The Impact of Youth Unemployment on Adult Unemployment in the NCDS

(2001) - Paul Gregg

Journal: The Economic Journal
Link:: https://academic.oup.com/ej/article/111/475/F626-F653/5139987
DOI:: 10.1111/1468-0297.00666
Links::
Tags:: #paper #NCDS #Unemployment #Transition #school-to-work #SocialClass
Cite Key:: [@Gregg2001]

Abstract

Using the National Child Development Survey, this paper looks at cumulated experience of unemployment, highlighting how unemployment experience is concentrated on a minority of the workforce over extended periods. Low educational attainment, ability not captured by education, Ænancial deprivation and behavioural problems in childhood raise a person's susceptibility to unemployment, there is strong evidence of structural dependence induced by early unemployment experience for men but only minor persistence for women. Attacking low educational achievement, and preventing the build-up of substantial periods in unemployment as youths, may reduce the extent to which a minority of men spend a large part of their working lives unemployed.

Notes

“highlighting how unemployment experience is concentrated on a minority of the workforce over extended periods. Low educational attainment, ability not captured by education, Ænancial deprivation and behavioural problems in childhood raise a person's susceptibility to unemployment, there is strong evidence of structural dependence induced by early unemployment experience for men but only minor persistence for women.” (Gregg, 2001, p. 626)

“More speciÆcally, whether unemployment damages future employment chances or earnings. In Britain, such a damaging effect on people is commonly described as scarring” (Gregg, 2001, p. 626)

“Long-term unemployment could simply be serving as a way identifying those with otherwise unobserved characteristics (heterogeneity) which leads to them having low exit probabilities.” (Gregg, 2001, p. 627)

“The paper draws on the National Child Development Survey to look at whether the cumulated experience of unemployment before age 23 drives individuals unemployment patterns between 5 and 10 years later, up to the age of 33” (Gregg, 2001, p. 627)

“The long panel with complete histories” (Gregg, 2001, p. 627)

“allows the use of an instrumental variables approach to identify structural dependence from any residual unobserved heterogeneity as proposed by Heckman and Borjas (1980).” (Gregg, 2001, p. 628)

“The results provide strong evidence of structural dependence induced by early unemployment experience for men. The paper also suggests that low education attainment, a disadvantaged family background and individual ability (not captured by educational attainment) and behavioural characteristics raise the underlying susceptibility to unemployment” (Gregg, 2001, p. 628)

“If a record suggests two weeks of a month was spent in employment and two weeks unemployed, the monthly record will show the person as being employed. Hence, short spells of unemployment will be under-recorded in the data.” (Gregg, 2001, p. 628) Important to remember when modelling unemployment

“The NCDS has a wealth of pre-labour market information normally unobserved in labour market databases. For clarity, we group this into parental/ family characteristics, and individual ability/behaviour measures. The Ærst covers the educational attainment of parents, family income at 16 and reported family Ænancial distress in childhood, ethnicity, etc. It also includes children taken into local authority care. The children participating in the NCDS had a number of tests administered to them at ages 7, 11 and 16. The measures of ability utilised in this paper are test scores in maths and vocabulary, administered at age 7 and a non-vocabulary reasoning test (akin to an IQ test) at age 11.” (Gregg, 2001, p. 630)

“Behavioural information covers the Bristol Social Adjustment category scores at age 7, and school attendance and getting into trouble with the police or probation service, by age 16.” (Gregg, 2001, p. 630)

“There are three possible reasons why unemployment falls disproportionately on the same people both as youths and prime age adults: ∑ Some people are always more prone to unemployment because of low education or other less easily observed heterogeneity. ∑ Area unemployment differences persist in the medium to long term.3 ∑ Unemployment when young damages your future employment prospects (ie structural dependence or scarring effects).” (Gregg, 2001, p. 631)

“Disentangling scarring (structural dependence) from unobserved heterogeneity is not straightforward. To date four approaches have been used in the context of unemployment” (Gregg, 2001, p. 632)

“Here it is assumed that the unobserved heterogeneity is constant across a number of cohorts and the relationship between this heterogeneity and unemployment is also constant. However, the cohorts do differ in their unemployment experience because of the economic cycle” (Gregg, 2001, p. 632)

“Duration dependence ± The intuitive evidence for duration dependence is the way observed outØows from unemployment decline rapidly with elapsed duration. After the1970s, average outØow rates declined (at all, durations) and Layard et al. (1991) argue, plausibly, that this could not be attributed to changes in population characteristics.” (Gregg, 2001, p. 633)

“Lagged unemployment dependence ± Arulampalam et al. (2000) and Muhleisen and Zimmerman (1994) both apply a random effects probit model to British and German household panel data, respectively. The data are presented as annually observed dummy variables over a short time window. A two-step procedure is followed where, in the Ærst step, IV are used to estimate a reduced form model for the initial observation (to overcome the initial conditions problem). In the second step, a random effects probit model is used with the random effects applied to control for unobserved heterogeneity” (Gregg, 2001, p. 633)

“All studies in this genre Ænd short-run structural dependence.” (Gregg, 2001, p. 633)

“This method is commonly used in looking at the wage reductions resulting from job loss” (Gregg, 2001, p. 633)

“Heckman and Borjas (1980) suggest that the conditions required to identify lagged duration dependence are less stringent and do not require the strong distributional assumptions for duration dependence in individual spells. They suggest an instrumental variables approach using time varying exogenous information.” (Gregg, 2001, p. 634)

“Note that (iii) and (iv) have some strong similarities. In both cases, some portion of a person's unemployment experience is deemed to be independent of individual heterogeneity and then the subsequent effects of this exogenous portion are studied” (Gregg, 2001, p. 634)

“The instrument utilised is the local unemployment rate just as entry into the labour market can Ærst take place at age 16. This is ward level data derived from the 1971 Census (as discussed above).” (Gregg, 2001, p. 636)

“Residence at age 16 is clearly exogenous to the individual, as nearly everyone lives in the parental home, but it does have limitations as an instrument.” (Gregg, 2001, p. 636)

“Second, the instrument is highly disaggregated and thus might start to pick up socio-economic factors and not just local labour market conditions.” (Gregg, 2001, p. 637)

“The third potential problem is that there might be a correlation between unemployment after age 28 and the instrument which captures patterns of local labour demand when the person was age 16.” (Gregg, 2001, p. 637)

“This variation over time of the coefÆcients attached to background factors may also reduce, but cannot eliminate concerns over multicollinearity” (Gregg, 2001, p. 648)

“While being able to control for individual ability and behavioural test scores and family background so extensively is relatively unusual, there may still be determinants of the underlying propensity to experience unemployment that remain unobserved” (Gregg, 2001, p. 648)

“So an instrumental variables approach offers reassurance if a valid instrument can be found” (Gregg, 2001, p. 648)

“A valid instrument must be causally determining early unemployment experience, exogenous to the individual and must change over time to be separable from the individual Æxed effect.” (Gregg, 2001, p. 648)

“Men who experience unemployment as youths disproportionately go on to experience further unemployment when they are prime age adults.” (Gregg, 2001, p. 651)

“An IV approach is adopted to control for any remaining unobserved heterogeneity. Here, unemployment prior to age 23 is instrumented with local labour market unemployment conditions at age 16, just prior to or just on the cusp of entry into the labour market. It is argued that local labour market conditions when aged 16 will not directly inØuence later unemployment experience, except through a scarring effect from youth unemployment. These results suggest there is no upward bias to the conditional correlation between individual unemployment experiences before age 23 and after age 28 through unobserved heterogeneity. It must be noted though that the instrument is not without its problems.” (Gregg, 2001, p. 651)