Do college students make better predictions of their future income than young adults in the labor force?

Abstract

Several studies have considered whether American college students' hold ‘realistic’ wage expectations. The consensus is that they do not – overestimation of future earnings is in the region of 40–50%. But is it just college students who overestimate the success they will have in the labor market, or is this something common to all young adults? In this paper, I analyze National Educational Longitudinal Study (1988) data to consider whether 20-year-old college men are more realistic about their future income than their peers (of the same age) who are already in the labor force. My findings suggest that young people in employment actually make worse predictions of their future income (on average) than certain student groups, so long as the latter successfully obtain a university degree.

Keywords:

• income expectations
• occupational expectations
• college students

JEL Classification:

• I20
• I21
• I28

Notes

1. Recent work in the sociological literature by Morgan (2005) depicts young adults as ‘Bayesian learners’. In particular, he illustrates the accuracy of a ‘fast’ and ‘slow’ learner's expectations over time. Morgan suggests the difference between fast and slow learners could be to do with the different timing of key life events. This could include entry into the labor market, a period when young adults should receive a lot of information that will lead them to quickly (and more accurately) updating their expectations.

2. These students were not randomly selected, but drawn from schools where there were other second and third wave respondents. More details can be found on page 56 of Curtin et al. (2002).

3. Around 5000 individuals were dropped from the study. Two thousand of these were classed as ‘poor responders’, who were basically excluded because of the low chance of future contact. Hence, it may be more appropriate to consider these 2000 observations as nonrespondents. The other 3000 individuals dropped were not classified as poor responders, but excluded purely to lower costs.

4. The substantive conclusions of this paper remain unchanged whether or not one applies these survey weights in the analysis.

5. Further analysis (which can be found in Jerrim 2010) shows that most of these observations were graduate students who had never entered full-time employment.

6. Jerrim (2010) further investigates this issue, and demonstrates that results are robust to methods that attempt to correct for any selection induced in the sample in terms of observable characteristics.

7. Ideally, students would have been formally instructed not to consider inflation in the wording of the question, as per Dominitz and Manski (1996). They report that students generally adhere to this, and do not consider inflation in their wage expectations. Moreover, Brunello, Lucifora, and Winter-Ebmer (2004) use similar wording to the question asked in NELS, in that students are not directly informed how to deal with inflation. They also assume students report their expectations in current prices, and find inconsistencies in their data with the idea that respondents make an adjustment to their responses to try and account for inflation.

8. Jerrim (2010) is a working paper version of this paper. It contains a set of additional results, along with further details of the NELS data that I analyze. This includes an explanation of how I predict individuals age 30 income from the information in the dataset that is available. Throughout this paper, I present results that refer to prediction ‘method 2’.

9. The exact wording of the question was: ‘What job do you expect or plan to have when you are 30 years old?’ Respondents were asked to write in an occupational description into an open text field.

10. It is, of course, possible for students to also hold a job while they study.  In auxiliary analysis not presented, I investigated whether the employment status of students was associated with the accuracy of their wage expectations. I found little evidence that this was the case.

11. As explained at the end of Section 2, ‘actual’ age 30 income is unobserved in NELS and so must be estimated from the data available. See Jerrim (2010) for full details.

12. Standard errors are based on 200 bootstrap replications. The NELS complex sample design (children clustered within schools) has been taken into account by clustering the bootstrap by school.

13. I have investigated the sensitivity of my findings to using quantile regression (focusing upon the median) rather than OLS. The substantive conclusions that I reach remain largely unchanged.

14. This refers to the school the respondent was in when selected in take part in the study at age 14. I have taken this clustering into account by making the appropriate adjustment to the estimated standard errors.