Impact of first birth career interruption on earnings: evidence from administrative data

ABSTRACT

This article makes use of unique administrative data to expand the understanding of the role women’s intermittency decisions play in the determination of her wages. We demonstrate that treating intermittency as exogenous significantly overstates its impact. The intermittency penalty also increases in the education level of the woman. The penalty for women with a high school degree with an average amount of intermittency during 6 years after giving birth to her first child is roughly half the penalty for college graduates. We also demonstrate the value of making use of an index to capture multiple dimension of the intermittency experience, and illustrate the importance of firm dynamics in the determination of a woman’s wage.

KEYWORDS:

• Intermittency
• administrative data
• career interruptions
• fertility
• labour supply
• wage differentials

JEL CLASSIFICATION:

• J13
• J31
• J22

Acknowledgements

The opinions expressed in this article do not necessarily reflect those of the Federal Reserve Bank of Atlanta nor the Federal Reserve System. The article has benefited from comments from Christopher R. Bollinger and from participants of the University of Colorado Department of Economics seminar series. Thanks to Keyung Wang and Chunying Xie for excellent assistance.

Disclosure statement

The authors declare they have no conflict of interest.

Notes

¹ While not focused on intermittency per se, Anderson, Binder, and Krause (2002) find a higher ‘motherhood’ penalty among higher-educated women, which is fully explained by time spent out of the labour force. This article will be able to quantify that relationship more explicitly and illustrate the differential importance of selection across education levels.

² White et al. (1990) provide an extensive discussion about the use of these employment data, commonly referred to as the Quarterly Census of Employment and Wages (QCEW), or ES-202 data.

³ Included in earnings are pay for vacation and other paid leave, bonuses, stock options, tips, the cash value of meals and lodging, and in some states, contributions to deferred compensation plans (such as 401(k) plans). Covered employer contributions for old-age, survivors, and disability insurance (OASDI), health insurance, Unemployment Insurance (UI), workers’ compensation, and private pension and welfare funds are not reported as wages. Employee contributions for the same purposes, however, as well as money withheld for income taxes, union dues, and so forth, are reported even though they are deducted from the worker’s gross pay.

⁴ We have no way of knowing how much of the non-matched 30% is the result of imprecise matching or the result of the mother simply not being in the Georgia workforce.

⁵ Heeding warnings of Bollinger and Chandra (2005), we do not eliminate outliers, although doing so does not affect the results.

⁶ While kindergarten is not required in Georgia, all public school systems begin at least in kindergarten, with public pre-k provided for only a select few lottery winners.

⁷ Since Kahn, García-Manglano, and Bianchi (2014) find that selection into the labour market is most strongly influenced by children when women are young, we re-estimate the model on a subsample of older women (age ≥ 34) and find similar results to those reported here (results available upon request). The wage penalty is larger, but varies similarly across education levels. Also see Han et al. (2008) who find that employment pre-birth is a strong indicator of post-birth return to the labour market.

⁸ Ideally, we would use hourly wages as the dependent variable, but hours of work are not available in the data. Implications of this limitation were discussed in the previous section.

⁹ Of course, in 6 years it is possible that the woman has given birth to a second, or third, child. We repeat the estimation on a reduced sample controlling for the number of siblings and obtain operationally equivalent results. Results of this robustness analysis are detailed below.

¹⁰ See Hotchkiss, Quispe-Agnoli, and Fernando (2015) for the consistency with earlier wage determination literature.

¹¹ These results are consistent with Hotchkiss, Pitts, and Robertson (2004) who find that new employment with an expanding firm results in greater earnings gains than new employment with a contracting or dying firm.

¹² Since the data do not contain information about hours worked earnings penalties estimated here combine any incidence of lower wages with lower hours. This will only be a concern if women likely to exhibit greater intermittency are also more likely to work fewer hours. Consequently, earnings estimates reported in this article should be considered upper bound estimates of a wage penalty.

¹³ Parameter estimates for the model excluding tenure are −0.4715, −0.0371, −0.4418, and 0.3530 for the intermittency index, number of periods of absence, total per cent of time absent, and per cent of time since last spell, respectively. All estimates are statistically significant at the 99% confidence level.

¹⁴ Results from Miller (2011) suggest that the timing of fertility decisions within one’s career impacts the penalty and that the benefit of delaying fertility also varies by education level. Other than controlling for age, we do not account for the specific timing of fertility decisions by education status. Note that based on a chi-squared test, the exogenous versus fixed-effects estimates of the penalty are statically significantly different from one another for each education level at the 99% confidence level.

¹⁵ These estimation results are available upon request.