How relative income affects work hours preferences

ABSTRACT

In this article, we look at explanations for why people want to change their working hours. We focus on the role of income and differentiate between the effect of household income, personal income and self-perceived relative income. Using Flemish data on 1435 workers, we perform binary logistic regressions in which we compare those who are over- or underemployed with those who are currently working their preferred number of hours. Our results show that the desire to work fewer hours is mostly related to a bad work–life balance, while the wish to increase working hours is associated with relative income rather than absolute income. Based on our findings we recommend governments to not only focus on increasing flexibility at the individual level but to also consider these positional effects by taking measures (e.g. decreasing the duration of the standard working week) at the population level.

KEYWORDS:

• Relative deprivation
• relative income
• work hours mismatch
• work hours preferences

JEL CLASSIFICATION:

• D31
• J22

Disclosure statement

The authors declare that they have no conflict of interest.

Notes

¹ The mismatch between actual working hours and preferred working hours has been studied extensively (e.g. Van Echtelt, Glebbeek, and Lindenberg 2006; Wooden, Warren, and Drago 2009) and has been found to negatively impact life satisfaction (Başlevent and Kirmanoğlu 2014; Wooden, Warren, and Drago 2009), job satisfaction (Angrave and Charlwood 2015; Iseke 2014), psychological well-being (Angrave and Charlwood 2015; Friedland and Price 2003), physical health (Bell, Otterbach, and Sousa-Poza 2012; Friedland and Price 2003) and productivity at work (Wang and Reid 2015).

² Research shows that work hours mismatches are constantly created and resolved, but while part-time workers who want to work more hours often find another job, the desire for fewer hours is often long-lasting (Breunig, Gong, and Leslie 2015; Reynolds and Aletraris 2006, 2010).

³ This does not mean that there are no economic theories that could explain the existence of work hour mismatches. The heterogeneous agent search and matching literature, for instance, provides some tools that could account for these mismatches (see Yashiv 2007).

⁴ Wielers, Münderlein, and Koster (2014) point to differences in quality and availability of part-time jobs as an explanation for the higher popularity of these jobs in some countries. An alternative explanation is that there is an impact of relative income on the hours people prefer to work. If more people earn less, people themselves also have a smaller desire to obtain a higher income. Yet another explanation could be found in Alesina, Glaeser, and Sacerdote (2005) who use a ‘social multiplier’ as an explanation for a similar effect. They argue that, as more people work fewer hours, the increased possibility to spend leisure time with others will incentivize people to work fewer hours.

⁵ LEVO is the Dutch acronym for ‘LEvensomstandigheden in Vlaanderen Onderzocht’ (research on living circumstances in Flanders).

⁶ These categories can also be recoded to reflect the average income of each category. This would allow us to experiment with different values for those in the highest possible income categories. We have explored this option and have tried alternative values for the highest income categories (e.g. 4000 or 5000 for those with the highest personal incomes, instead of 3250), but this did not change our results. Therefore, we have chosen to simply work with the nominal values of the income variables.

⁷ It could be argued that absolute household income does not adequately reflect purchasing power because it ignores household composition. To address this concern, we have also used the modified OECD equivalence scale to calculate household incomes corrected for family size (as in Hagenaars, de Vos, and Zaidi 1994). These corrected household incomes are obtained by dividing standard household income (which can be calculated by taking the mean value for each category as an estimate for their household income) by 1 plus 0.5 for each additional adult and 0.3 per child in the household. For example, a household with two adults and two children would have its income divided by 2.1. Using this measure instead of our measure for household income did not change the results of our analyses.

⁸ The difference between two categories can be calculated as follows. First, we take the difference of the logs of both categories. For the fourth and third category, for example, this would be ln(4)-ln(3) = 0.2877. Next, we have to exponentiate the odds ratio with this number. We then find that 0.408^0.2877 = 0.773. This number indicates that those in the fourth category are only 0.773 times as likely as those in the third category to want to work more hours. Hence, they are 22.7% less likely to want to work more hours.

⁹ This could possibly lead to a biased estimate of the coefficient for income in our analyses. The standard approach to these types of concerns is to use an instrumental variables regression. Unfortunately, our dataset does not contain adequate instrumental variables, making it impossible to apply this method.