comments from Bruno Larue

Post Reply
Larue_Bruno
Posts: 8
Joined: Tue Jun 05, 2018 8:28 pm
Contact:

comments from Bruno Larue

Post by Larue_Bruno » Fri May 31, 2019 10:02 am

The presentation was very good. You addressed some of the points I bring up below from reading your paper. You showed a very good understanding of your methodology and the discussion about the results was better in the presentation than in the paper.


Comments on « Work-study and Educational Mismatch among Youths: Evidence from Zambia” by Chitalu Miriam Chama-Chiliba, Hillary Chilala Hazele, Mwimba Chewe, Kelvin Chileshe and Araar Abdelkrim

The paper is about the incidence of working while in college or at university on mismatch between the skills of the young workers and the skills required by their jobs. A subjective mismatch measure is used along with an empirical mismatch measure. Workers may feel under-qualified, adequately matched or over-qualified. The dependant variable is categorical and this calls for the estimation of an ordered choice model. The treatment is the work experience of youth while in school. Two econometric issues may arise. A mismatch outcome can only be observed if the youth is actually employed. If the sorting of employed and unemployed is not done randomly, then a selection bias may arise. This can be addressed by adding a selection equation with exclusion variables to be able to sort out the effects of variables that enter both the selection and mismatch equations. The pertinence of adding a selection equation can be ascertained by testing the significance of the correlation between its residuals and the residuals of the ordered probit. The other econometric issue that might arise has to do with factors potentially impacting simultaneously the mismatch and treatment outcomes. The presence of an endogeneity bias can be ascertained by modeling the treatment or the probability of working while enrolled in school.

General comments:

The problematic is interesting. Matching skills of workers with the skills sought after by employers wishing to maximize productivity is not that obvious even though workers can signal (with some noise their abilities) and employers can screen to reduce information asymmetries. Work experience can help workers develop new skills and find out what they are good at and what they like. Therefore, one would expect the treatment to lower the probabilities of mismatching.
The motivation offered by the authors dwells on youth unemployment in Zambia being exacerbated by skills mismatching. Statistics presented show that youth unemployment has been decreasing over time, from a high of 37% in 1993 to 15% in recent years. While this is a huge improvement, a 15% unemployment rate is high enough for employers not to hesitate to fire unqualified youth and for overqualified (risk averse) youth to hold onto their job. Hence, it is not surprising to see that 26% of employed youth feel overqualified against 19% who feel underqualified. The implications for productivity and national income growth are quite clear: underqualified workers can be replaced by more qualified ones and overqualified workers must not allow frustration to pull their productivity down if they want to remain employed. This suggests that the economy suffers from too weak (strong) of a demand (a supply) for (of) skilled labor. It is not clear that programs adding qualifications to youth can help. Is the problem a vertical mismatch one with too many workers not having enough skills? Perhaps there is a horizontal mistmatching problem with youth having skills that they cannot exploit while lacking in skills that the labor market could reward. The concepts of vertical and horizontal mismatching are mentioned, but there is little in the way of empirical evidence to precisely sort out the nature of the mismatching problem in Zambia and how it affects unemployment. Perhaps, if there was information about the skills of the unemployed, you could make the case for a link between skills offered and skills demanded not matching at all and contributing to unemployment. Otherwise, I think that the motivation should focus on foregone national output from mismatching.

Youth unemployment does not vary much by gender, but it varies a lot across provinces (8.3% to 29.2%). Why does it vary so much regionally? Looking at Table 1’s mismatching by province, mismatching differences across provinces does not seem to explain youth unemployment differences across provinces.

Theoretical perspectives could be strengthened, perhaps better adapted to the Zambian context, for example explaining how the overall regional unemployment conditions mismatch outcomes.

The data must be described more thoroughly to better distinguish the selection bias concern from the endogeneity bias concern. Once the employed are sorted out, what is the proportion of treated, that have worked while enrolled in school? It must be made clearer that not all of the employed are treated and that some of the employed did not work while enrolled in school. Descriptive statistics about the treatment are not reported. It would be useful to know the proportion of youth who worked while in school.

Exclusion variables must be better motivated. Why is the mother’s education in the selection equation and not the father’s? What is the intuition? To what extent are poverty headcount and unemployment rate by ward correlated? They might be picking up the same thing. Poverty headcount is not significant in any of the models reported in tables A2 and A3. It could simply be removed or replaced by an alternative exclusion variable. Why are the father’s education and occupation exclusion variables in the treatment equation? Since youth is defined as someone between 15 and 29 years old, does it make sense to consider the youth’s number of children as an exclusion variable?

Coefficients are difficult to interpret in ordered regressions because they typically inform about directions for end categories and not the middle one(s). In this paper, the middle category (matched) is the good one and is of greater interest. W. Greene encourages reporting and discussing marginal effects about the probability for each category and not spend too much time on coefficients. Since full results (coefficients for all of the variables) are reported in tables A2 and A3, there is no need to add tables in the text for coefficients.

The marginal effects of the preferred model (the one correcting for selection and treatment endogeneity and using an artificial instrument as exclusion variable) in Table 3 are puzzling. People who work while in school acquire skills, learn to deal with job stress… and yet the probability of feeling less qualified increases by 11% for the treated youth! The marginal effects from the other models are more intuitive. The marginal effects from the empirical measure of mismatching do not have this problem, with the probability of overqualification increasing substantially with treatment. In this case, the interpretation is not about “feeling” overqualified, but it is about “being” overqualified. Youth that worked while in school invested in more years of schooling. Put in this light, the two definitions are less about robustness checks or alternatives and more about means to acquire different but complementary information.

One model has a cutoff point that is not statistically significant. It would also be useful to report the standard errors of the cutoff points to see whether there is overlap. Can the data deal with three categories or should there just be two?

The econometric interpretation of the results follows along the lines of the Stata documentation about extended regression models and may be this is why you seem so keen on having an endogenous treatment.

The policy implications are weak. You have to put yourself in the position of policymakers. What can they do with your results?

Minor comments:
p.6: “… it has been found that women are generally more risk averse than men … “ First, one or more reference is needed to support the assertion that “it has been found”. Second, if women are more risk averse, they would be less prone to try to quit their job when overqualified or underqualified. As such, one would expect a lower unemployment rate for women if they are more risk or loss averse.
p.7 … to determine the students that are best suited… replace to determine by to identify.
p.7 … Spence (1973) provides some insight … change insight for insights.
p.8 …aid their career path decisions… change aid for improve
p.10 Cameroon and Trivedi… change Cameroon for Cameron.
p.10 Where yi … You cannot start a sentence with Where unless you ask a question. Replace W with w in where.
p.12-13 There should be references about the construction of artificial instruments and a word or two about the limitations of such instruments.
p.14. In literature, there … replace with In the literature.
p.14 … human capital where as overeducated… replace where as with whereas.
p.15. I do not understand what you mean by “context specific case”. I would rephrase.
Is table 1 based on the subjective mismatching definition or on the empirical definition? How similar or different are the distributions of the two mismatching variables?
p.19. There is an incomplete sentence: Therefore, an individual would have acquired.
p.19. Table 4 should include also the probabilities of being underqualified and overqualified.
p.21. Table 2 depicts … replace with Table presents…
p.22 Some pseudo-R2s have not been computed for some estimators.
p.24 working while studying results are in bold in table 4 but not in table 5.
p.25. … empirical measue… replace with measure.
p.25. Are the F-statistics exceeding the threshold of 10 reported anywhere? Table 6 has the p-values. One can probably associate very low p-values with high F-stats but it would be best to report the F-stats if this is what you discuss.
References:
For books, the city where the publisher is located must be added.
The journal information is missing for the Boudarbat and Chernoff (2010) reference.
The style seems to vary in this section. Sometime you have pp, p or just the page numbers. It is important to be consistent.

Post Reply

Return to “PMMA-20280 - Occupational Preferences and Labour Market Skills Mismatch among Youths in Zambia”

Who is online

Users browsing this forum: No registered users and 1 guest