2.774 6.21 .246 .557 -.022 -.187 .523 2.362 •8.174 -.121 -.045 .237 -.154 -.460 .440 -.341 2.84 (.822) (7.21) (22.247) (.862) (.193)* (.397) (.696) (.248)* (1.179)+ (16.828) (.216) (.287) (.325) (.200) (.316) (.325) (-325) (1.689) - .475 2.283 4.612 .082 .457 -.047 .471 -.563 1.293 -8.55 -.106 -.070 .249 - .197 .030 -.245 .014 .019 (.182)* (3.859) (19.834) (.236) (.162)* (.231) (.146)* (.186)* (.842) (4.769)+ ( 142) 143) 174) 149) 148) 187) 419) 395) -1.77 2.372 5.914 - .102 - .249 - .102 - .071 .250 1.266 -7.669 -.135 .337 -.271 .105 .288 - .213 - .339 .600 (.320)* (5.43) (28.781) (1.57) (.215) (.335) (.487) (.208) (1.81) (12.184) (.175) (.200) + (.713) (.233) (.235) (.201) (.244) (.971) -.745 ( -.08J) (2 _ 5.41 -.329 - .296 .050 -.267 -.066 1.221 -8.397 .107 .010 .053 .198 .364 .103 .513 .361 .232)* .561) (1.926) .416) .142)* .239) .226) .137) .864) (5.382) .131) .131) .248) .155) .173)* .168) .165)* .635) 129 -ahe 1.86 (3.187) 3.52 (1.147)* 3.21 (2.09) 1.21 (1.65) cavrd -.044 (.198) -.180 (.128) -.350 (.219) -.036 (.141) quit2emp -.006 (.750) .646 (.192)* 1.234 (.241)* .596 (.180)* quit2un -.388 (.421) .266 (.217) .065 (.338) .018 (.177) quit2olf -.254 (.724) .059 (.213) .403 (.362) -.060 (.223) seasonal .780 (.471) .013 (.157) -.485 (1.116) .414 (.330) other .186 (.455) .125 (.191) .962 (.264)* .346 (.186)+ rho -.541 (1.779) .889 (1.191) .661 (2.126) .858 (1.362) in this sample are relatively less likely to move industries, given re-employment. Another difference between the import affected declining industry group and the others was the coefficient on the quit-to-employment variable, which was close to zero for the former and positive and significant for the latter. This means that those make direct job-to-job transitions were less likely to stay in the same industry if they came from the import affected declining industry group. This finding is what we expect: since workers know the industry is declining those who leave of their own volition will be more mobile relative to those who wait to be laid off. The rho parameter proved, like many other parameters, to be insignificant. However there is a disparity between the import affected declining industry sample and the rest, because in the former case rho is large and negative, whereas in the latter cases it is large and positive. What the implies is that for sample 1 those who are, for some unobservable reason, less likely to be re-employed, 'are more likely to remain in their industry. In other words, those in the import affected declining industry sample who move industry are more likely to be re-employed, whereas for the other groups it is those who remain in the industry who are likely to be re-employed first. This is consistent with our view of declining industries as ones where the "best" (in the sense here of most employable) workers are relatively more likely to move industries. In an expanding or stable industry one might expect the - 130 -reverse: the best workers will wish to be re-employed in their current industry in order to retain their industry specific human Gapital: those who leave may be those who are not well matched to the industry. Indeed it is possible that there might be some sort of stigma effect associated with moving industries, if employers use this fact as a signal of lower quality. - 131 -X. CONCLUSIONS. The objective of this thesis was to examine the subsequent labour market experiences of those who were separated from import affected industries. In the theoretical portion we developed two models of the labour market. The first assumed a finite horizon and no (frictional) unemployment. The intent was to understand when the market response to increasing foreign competition might become sub-optimal from a social perspective, and therefore under what conditions there might be a rationale for government intervention - either in the form of tariff protection so that workers remained in their jobs, or displacement assistance for those exiting from declining industries. What we found was that it could be optimal to prevent plant closures in cases where high mobility costs meant that workers were earning considerable rents in their jobs relative to their next best alternative. Also affected would be those who quit the industry in the face of declining relative wages: although one could not justify tariffs merely to maintain wages (as opposed to preventing layoffs), there might be case for financial compensation of those affected. This would be particularly appropriate if the changes in trade pressure's were the deliberate outcome of government policy. An interesting feature of the model developed in chapter two was the implication of assuming a heterogeneous workforce. Over time as an industry declines there is a self-selection process going on, as the more mobile workers leave. This emphasis on heterogeneity was continued in chapter three, where we examined the labour market behaviour of agents - 132 -conditional on being laid off from a given industry. A search framework was adopted, so that agents might have differing non-employment spells depending upon how many offers they receive and which they choose to accept. We found the exit rate from non-employment into a new job would differ not only across industries but also over time for agents in the same industry. Thus it was possible for the re-employment rate to be higher for a depressed industry (with low wages and a low arrival rate of new offers) than for an expanding industry, at least initially, because mobile workers in the latter industry would be tempted to turn down jobs elsewhere in the knowledge that they could soon expect an offer of a job in their own industry. In the long run however exit rates would be lower for the depressed industry. We also showed that the responsiveness of an agent to the parameters of the UI system was greater in industries with higher wages. In chapters four and five we constructed the data set used in subsequent chapters. We assembled a group of four different industry samples according to the employment history of the industry and the vulnerability to foreign competition. Of particular interest was sample one, consisting of four import affected declining industries: leather, textiles, clothing and primary metals. All four have been the foeus of government policy in the past. This group was notable for being somewhat lower paid on average than the other three groups, and also employed a higher proportion of females. The theoretical model in chapter three led us to expect that those quitting from a declining industry would be of higher quality than those laid off. This was somewhat confirmed by the data: .whereas in the - 133 -expanding industries there was little difference between the two groups, in the declining industries those who quit were earning higher wages than those laid off, even allowing for other characteristics. The seventh and eighth chapters were devoted to an empirical analysis of labour market behaviour after displacement. It was found that the sample of individuals who lost or left jobs in trade affected declining industries fared worse than their counterparts in other industries, both in terms of the length of non-employment, and in terms of their eventual re-employment wages. This group also showed no increase in the probability of finding a job after the likely end period for UI: in light of our model in chapter three we interpreted this as implying that labour market conditions for this group were bad enough that individuals would not postpone job search until benefits ran out, nor turn down suitable jobs even though the receipt of UI would subsidise unemployed job search. Older workers in these industries also displayed relatively high reservation wages compared to their younger counterparts: our explanation of this was that the individuals believed that costly retraining would be required for the new jobs they were offered, and thus re-employment was relatively less attractive to them than to their younger colleagues. In absolute terms our results confirm those of earlier researchers who found significant losses from joblessness. After correcting for bias due to censoring we found that mean completed joblessness was about a year, and that the mean re-employment wage was 75-80 percent of the former wage. The latter result should be tempered by the fact that we do not estimate the extent to which the re-employment' wage may have - 134 -subsequently caught up with the wage the worker would otherwise have been earning. Finally in chapter nine we examined the type of job obtained after separation from an industry, and the relation between re-employment and the decision to change industries. We found a large degree of inter-industry mobility on the part of those who were observed to be re-employed. After estimating a model which allows for simultaneous determination of re-employment and industry change, we found some evidence that for those in the import affected declining industry group re-employment was positively correlated with changing industry, while for the other industries the correlation was negative. Unfortunately these effects were statistically insignificant, owing perhaps to the poor performance of the Bivariate Probit model. Our overall conclusion is that those displaced from import affected declining industries fare poorly, in both absolute terms and relative to other similar industries. The question then becomes: why? Demographic differences between the industries were not able to account for the result: instead we place the blame upon mobility costs. Our contention is not that these costs differ across industries, but rather that even when they are identical for all workers, the impact will be different for those in expanding industries who do not "need" to move and those for whom finding employment in the industry for which they have skills and experience is increasingly different. In our theoretical work we found that different adjustment costs for workers in the same industry could actually exacerbate displacement - 135 -costs, because the existence of some workers with low costs keeps the wages in the industry higher ensuring that all have more to lose. We found evidence that some workers in the import affected declining industries did indeed find a job quickly: the hazard function for the first month or so for this group was not too dissimilar from the other groups. We also found weak evidence from the Probit model that re-employment was positively associated with finding a job only for the former group, which tends to indicate that sectoral mobility is particularly important for reducing non-employment duration and thus displacement costs for this group. These hypotheses do not however explain why those from the import affected industries should have had a harder time than those from non-import affected declining industries. Part of the reason may be the particular nature of this latter group, which is composed mainly of forestry workers. It may be that the availability of seasonal work in this industry means that non-employment spells, although more frequent, may be of shorter duration. This effect would not necessarily be eliminated by excluding those who left for seasonal reasons, because those permanently laid off in forestry may find it easier to find work than those in textiles because of the availability of seasonal or*other temporary employment. Because our approach models duration given separation from a job, rather than frequency of separation per se, this sort of intermittent employment would show up as shorter average non-employment duration, even though over the course of a year a worker from each industry might have the same total non-employment. - 136 -There are other possible explanations for the rents which we appear to observe, and these explanations may have different implications for policy. For example, if the rents are due to trade unions, which maintain wages above the market clearing level, then the elimination of these rents through foreign competition may be thought desirable. This is not necessarily the case however; Mezzetti and Dinopolous (1991) show that the welfare effects of freer trade depend upon how much unions value employment relative to wages. Also, unionisation rates were less than 50% for all the industry groups studied, although it is possible that unionised firms influence strongly the wages at non-union firms. Another framework for examining the welfare implications of employment reductions due to imports is the implicit contracts paradigm, in which firms insure their risk neutral employees by providing a constant wages, which in turn leads to layoffs if demand is low. In Matusz (1986) , the author uses this framework to examine the effects of international trade. He shows that the move to freer trade may increase unemployment or reduce real wages. However, expected utility will always increase because increases in unemployment are offset by higher wages, and lower wages are accompanied by an increase in employment security. In our empirical work though, we did not find an increase in wages in the industry with longer non-employment durations. It is possible however that once re-employed the employment security of those exiting import affected firms was increased: we do not have the data to examine this possibility. A third way of modelling labour markets that allows for the existence of rents is the efficiency wage paradigm, in • which rents are - 137 -used a device to induce effort on the part of a firm's workforce. In this case, Bulow and Summers (1986) have shown that freer trade can reduce wages and employment in affected industries and reduce overall societal welfare. Thus the predictions and policy conclusions are not dissimilar from our theoretical model, although the source of rents is very different. From a policy perspective our results point to possible cause for concern, because they indicate that costs of displacement are high for many workers, at least in the short run. Ideally one might like to target aid at those most affected. However if our hypothesis about the importance of mobility costs is true, then it may be hard to precisely target individuals with a lot to lose. This is because it is unobserved costs that are important in leading to rents and thus displacement costs: observed costs will simply be reflected in lower wages and thus lower displacement costs - at least in a competitive market. Even if aid cannot be precisely targeted, the magnitude of the losses that we have found for those displaced in certain industries does mean that some workers are hit particularly hard by changes in the international trading regime. To the extent that these changes are "due to deliberate policy choices such as a high exchange rate or trade agreements that reduce protection for these industries, perhaps those who benefit from these changes should compensate those who lose. Whether these losses outweigh the benefits of these policies is another question: however it is worth noting that the displacement costs we identify are often ignored when these cost-benefit analyses are done1. Other factors 1See for example Hazeldine (1990) . He uses five different models to asses the - 138 -that need to be taken into account when deciding upon policy responses are the effects upon the distribution of income: if workers in import affected industries are relatively well paid, then policies that help these individuals at the expense of less well paid persons will have redistributional effects that may be undesirable. There are three main directions for further research. Firstly a more sophisticated econometric technique could be used to better integrate the estimation of post-separation joblessness and the wage paid on a new job. In chapter eight we discuss how such a technique would be implemented. Secondly, it would be useful to provide a better characterisation of what mobility costs actually comprise: as well as retraining costs there are other factors such as whether one would need to move to another part of the country that would need to be taken into account. This would also allow us to decide whether the losses that we observe are indeed due to these mobility costs, or whether instead they are due to other factors such as rents in the job they have just left. Thirdly there is the question of what happens several years after separation. Are workers still paid less than they would have had they remained in their original industry or are the effects of the separation temporary? A thorough investigation would require not only a data set with a longer time frame, but also a proper model of the returns to job tenure, both on the old job and the new. costs and benefits of the U.S.-Canada FTA, but not one incorporates the possible losses to employees. This is especially odd given the strong opposition to the FTA and to NAFTA by organised labour. - 139 -BIBLIOGRAPHY. Abowd, J.M. and T. Lemieux, "The Effects of International Competition on Collective Bargaining Outcomes: A Comparison of the United States and Canada." in J.M. Abowd and R. Freeman (eds.) Immigration, Trade and the Labor Market, Chicago: University of Chicago Press (1991), pp.343-367. Addison, J.T. and P. Portugal "Job Displacement, Relative Wage Changes and Duration of Unemployment", Journal of Labor Economics 7, no.3 (July 1989) , pp.281-302. Becker, G. "Investment in Human Capital: A Theoretical Analysis." Journal of Political Economy 70, no.l (February 1962), pp.9-49. Blanchflower, David "Fear, Unemployment and Pay Flexibility," Economic Journal 67, no.2 (May 1991), pp.483-96. Blau, D.M. "An Analysis of Employed and Unemployed Job Search Behavior.", University of South Carolina Working Paper. Presented at Econometric Society Annual Winter Meeting, New Orleans, (December 1986) . Blau, D.M. and P.K. Robins, "Job Search Outcomes for the Employed and the Unemployed," Journal of Political Economy 98, no.3 (June 1990), pp.637-655. Bulow J. and L. Summers "A Theory of Dual Labor Markets with Applications to Industrial Policy, Discrimination and Keynesian Unemployment." Journal of Labor Economics 4, no.3, pt.l (July 1986) Burdett, K. and J. Ondrich, "How Changes in Labour Demand Affect Unemployed Workers," Journal of Labor Economics 3, no.l (January 1985), pp.1-10. Christofides, L.N. and C.J. McKenna, "Employment Tenure in Canada.", paper presented to the Canadian Employment Research Forum conference, Aylmer, Quebec, (March 19 92). Canadian Labour Market and Productivity Centre, "The Role of Business-Labour Sectoral Initiatives in Economic Restructuring.", Quarterly Labour Market and Productivity Review (1992), no.1-2. Carrington, W.J. and Asad Zaman, "Interindustry Variation in the Costs of Job Displacment." Journal of Labor Economics 12, no.2 (April 1994), pp.243-275. Doiron, D.J. "Layoffs as Signals: the Canadian Evidence." mimeo. Department of Economics, University of British Columbia (November 1993) . - 140 -Economic Council of Canada, "Adjustment Policies for Trade-Sensitive Industries." Research Report, Ottawa: Minister of Supply and Services (1988) . Fallick, B.C. "Unemployment Insurance and the Rate of Re-employment of Displaced Workers", Review of Economics and Statistics (1991) pp.228-35. Fallick, B.C., "Job Security and Job Search in more than One Labor Market," Economic Inquiry XXX, no.4 (October 1994), pp.742-745. Farber, H.S. "The Incidence and Costs of Job Loss: 1982-1991." Industrial Relations Section Working Paper No.3 09, Princeton University (January 1993) Follman, D.A., M.S. Goldberg and L. May, "Personal Characteristics, Unemployment Insurance and the Duration of Unemployment." Journal of Econometrics 45, (1990), pp.351-366. Gibbons, R. and L.F. Katz, "Layoffs and Lemons." Journal of Labor Economics 9, no.4 (October 1991), pp.351-380. Glenday, G., G.P. Jenkins and J.C. Evans, Worker Adjustment Policies.- An Alternative to Protectionism, Ottawa: The North-South Institute (1982) . Government of Canada, A Review of Canadian Trade Policy, Ottawa: Minister of Supply and Services (1983). Gray, D. and G. Grenier, "The Determinants of Jobless Durations of Displaced Workers in Canada.", mimeo, Department of Economics, University of Ottawa (October 1993) . Green, D.A. and W.C. Riddell, "Qualifying for Unemployment Insurance: An Empirical Analysis.", Mimeo, Department of Economics, University of British Columbia (September 1993). Hazeldine, T., "Why do the free trade gain numbers differ so much? The role of industrial organisation in general equilibrium." Canadian Journal of Economics, XXIII, no.4 (Novemeber 1990), pp.791-806. Jacobson, L.S., R.J. LaLonde and D.G. Sullivan, "Earnings Losses of Displaced Workers." American Economic Review 83, no.4 (September 1993), pp.685-709. Jones, S.R.G. and P. Kuhn, "Mandatory Notice and Unemployment." Department of Economics Working Paper No. 92-15, MacMaster University (July 1992). Jones, S.R.G. and W.C. Riddell, "The Measurement of Labour Force Dynamics with the Labour Market Activity Survey: The LMAS filter.", Department of Economics Discussion Paper No.91-17, University of British Columbia (June 1991). - 141 -Kletzer, L., "Industry Wage Differentials and Wait Unemployment," Industrial Relations 31, no.2 (Spring 1992). Kruse, D.L., "International Trade and the Labor Market Experience of Displaced Workers.", Industrial and Labor Relations Review 41, no.3 (April 1988), pp.402-417. Lancaster, T., The Econometric Analysis of Transition Data, Cambridge: Cambridge University Press (1990). Lawrence, C , and R.Z. Lawrence, "Manufacturing Wage Dispersion: An End Game Interpretation," Brookings Papers on Economic Analysis 1985:1, pp.47-116. Lindbeck, A. and D.J. Snower, The Insider-Outsider Theory of Employment and Unemployment, Cambridge, Mass.: MIT Press (1988). McBride, S. Not Working: State, Unemployment, and Neo-Conservatism in Canada, Toronto: University of Toronto Press (1992). Matusz, S.J., "Implicit Contracts, Unemployment and International Trade," Economic Journal 96, no.2 (June 1986), pp.307-322. Meyer, B.D., "Unemployment Insurance and Unemployment Spells.", Econometrica 58, no.4 (July 1990), pp.757-782. Mezzetti, C. and E. Dinopolous, "Domestic unionization and import competition," Journal of International Economics 31 (1991), pp.79-100. Mortensen, Dale T., "A Structural Model of Unemployment Insurance Benefit Effects on the Incidence and Duration of Unemployment." in Y. Weiss and G. Fishburn (eds.) Advances in the Theory and Measurement of Unemployment, London: Macmillan (1990), pp.57-81. Osberg, L. "Unemployment and Inter-Industry Mobility in Canada in the late 1980s." Department of Economics Working Paper No. 88-04, Dalhousie University (June 1988) . Picot, G. and W. Pyper, "Permanent Layoffs and Displaced Wojrkers: Cyclical Sensitivity, Concentration and Experience following the Layoff." Analytical Studies Branch Research Paper No. 55, Statistics Canada (1993) . Podgursky, H. and P.Swain, "Job Displacement and Earnings Loss: Evidence from the Displaced Workers Survey", Industrial and Labor Relations Review 41, no.l (October 1987), pp.17-29. de la Rica, S. "Displaced Worker in Mass Layoffs: Pre-displacement Earnings Losses and the Unions Effect." Industrial Relations Section Working Paper No.303, Princeton University (June 1992). Ruhm, C.J., "Are Workers Permanently Scarred by Job' Displacements?" American Economic Review 81, no.l (March 1991), pp.319-324. - 142 -Weiss, A. "Job Queues and Layoffs in Labor Markets with Flexible Wages." Journal of Political Economy 88, no.3 (June 1980), pp.526-38. - 143 -

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0088379/manifest

Comment

Related Items