Printer Friendly

Job changing after displacement: a contribution to the advance notice debate.

I. Introduction

Considerable research effort has been devoted to modeling the effects of advance notice on postdisplacement outcomes within the framework set by the biennial Displaced Worker Surveys. Yet the mechanisms through which advance notice achieves the benefits attributed to it have been little explored in the literature. The standard argument that notice facilitates job finding by |bringing forward' search is of course consistent with the empirical conclusion from "first generation" studies that (broadly defined) notice reduced postdisplacement joblessness but had no discernible impact on postdisplacement wages.[1] But as an explanation it fails to address the empirical suggestion of time varying effects[1; 8] while glossing over the sensitivity of the unemployment result to sample construction and model specification. The picture has become more clouded with the publication of results from "second generation" studies using more recent data that are virtually the reverse of those first reported. That is to say, analyses of the 1988 and 1990 Displaced Worker Surveys report that that extended intervals of notice appear at best to be associated with minor reductions in joblessness but to lead to substantive earnings improvement[5; 7; 10; 11; 12].

There is, then, a pressing need to inquire into mechanisms. The present paper is offered as a contribution to this line of investigation, focusing as it does on job changing in the aftermath of the displacement event. We investigate the correlates of job changing in the wake of displacement and the relationship between number of postdisplacement jobs and end-of-survey earnings, advance notice being a determinant of both turnover and earnings.

To anticipate our results, we find evidence to suggest that formally notified workers, though they do not experience lower unemployment than their non-notified counterparts or have better overall reemployment prospects do appear to make better job matches and thereby enjoy higher relative earnings. This particular finding holds for written notice of more than two months' duration in equations estimated over a sample of all workers. Disaggregation of the data by reason of job loss suggests that this overall result masks potentially important differences among subgroups of the population either in their receptivity to the information conveyed by notice or because of the "quality" of notice provided.

II. Data and Estimation Procedure

Our data are taken from the 1988 Displaced Worker Survey (DWS), conducted as a supplement to the January Current Population Survey of that year. Since the basic format of the 1988 survey closely resembles that of its 1984 and 1986 precursors, we confine our remarks here to that which is new in the third survey, to the jobs variable, and to the restrictions imposed on the data.

The principal departure of the 1988 DWS from the earlier surveys is that it contains some indication of the extent of notice and not simply its incidence as heretofore. Thus, in addition to the composite or broadly defined notice question inquiring of workers whether they "expect[ed] a layoff or had received advance notice of a plant or business closing," the 1988 survey goes on to ask of those who replied in the affirmative whether they had received formal written notice and, if so, the length of notice provided. Weeks of notice are coded not in continuous form but rather in three grouped intervals: less than one month, between one and two months, and greater than two months. We may thus distinguish between what we shall term informal notice (INFONOT) and the three categories of written notice (WRITNOT1, WRITNOT2, WRITNOT3). INFONOT covers all those who responded to the composite notice question in the affirmative less those who subsequently indicated that they had received written notice.

The second innovation in the 1988 survey is that it provides information on the duration of the single spell of joblessness immediately following the displacement event for those who subsequently located a job. In the previous surveys the reported length of unemployment could include multiple spells of joblessness, and hence reflect multiple job holding, if these were associated in the mind of the respondent with the displacement event. In other words, it is now possible to examine the causal link between postdisplacement search and job transitions. This possibility is precluded for the 1986 DWS (which was the first survey to include a number of jobs held variable) because of the contamination of the jobless duration measure by job turnover.

Turning therefore to the number of jobs reported in the wake of displacement, the 1988 DWS identifies some six job holding categories, namely, 0 to 5 (or more) postdisplacement jobs. In what follows, we shall group jobs 3 to 5 within a single category. Unfortunately, there are signs of reporting/coding error in the responses to the job holding question. Thus, for example, it emerged that some workers who were classified as having 0 jobs since the displacement event could nevertheless be identified as employed on the basis of the January 1988 Current Population (CPS). To tackle this problem, it was decided also to use information from another question in the DWS that identified those who had claimed "never to have worked" since displacement, and which evinced a smaller reporting/coding error. A worker was duly classified as a 0 job holder if he or she claimed to be such in response to the number of jobs held question and replied in the affirmative to the relevant never found work question and was not otherwise reported as employed as of end-of-survey on the basis of the parent January 1988 CPS. Those respondents claiming at once never to have found work but yet reporting positive job holding were purged from the sample.

Although we wish to account for the behavior of those currently economically active workers who never found work (and passed the consistency test in this regard), as a purely practical matter it was decided to exclude those bona fide positive job holders who were currently unemployed as of January 1988. This exclusion was imposed to avoid the problem of having to model, with inadequate data for the purpose, two types of selection into reemployment status, namely, that of workers who were continuously unemployed following displacement, on the one hand, and that of workers who, though they found work following displacement, subsequently lost or left their jobs, on the other.

The remaining restrictions imposed on the data were as follows. Because the nature of displacement is not well defined for certain individuals and sectors, those employed part-time and in agriculture at the point of displacement were excluded, as were those who were aged above 65 years as of January 1988. Similar reasoning explains the exclusion of the self-employed together with those classified as displaced for "seasonal" or "other" reasons. Thus, our sample of displaced workers comprises those separated by reason of plant closing or relocation, slack work, and abolition of shift or position. Since there are grounds for believing that workers displaced by layoffs (i.e., the slack work and abolition of shift categories) may differ materially from their counterparts dislocated by reason of plant closings (all workers rather than a subset of them are "canned" in the latter case), we also allow for disaggregation of the data along these lines.(2) Descriptive statistics on the continuous and dichotomous variables employed in this inquiry are provided in Appendix Table I.
Table I. The Effect of Advance Notice on Job Changing, Ordered
 Probit Estimates
 Sample

Variable All Workers Plant Closings Layoffs

Constant -.068 .072 .264
 (.285) (.381) (.431)
DURATION -.002(*) .001 -.004
 (.001) (.002) (.002)
INFONOT .098(**) -.201(***) .026
 (.049) (.068) (.073)
WRITNOT1 .147 .102 .184
 (.097) (.142) (.136)
WRITNOT2 -.261(**) -.262(*) -.369(*)
 (.122) (.153) (.223)
WRITNOT33 -.316(***) -.316(**) -.428(*)
 (.122) (.146) (.237)
EDUC -.006 -.011 -.003
 (.011) (.015) (.017)
AGE -.015(***) -.016(***) .041(***)
 (.003) (.004) (.004)
TENURE -.006 -.007 .001
 (.005) (.006) (.008)
InOLDWAGE .046 .087 .014
 (.051) (.073) (.073)
UNEM -.021 -.037(**) -.002
 (.012) (.016) (.018)
WHITE .099 -.237(**) .048
 (.072) (.100) (.106)
MALE -.033 -.018 -.083
 (.055) (.076) (.082)
MARRIED .136(***) .136(*) .136(*)
 (.053) (.074) (.079)
KIDS -.006 .006 -.022
 (.023) (.030) (.036)
BLUE .181(***) .210(***) .145(*)
 (.055) (.075) (.083)
MANU .017 -.023 -.050
 (.052) (.073) (.075)
HEALTHIN .114(**) -.120 -.106
 (.054) (.074) (.081)
CLOSE -.013
 (.051)
ABOLISH -.003
 (.076)
YEAR83 1.090(***) 1.154(***) 1.026(***)

 (.083) (.119) (.121)
YEAR84 .920(***) .862(***) .996(***)
 (.076) (.106) (.113)
YEAR85 .737(***) .696(***) .792(***)
 (.071) (.099) (.107)
YEAR86 .480(***) .411(***) .558(***)
 (.072) (.103) (.105)
SMSAI .143(**) .235(**) .025
 (.068) (.093) (.105)
SMSA2 .068 .076 .061
 (.055) (.078) (.082)
SMSA3 -.252(***) -.330(***) -.149(***)
 (.077) (.104) (-.119)
NORTHE -.050 -.064 -.031
 (.075) (.100) (.117)
NORTHC .127(**) .112 .147(**)
 (.061) (.085) (.091)
WEST .183(***) .139 .229(**)
 (.062) (.085) (.092)
[mu] .781 .805(***) .763(***)
 (.026) (.036) (.038)
Log-likelihood -2716.3 -1450.3 -1252.6
[mu] 2967 1608 1359

Notes: asymptotic standard errors in parentheses
*, **, *** denote significance at the .10, .05, and .01 levels,
respectively


The analytical techniques employed are relatively straightforward. Investigation of the determinants of job turnover is conducted by means of an ordered probit equation estimated over the three positive job holding categories. The ordered probit was selected in preference to the multinomial logit for reasons of tractability in subsequently exploiting the number of jobs variable in the wage equation (see below). The ordered probit involves the latent variable "tendency to switch jobs." A positive coefficient on a notice variable in the ordered probit indicates that the particular class of notified worker is more likely to be at one of the higher values of the dependent variable (i.e., number of jobs), and conversely in the case of a negative coefficient.(3) One indication of a good job match is of course its longevity. If notification produces better job matches, then we would expect it to be negatively associated with job turnover.

Our postdisplacement wage equation includes, in addition to the four dummy variables identifying each type of notice and other conventional correlates, a variable representing the actual number of postdisplacement jobs,(4) a control for unobserved individual heterogeneity, and a selectivity argument. Each requires some comment. The principal innovation in the wage equation is of course the inclusion of the job changing variable. If notice of some interval enables workers to achieve better job matches then its coefficient estimate should be positive and that for job turnover should be negative. The justification for including the number of subsequent jobs is that it renders the job matching process more transparent, given that the survey only provides postdisplacement earnings on the last job as of end-of-survey date (namely, January 1988). More specifically, it is possible to retrieve the direct and indirect effects of written notice from the coefficient estimates for notice and number of jobs. The direct effect can be obtained by simply reading off the coefficient estimate for the relevant notice interval. It identifies the value of the initial job match made possible by notice, via extended search and/or on-the-job subsidized search. The indirect effect, which captures the longer-lasting nature of that job match, is obtained from the ordered probit and the coefficient estimate for the actual number of jobs in the wage equation. Specifically, it is the product of the partial derivative of the expected number of jobs with respect to the relevant notice variable and the latter coefficient.(5) Note that since the expected number of jobs is not linearly related to the regressors in the ordered probit one cannot compute the indirect effect of notice simply by multiplying the coefficient estimate for the number of jobs variable reported in the wage equation by the coefficient estimate for the notice argument in the ordered probit.

Second, the postdisplacement wage equation attempts to control for unobserved individual heterogeneity by using the residual from an estimated predisplacement wage equation (see Appendix Table II) as a separate regressor.(6) The estimated residual is the unexplained variation in the individual worker's predisplacement earnings and is orthogonal by construction to the explanatory variables. As we shall see, the coefficient estimate for the residual in the postdisplacement wage equation is both positive and highly significant, as would be expected if the residual is related to ability.
Table II. Selectively-adjusted Postdisplacement Wage Equations

 Sample

Variable All Workers Plant Closings Layoffs

Constant 4.123(***) 4.209(***) 3.813(***)
 (.111) (.141) (.174)
DURATION -.004(***) -.004(***) -.004(***)
 (.001) (.001) (.001)
NUMBERJOBS -.059(***) -.082(***) -.036(**)
 (.012) (.017) (.018)
INFONOT .043(**) .059(**) .024
 (.020) (.027) (.031)
WRITNOT1 .042 .057 .025
 (.035) (.047) (.052)
WRITNOT2 .009 .023 .024
 (.048) (.061) (.076)
WRITNOT3 .108** .169(***) -.066
 (.045) (.051) (.089)
EDUC .074(***) .068(***) .089(***)
 (.005) (.006) (.007)
AGE .002(*) .001 .003
 (.001) (.002) (.002)
TENURE -.003 -.003 -.003
 (.002) (.003) (.003)
RESIDUAL .553(***) .514(***) .604(***)
 (.031) (.033) (.050)
UNEM .004 -.005 .009
 (.006) (.007) (.008)
WHITE .090(***) .167(***) .063
 (.033) (.045) (.046)
MALE .362(***) .361(***) .360(***)
 (.022) (.030) (.033)
MARRIED .055(***) .038 .080(***)
 (.021) (.028) (.031)
BLUEC .064(***) .061(*) .063(*)
 (.023) (.031) (.035)
MANUC .109(***) .131(***) .075(***)
 (.020) (.029) (.029)
HEALTHINS .215(***) .206(***) .234(***)
 (.022) (.030) (.033)
CLOSE .013
 (.023)
ABOLISH .025
 (.031)
YEAR83 .057 .174(**) .103
 (.076) (.083) (.115)
YEAR84 .013 .116 .054
 (.071) (.080) (.104)
YEAR85 .019 .097 .099
 (.066) (.076) (.097)
YEAR86 -.034 .015 .032
 (.054) (.061) (.079)
SMSA1 .096(***) .132(***) .063
 (.029) (.037) (.045)
SMSA2 .246(***) 270(***) .221(***)
 (.023) (.032) (.035)
SMSA3 .379(***) .386(***) .371(***)
 (.032) (.042) (.049)
NORTHE .032 -.018 .056
 (.030) (.037) (.046)
NORTHC -.059(**) -.079(**) -.045
 (.026) (.035) (.038)
WEST .065(**) .071(**) .067(*)
 (.025) (.034) (.039)
[lambda] -.294** -.048 -.204
 (.129) (.172) (.153)
[R.sup.-2] .453 .451 .457
n 2728 1483 1245
Notes: standard errors in parentheses
*, **, *** denote significance at the .10, .05, and .01 levels,
respectively


Third, the postdisplacement wage equation incorporates the standard correction for selection into reemployment. The auxiliarily probit equation is provided in Appendix Table III.

[TABULAR DATA OMITTED]

Finally, the equation also includes a predicted duration of unemployment variable. The individual worker's expected duration of joblessness was computed from a flexible specification of the accelerated failure time model, namely, the three-parameter extended generalized gamma distribution[2]. Use of a predicted duration of unemployment variable may be justified on the grounds that it provides more information on the constant reservation wage assumption of the standard search model than is permitted by inspection of the lambda coefficient alone. The duration equation is given in Appendix Table IV.(7)

[TABULAR DATA OMITTED]

III. Findings

The effect of notice on the number of jobs held following displacement is reported in Table I for the whole sample and also for the disaggregations based on reason for job loss. Beginning with the all worker sample, it is apparent that all types of notice with the exception of the shortest interval of written notice are associated with reductions in subsequent job turnover. The influence of informal notice is weak compared with that of the two longest intervals of written notice. The coefficient on the longest notice interval is particularly strong and well determined. There are few surprises with respect to the other correlates of job changing. Thus, for example, older workers and married workers evince lower job changing, while turnover is higher among blue-collar workers. Interestingly, the length of the spell of unemployment following displacement is only weakly associated with reduced turnover.

A somewhat different pattern of results obtains once we disaggregate by reason for job loss. In particular, although extended written notice is associated with reduced job changing across both samples its effects are noticeably less well determined for layoffs than plant closings. Note, too, the positive albeit insignificant coefficient estimate for informal notice among those displaced by layoffs. Differences between the results reported for the plant closing and layoff samples with respect to the role of advance notice may in part reflect unrealistic expectations of recall among members of the latter group, leading them to locate "first" jobs that they view as temporary.

We preface our discussion of the impact of notice on postdisplacement earnings by noting a potential problem associated with our use of the actual number of jobs rather than the predicted value from the ordered probit: the estimates of the wage equation will be consistent only if the error terms of the number of jobs and the wage equations are orthogonal. We tested for orthogonality using a modified Hausman test. Specifically, the postdisplacement wage equation(s) was reestimated using all of the arguments documented in Table II plus the residuals from the ordered probit equation as a separate regressor. The coefficient estimates for the latter regressor were never statistically significant, suggesting the absence of correlation between the residuals from the two equations. The results of this test procedure are not provided here but are available from the authors on request.

Having established that this problem is not pressing, we next consider the impact of notice on postdisplacement earnings, beginning with the results for the all-worker sample contained in the first column of Table II. The longest interval of written notice appears to be associated with a substantive improvement in relative earnings in the order of 11 percent. Shorter intervals of written notice do not "yield" higher earnings, although the small, positive coefficient estimate for informal notice is significant at conventional levels. For its part, heightened job turnover serves to reduce earnings. This variable negatively proxies the role of tenure on the current job for the generality of workers and in part captures the indirect effect of notice on earnings. Other things being equal, longer postdisplacement joblessness is associated with lower reemployment earnings for reasons that presumably have to do with declining reservation wages and/or human capital depreciation and stigma effects. That said, there is also evidence of negative selection, at least for this level of aggregation, which would suggest that those currently unemployed have higher reservation wages than their reemployed counterparts.

The disaggregated results presented in the last two columns of the table indicate that the allworker regression masks important differences between layoffs and plant closings. In particular, the already weak evidence of modified job changing behavior among notified laid-off workers is underscored by their earnings development after displacement. For the plant closing sample the positive impact on earnings of the longest interval of written notice is both sizeable and highly significant, whereas for laid-off workers the corresponding coefficient estimate is negative. For neither sample are shorter intervals of written notice productive of income, although the small positive coefficient estimate for informal notice retains significance in the plant closing sample.

The pattern of many of the other correlates of postdisplacement wages is consistent across the subsamples. Interestingly enough, the evidence of negative selection earlier reported for the all-worker sample is no longer apparent at the disaggregated level, a result that is in conformity with other studies that also include a jobless duration argument in the postdisplacement wage equation [3].

The broad conclusion is that, although notice can apparently promote better job matches, its effects hinge not only on the length of written notice but also on reason for job loss. In a concluding section we will address possible reasons for these disparate results. As noted earlier, the total effect of notice on earnings development is the result of, first, an initial job match effect, and, second, a turnover effect resulting from that match. The direct and indirect effects of notice are charted in Table 111. The main conclusions of this exercise are as follows. First, the shortest interval of written notice appears to be too short to yield better job matches: the indirect effects are uniformly negative and the direct effects though positive are derived from insignificant coefficients in the wage equation. Second, informal notice may be of insufficient precision to materially impact earnings development in the wake of displacement. Third, and relatedly, the direct and indirect effects of notice strengthen with the duration of the notice interval, except in the case of layoffs. For this latter group of workers, then, any favorable impact of the longest interval of notice on job changing does not translate into overall earnings improvement. Indeed, there are few signs to suggest that laid off workers overtly benefit from receipt of any kind of notice.

IV. Discussion

The improved job matches facilitated by lengthy advance notice provide an obvious but hitherto unexploited link between two separate findings reported in analyses of the most recent Displaced Worker Surveys, namely, the failure of such notice to ameliorate jobless duration on the one hand and its success in raising postdisplacement earnings on the other. Even identical reported spell lengths of joblessness between notified and non-notified workers will conceal real differences in search time. Extended search is of course consistent with the location of better job matches.

Better job matches imply earning improvement (or reduced earnings losses) by reason of the initial job match and its (longer) duration. We found evidence to suggest that extended notice may indeed be associated with significant reductions in postdisplacement job changing and that reduced job turnover contributes positively to earnings development. Based on these results, we sought to distinguish the direct and indirect effects of notice on earnings. Clearly, the finding that notice favorably impacts earnings, noted in the recent literature, implies better job matches on the part of notified workers. But inclusion of the number of postdisplacement jobs in the wage equation not only renders this process more transparent but also provides a more correct specification of that equation.

One nagging feature of the advance notice literature has been the sensitivity of the results obtained to sample construction. Much of the problem doubtless stems from the lack of firm priors on which to base sample stratification. Here, we distinguished between the displacement occasioned by plant closing and that resulting slack work or abolition of a shift or position. It was felt that results from a largely employer-selected group could well differ from a more representative displaced worker sample (i.e., plant closing affect the whole work force rather than a subset thereof). Markedly different results were duly reported for the two groups.

If it is not surprising that the longest interval of written notice can favorably impact earnings, it is nevertheless something of a puzzle that its effects vary so widely as between the two subsamples. Observed compositional difference between the two groups cannot explain the disparities in outcomes, although the unfavorable information conveyed to the other side of the market could clearly operate in the hypothesized manner to negate any informational edge enjoyed by the notified laid off worker. There are two remaining possibilities. The first, noted earlier, is that significant numbers of workers on layoff expect to be recalled, which (unrealistic) expectation duly affects their search behavior. This possibility is not ruled out by the DWS requirement that workers be permanently displaced. It is worker respondent who identifies that displacement event and this can clearly be defined ex post. The second explanation would emphasize quality differences in notice rather than differences in the receptivity of workers to the same notice. Although the DWS is silent on the issue of the other forms of assistance that accompany written notice, a recent survey of establishments by the United States General Accounting Office (GAO) [13] confirms that notice is often just one component of a package of reemployment assistance. Moreover, and more relevantly, the GAO study demonstrates that establishments that closed were more likely to offer their workers assistance than establishments that had a mass layoff, although as shown in Table IV only in the case of income maintenance benefits are the differences between the two groups statistically significant at the .05 level. It is unfortunate that the GAO does not further refine the cross tabulations by length of notice, but since supplementary assistance of this type is functionally related to length of notice (see the last two columns of the table), it remains entirely possible that significant qualitative differences could attach to the same quantity of notice as between plant closings and layoffs in favor of the former. In short, part of that which has been attributed to notice, and in particular the longest interval of notice, may reflect the impact of other forms of reemployment assistance jointly provided with explicit notice.

This attempt to peek inside the black box of mechanisms through which advance notice "works" is clearly very much a first step. Thus, we have not examined the search process directly and, in particular, we have not modeled the intensity of on-the-job versus unemployment search and the effects of notice on reservation wages. Moreover, the identity of the "good job match" is necessarily imprecise in this treatment for reason that reflect the absence of wage data for intervals other than point of displacement and survey data and our disinclination to posit that the key match resulting from notice is inevitably that reported immediately following the displacement event. In our treatment multiple job holding in the wake of displacement is viewed as an unproductive activity that can be ameliorated by a sufficient quantity of notice, other things being equal. The provision of lengthy notice is not a panacea, however, and beneficial effects may hinge in part on other forms of reemployment assitance (the quality of notice) and differences in the quality, actual or perceived, of the group receiving notice.(8)

The present treatment offers only a partial reconcialiation of the unemployment and earnings outcomes separately reported in empirical analyses of the most recent Displaced Workers Surveys. Worker Surveys. For the future, these outcomes must be accommodated within a common search-theoretic framework that takes the arrival rate of jobs offers, the offered wage distribution, and search costs explicitly into account.(9) It should not go unsaid, however, that the Displaced Worker Survey remains a very imperfect vehicle for ascertaining the specific contribution of advance notice in this regard.

[TABULAR DATA OMITTED]
Appendix Table III. Probit Probability of Reemployment

 Sample
Variable All Workers Plant Closings Layoffs
constant -.045 -.691 .569
 (.412) (.678) (.530)
INFONOT -.037 -.068 -.007
 (.075) (.120) (.099)
WRITNOT1 -.081 -.171 .009
 (.143) (.265) (.175)
WRITNOT2 .188 .154 .368
 (.206) (.252) (.384)
WRITNOT3 -.154 -.236 .019
 (.165) (.198) (.332)
EDUC .051*** .051** .052**
 (.016) (.024) (.023)
AGE -.019*** -.024*** -.014***
 (.004) (.006) (.005)
TENURE .002 .002 .006
 (.006) (.009) (.010)
lnOLDWAGE .110 .327*** -.014
 (.073) (.120) (.096)
UNEM -.091*** -.076*** .106***
 (.019) (.029) (.026)
WHITE .496*** .617*** .430***
 (.094) (.141) (.128)
MALE -.091 -.181 -.027
 (.082) (.126) (.111)
MARRIED .118 .173 .067
 (.081) (.127) (.107)
KIDS -.035 -.064 .004
 (.033) (.510) (.045)
BLUE -.164* .011 -.302***
 (.086) (.137) (.114)
MANU -.038 -.246* .054
 (.080) (.134) (.104)
HEALTHINS .089 .088 .083
 (.081) (.129) (.105)
CLOSE .428***
 (.078)
ABOLISH .280**
 (.113)
YEAR83 2.113*** 1.724*** 2.615***
 (.171) (.232) (.297)
YEAR84 2.137*** 1.968*** 2.302***
 (.183) (.247) (.286)
YEAR85 1.849*** 1.789*** 1.945***
 (.134) (.208) (.182)
YEAR86 1.121*** 1.011*** 1.199***
 (.086) (.137) (.114)
SMSA1 .012 .018 .006
 (.108) (.174) (.142)
SMSA2 -.093 -.199 -.101
 (.087) (.135) (.116)
SMSA3 -.216* -.170 -.222
 (.114) (.178) (.151)
NORTHE -.401*** -.279 -.481***
 (.117) (.178) (.159)
NORTHC .090 -.163 -.016
 (.092) (.143) (.124)
WEST .085 -.019 .143
 (.100) (.164) (.129)
Log-likelihood -866.72 -345.35 -510.09
n 3385 1746 1639

Notes: asymptotic standard errors in parentheses
*, * *, *** denote significance at the .10, .05, and .01 levels, respectively

Appendix Table IV. Duration Equation, Accelerated Failure Time Model, Extended G
eneralized Gamma
Specification

Variable Coefficient Estimate Variable Coefficient Estimate

Constant 1.882*** HEALTHINS .029
 (.354) (.067)
INFONOT -.080 CLOSE -.436***
 (.061) (.065)
WRITNOT1 .035 ABOLISH -.272***
 (.124) (.093)
WRITNOT2 -.137 YEAR83 -.125
 (.146) (.105)
WRITNOT3 -.122 YEAR84 -.155*
 (.139) (.092)
EDUC -.029** YEAR85 -.038
 (.014) (.084)
AGE .023*** YEAR86 -.019
 (.003) (.082)
TENURE .011** SMSA1 -.120
 (.006) (.086)
InOLDWAGE -.091 SMSA2 -.017
 (.062) (.071)
UNEM .137*** SMSA3 -.022
 (.015) (.094)
WHITE -.486*** NORTHE .301***
 (.091) (.092)
MALE -.239*** NORTHC .178**
 (.068) (.076)
MARRIED -.079 WEST -.057
 (.067) (.078)
KIDS .044 [sigma], scale parameter 1.563***
 (.028) (.024)
BLUE .217*** q shape parameter .183*
 (.069) (.066)
MANU .159*** Log-likelihood -5829.74
 (.065) n 3385

Notes: asymptotic standard errors in parentheses
*, **, *** denote significance at the .10, .05, and .01 levels, respectively


References

[1.] Addison, John T. and McKinley L. Blackburn. "Search Duration and Joblessness." Unpublished Paper, Department of Economics, University of South Carolina, February 1992. [2.] _____ and Pedro Portugal, "On the Distributional Shape of Unemployment Duration." Review of Economics and Statistics, August 1987, 520-26. [3.] _____ and _____ , "Job Displacement, Relative Wage Changes, and Duration of Unemployment." Journal of Labor Economics, July 1989, 281-302. [4.] _____ and _____, "Advance Notice," in Job Displacement: Consequences and Implications for Policy, edited by John T. Addison. Detroit: Wayne State University Press, 1991. [5.] _____ and _____, "Advance Notice and Unemployment: New Evidence from the 1988 Displaced Worker Survey." Industrial and Labor Relations Review, July 1992, 634-64. [6.] _____ and _____, "Advance Notice: From Voluntary Exchange to Mandated Benefits." Industrial Relations, Winter 1992, 159-78. [7.] _____, Douglas A. Fox, and Christopher J. Ruhm, "The Impact of Advance Notice on the Probability of Unemployment: A Comment." Industrial and Labor Relations Review, July 1992, 665-73. [8.] Ehrenberg, Ronald G. and George H. Jakubson. Advance Notice Provision in Plant Closing Legislation. Kalamazoo: W.E. Upjohn Institute, 1988. [9.] Kletzer, Lori. "Sectoral Mobility and Jobless Durations: Evidence from the Displaced Workers Surveys." Unpublished Paper, Department of Economics, Williams College, September 1990. [10.] Nord, Stephen and Yuan Ting, "The Impact of Advance Notice on Plant Closings on Earnings and the Probability of Unemployment," Industrial and Labor Relations Review, July 1991, 681-91. [11.] Ruhm, Christopher J. "The Impact of Formal and Informal Notice on Postdisplacement Earnings." Unpublished Paper, Department of Economics, University of Nordi Carolina at Greensboro, 1991. [12.] _____"Advance Notice and Phostdisplacement Joblessness." Journal of Labor Economics, January 1992, 1-32. [13.] United States General Accounting Office. Plant Closings - Limited Advance Notice and Assistance Provided Displaced Workers. GAO/HRD-87-105. Washington, D.C.: GAO, July 1987.

(1.) For a review of first generation studies using the 1984 and 1986 surveys, see Addison and Portugal [4; 5]. (2.) Even if, as was indeed found to be the case, these workers have no less favorable observed characteristics than the plant closing sample, it is likely that the form of their job termination conveys very different information to the market; that is, laid off workers may well be viewed as "lemons" by prospective employers. (3.) Let [y.sup.*] equal the latent variable "tendency to change" jobs. The influence of the latent variable on the expected number of jobs (j = 1, 2, 3) may be expressed [y.sup.*] < 0 if j = 1 [mu] > [y.sup.*] [greater than or equal to] 0 if j = 2 [y.sup.*] [greater than or equal to] [mu] if j = 3, where [mu] is a threshold value (and is the estimated [mu] in the ordered probit) and [y.sup.*] = [beta]'X + [epsilon], [epsilon] N(0, 1). Thus we may write e < - [beta]'X for j = 1 [mu] - [beta]'X > e [greater than or equal to] - [beta]'X for j = 2 e [greater than or equal to] [mu] - [beta]'X for j = 3. Since [y.sup.*] and e are normally distributed Prob(j = 1) = F(- [beta]'X) Prob(j = 2) = F(- [beta]'X + [mu]) - F(-[beta]'X) Prob(j = 3) = 1 - F(- [beta]'X + [mu]) where F(*) is the cumulative standard normal distribution. (4.) We used the actual number of jobs rather than the instrumented value because of our concern with the reliability of the ordered probit as a predictor of the number of postdisplacement jobs. Justification for treating the actual number of jobs as an exogenous variable in the postdisplacement wage equation is provided below. (5.) The partial derivative of the expected number of jobs with respect to the notice variable is calculated using [Mathematical Expression Omitted] where [beta.sub.1] is the coefficient of the notice argument in the ordered probit and [phi] is the standard normal density function. (6.) Replacing the residual by the natural logarithm of the predisplacement wage - its coefficient being either freely estimated or constrained to unity - did not disturb the basic results reported below. Moreover, much the same pattern of results obtained for the coefficient estimates of principal interest when we excluded the previous wage variable. (7.) Note that the duration equation, the reemployment probit, and the ordered probit each include controls for the previous occupation and industry. The postdisplacement wage equation instead uses current occupation and industry as regressors. In addition, the former three equations substitute the natural logarithm of the predisplacement wage for the wage residual term used in the postdisplacement wage equation. (8.) Since we would not seek to argue that notice actually impedes the reemployment process, "perverse" effects of notice, where observed, may be associated with unobserved characteristics that would yield even less favorable outcomes in the absence of notice. As a practical matter, attempts to model the endogeneity of notice have proved elusive because of the absence of establishment and other variables in the Displaced Worker Survey [4; 6; 12]. [9.] For an interesting discussion of the determinants of sectoral mobility along these lines, see Kletzer [9].
COPYRIGHT 1993 Southern Economic Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1993, Gale Group. All rights reserved. Gale Group is a Thomson Corporation Company.

Article Details
Printer friendly Cite/link Email Feedback
Author:Fox, Douglas A.
Publication:Southern Economic Journal
Date:Jul 1, 1993
Words:5484
Previous Article:An economic analysis of recidivism among drug offenders.
Next Article:Do legal minimum wages create rents? A re-examination of the evidence.
Topics:


Related Articles
The impact of managerial layoffs.
Clintongue.
Lasting consequences of wrongful dismissal.
Current developments in employee benefits.
Technology, Trade Sensitivity, and Labor Displacement.
CITY ATTORNEY HOPEFULS PRESENT ARGUMENTS : HAHN, STEIN CLASH AT CSUN DEBATE.
WELLS LAYS OFF 1,750 WORKERS.
KATZ CALLS FOR HAHN TO DEBATE SPLIT.
Giving staff advance notice of large layoffs. (Legal).

Terms of use | Copyright © 2016 Farlex, Inc. | Feedback | For webmasters