Trade-offs are more complicated than Okun’s leaky bucket

A popular story in macroeconomics and broad public economics is that of Okun’s leaky bucket (Okun 1975).  This essentially states that we can think of redistributing income like using a leaky bucket to move water – where the loss of income associated with the loss of efficiency is the water that falls out of the bucket on the trip.  This gives us our basic “equity-efficiency trade-off“.

Both the Spirit Level and a lot of the “new economics of income inequality” (namely the work on rent seeking) is trying to turn this on it’s head by saying that, not only does the efficiency vs equity/inequality trade-off not exist, but it is often the other way around.  When a mechanism is given, which can in turn be tested, I am fine with this (eg I am not discussing the types of studies Offsetting touches on here).

However, when it comes to directly determining policy about redistribution this use of the leaky bucket has stretched the metaphor too far.  Policy has more dimensions than a couple of synthetic aggregreates (GDP and inequality measures) which get confused for the much broader terms ‘efficiency’ and ‘equity’.  The “leaky bucket” was a metaphor to describe a likely trade-off between efficiency and the equality of income via the use of policy to redistribute – but it was solely a starting point for thinking about the issue.  For some reason many authors have taken that and created a welfare function that embodies a “trade-off between equity and efficiency” focusing solely on mean income and some measure of its statistical dispersion.  The feeling is that we can use that to create a “abbreviated social welfare function” which helps with policy.

But when it comes to actual policy, policy settings impact upon welfare in a myriad of different ways not captured by this.  The more heterogeneous populations are, the worse this problem becomes (eg Lockwood and Weinzierl 2013).  Ignoring heterogeneity of choice is dangerous, and is the very issue I am most concerned about with the discussions of inequality that sometimes take place.

Note:  For people who are inclined towards thinking about causal social science I would suggesting thinking of it this way.  Our analysis based on a “leaky bucket trade-off” tells us about average effects over a population, but heterogeneity both tells us that the individual effects will differ greatly (which can mean very different welfare effects given the relationship between our aggregates and welfare is likely to be non-linear) and that the preferences of individuals will differ (over consumption, leisure, and non-pecuniary job attributes) – implying that some inequality in terms of the measures we quote is good, and we need to understand the nature of this.  Heterogeneity makes the “functional relationship” between observed variables and our target variable (welfare) more complex, and importantly it also implies that even the direction of any ideal counterfactual is up for debate on an a priori basis.

Let us take New Zealand as an example.  We recently touched on some of the literature in New Zealand with regards to inequality.  One particular paper was Podder and Chatterjee (2001), which discussed how to look at generalised Lorenz dominance in the New Zealand income distribution – in other words, they were able to discuss how, for any conceivable view of the equity-efficiency trade-off, New Zealand at the start of the reforms was preferable to 1991/92 New Zealand.

This paper was well put together, methodologically strong in terms of its discussion of these concepts, and it was transparent about the fact that they viewed much of the drop in income over this period as a “policy choice”.  Furthermore, generalised Lorenz dominance is a strong result – especially if you do view the slowdown as a policy choice (I would note that I’m not as certain I would view it all as policy choice, and I would also point out that much of the cost could be seen as transitory – which is relevant when 1991/92 was the cyclical bottom of the cycle).  However, the reforms changed a number of other characteristics of New Zealand which we cannot view as “welfare irrelevant” – which is why we need to be significantly more careful.

Evans, Grimes, and Wilkinson (1996) discussed the nature of the reforms with an efficiency bent [Note:  Brian Easton was unhappy with their choice of argument].  However, of the main other areas where we saw significant policy change was in how “open” New Zealand was for trade.  The impact of this was a significant lift in the choices available to New Zealander’s (just think of the change in the availability and price of clothing and footwear, as Will Taylor says ‘variety is the spice of life‘) – a factor that isn’t captured well in GDP statistics.  The sheer, massive, change in the structure of the New Zealand economy makes welfare comparisons based on Lorenz Curves or abbreviated social welfare functions fraught – which is why analysis is (and should be) focused on details of the reform program, and the corresponding distributional effects of what occurred.

The reforms were not perfect, the reforms were not all bad.  And an oversimplified trade-off in the form of “Okun’s leaky bucket” can only take us so far in trying to understand them!

Note:  Looking up some of the papers I was after for this post lead me to this from Motu – a paper I previously hadn’t seen.  The scale of the employment and income shocks due to the reforms have had long-lasting effects.  I think this is a valuable perspective.

If structural shocks have long term consequences, we better make sure we have a clear idea of what we are doing before we do it 😉