The European Social Survey asks individuals how much control they have over how their daily work is organized (on a scale of 0 to 10). As an alternative to looking at health, one could also consider subjective well-being: in the data happiness is also ranked on a scale of 0 to 10.
The graph plots the mean of the happiness variable against how much control individuals say they have, using the data for Ireland. This is purely a bivariate comparison.
So there is a correlation with those at the top recording greater happiness but the gradient seems fairly gentle to me. More importantly, those at the bottom of the ladder seem happier than those immediately above them. So it may be that those at the bottom just don't worry and get on with things. A multivariate analysis would be needed to consider this more throughly.
If one looks at subjective general health with the same data, the relationship is rather more monotonic with those at the top having better health.
Interesting. One thing you should keep in mind with happiness scales is that there is very little dispersion. I can't see it really with that graph but a mean difference of half a point in a decent sample size would be a lot for happiness data. Things like divorce, unemployment, chronic illness etc., tend to "cost" about 2 points or so.
ReplyDeleteI wonder is there a case for starting the y-axis at 5 or thereabouts to get a sense of the differences.
One of the main ways the literature has tried to provide a meaningful scale is to compare coefficients on these variables with income coefficients, to get some kind of compensating variation. My view now is that this tends to lead to overestimates particularly if there is measurement error in income that attenuates the income coefficient.
Again, I know this is a blogpost not a seminar but another thing to think about straight off is the sample size at each point of the control variable. Is there enough in the first four categories?
ReplyDeleteYup, all good points. Leaving aside the usual causality concerns, a multivariate analysis would tell us something. For example, women are found more in the lower categories of work-control. The income data in ESS isn't great but one could so something.
ReplyDeleteThere is actually a reasonable spread over the categories of the work control variable: the lowest is 17%; the lowest 4: 35%.
So this could make a nice little project for a good student. If anyone is interested, I am happy to help.
I've been looking at this in relation to SHARE, will post at some stage
ReplyDelete