It turns out that maybe it really doesn't. At least, not according to research that I am completing with Matt Roskruge at NIDEA (the National Institute of Demographic and Economic Analysis, at the University of Waikato). In the journal Policy Quarterly in August (PDF), we wrote about some of our preliminary results. Our key research question was essentially: "Does working make older New Zealanders better off?".
We were interested in this question because labour force participation at older ages has been increasing substantially over time (and because I was funded by MBIE to investigate the effects of labour force participation among older people). To give you a sense of the scale of increase in participation, see Figure 1 below, which shows the labour force participation rate of each five-year age group 55 years and over, across the last five Censuses. This figure is taken from a working paper I wrote (PDF) earlier in the year, which explores a lot of different aspects of labour force participation among older workers in New Zealand.
Figure 1: Labour Force Participation Rate by Age, 1991-2013
Obviously, this is also a really important question, because the New Zealand population is ageing rapidly (see here or here for the national-level, or see some pretty graphic pictures (pun intended!) of the changing age structure at the sub-national level in this report by Bill Cochrane and I for the Waikato Region).
To look at the question of wellbeing and working among older people, Matt and I used data from three waves (2008, 2010, and 2012) of the New Zealand General Social Survey (GSS), which is a nationally-representative survey that collects data on a range of social and economic indicators of well-being. The key question is on life satisfaction (a proxy for overall wellbeing): "How do you feel about your life as a whole
right now?" Responses are measured on a five-point Likert scale (1 = very satisfied; 2 = satisfied; 3 = no feeling either way; 4 = dissatisfied; and 5 = very dissatisfied). To keep things simple (and avoid having to run ordinal models), we reduced this into a variable that was equal to one if the respondent was very satisfied, and zero otherwise.
Now, there are two key problems to overcome with trying to analyse the relationship between working an wellbeing. First, there is self-selection - people choose whether or not to work, and those who choose to work would likely be those who believe that it will increase their wellbeing. This would lead to a bias in any attempt to evaluate the effect of working on wellbeing. Second, there is an endogeneity problem, because health status affects whether an individual is able to work or not, and also directly affects the individual's wellbeing. We use instrumental variables regression to overcome both problems - using gender as an instrument for full-time work status. Now, I'm not the biggest fan of instrumental variables (see here for example). We think that gender is a good instrument because it is closely correlated with full-time work (men have higher labour force participation than women), and it meets the exclusion restriction because there is no theoretical reason why men should have higher (or lower) wellbeing than women (indeed, from having looked through the literature as part of the Enhancing Wellbeing in an Ageing Society (EWAS) project, there is little consistency in effect of gender on wellbeing). Incidentally, if you are interested in what correlates with wellbeing among older people in New Zealand, read this monograph (PDF) from the EWAS project (or this one (PDF) on people aged 40-64).
What Matt and I found is interesting. We had three groups of labour force status - full-time employed, part-time employed, and not working (which combines the small number of people who reported as unemployed, as well as those who reported as being retired). Figure 2 shows the raw results in terms of life satisfaction, and it looks like those not working have the lowest life satisfaction (smallest proportion very satisfied, and largest proportion not satisfied).
Figure 2: Life Satisfaction, by Labour Force Status
However, one we ran our instrumental variables model, we found that (after controlling for health status) full-time work is associated with significantly lower life satisfaction than either part-time work or not working. Working full-time is associated with an about 49% lower probability of reporting being very satisfied. Those are the results we report in Policy Quarterly. We've been doing some follow-up work since that article, looking at some of the mechanisms through which this might be working.
Is it because wealthier older people don't need to work, so those who are working are doing so because they have to in order to get by? It doesn't seem so - the results are robust to the inclusion of area deprivation (as a proxy for wealth - the GSS doesn't have a direct measure of wealth). Including wealth makes part-time work marginally statistically significant and negative (working part-time is associated with about a 7% lower probability of reporting being very satisfied).
Is it because older workers who are working full-time are dissatisfied with their jobs? If we restrict the sample to only those who are working, job satisfaction is significantly positively related to life satisfaction, but it doesn't make full-time work status any less statistically significant (or any less negative). There's no difference in the job satisfaction-life satisfaction relationship between full-time workers and part-time workers.
Is it because older workers who are working full-time want to work less but for some reason can't? The GSS asked people if they wanted to work more hours, or fewer hours. Wanting to work more hours is associated with lower life satisfaction. Wanting to work fewer hours is not. These results don't differ between full-time and part-time workers, and they also don't make make full-time work status any less statistically significant (or any less negative).
So, it appears to be a fairly robust result - working full-time when you're older makes you less happy. Or maybe less happy older people prefer to work (reverse causality is still a possibility). I know I'd rather be working when I'm older than doing this.
Matt and I are completing the write-up of a working paper on this at the moment. Keep an eye out for it here (it should be up by the end of October). We'll also be presenting on this at the Labour, Employment and Work conference in Wellington in November.