Friday 28 October 2016

Mismeasured: The Case of British Inequality

In preparing for an upcoming Uni essay on inequality, I was reading through contributions by various American economists talking about inequality especially John Cochrane's amazing rant (also available on his blog, The Grumpy Economist). The collection of essays itself, published last year by the Hoover Institution at Stanford University was originally a series of lectures in memory of Gary Becker.

What particularly stood out to me in Cochrane's text was a point I have repeatedly made in the past; So what? Does it matter to us if inequality has increased? For most people, economics professors and media pundits alike, the question is meaningless and somehow self-evident. Of course it does – and governments need to do something about it! In our Post-Piketty, Post-GFC era there isn't even much discussion over the objective issue: has inequality even increased? Measured from when and where?

Of course, unpacking that question reveals that it's not as straight-forward a question as you may think. What kind of inequality are we looking at? Let's go through some problems.

Leaving moral or political or even bigger economic questions aside, not even "purely statistical" inequalities are very easy to figure out. Are we looking at hourly wages? At total earnings? Since most households have some substantial intra-household sharing of resources, maybe we should look at households' rather than individuals' total earnings? Then again, in the real world, governments act, ensuring that whatever "market outcome" we could statistically show is never ever observed, and so we'd have to look at post-tax, post-transfers for accurate figures of what people face.

But our issues don't stop there. Let's say we settled on one of the above versions, what's the correct measure to use? Is it the widely-used GINI coefficient? Variances in distributions? Or perhaps the P90/P10 or P90/P50 or P90/P10 metrics, where we compare the incomes of the 90th percentile with the 50th or 10th percentile (in other words, if there are only 100 income earners in our society, lining them up according to income, we'd compare the income of the 10th highest earner – the 90th percentile  with the 50th highest – the 50th percentile – to construct the P90/P10 statistic).

Normally, researchers study all of these metrics since they can reveal different things. For instance, we can know that the top-income earners have increased their incomes faster than others if the P90/P50 statistic increases faster than the P50/P10. But even this doesn't capture all the problems. The next issue is the raw data; if we look at tax returns only, the lowest earners are excluded, since most countries have a minimum threshold below which you pay 0% income tax (and so they don't show up in data over tax returns); if we rely on survey data there's the accuracy problem and the fact that we seldom capture the very richest – which to a certain extent is true for tax returns too, my leftie friends never cease to point out. Both these data sources can bias our findings upwards or downwards.

Three more pitfalls, before I'll dive into some comparisons. First, all of the above reasoning concerns income inequality only, while many commentators (including Oxfam's infamously moronic survey) emphasize wealth or some combination of accessibility to, say, higher education. Second, there's the problem I discussed while reviewing Ellenberg's book: comparing percentage changes (say income gains) that can be negative can royally mess up your findings. Third, and most importantly, there's such a thing as social mobility, which means that people are moving in and out of whatever income bracket one is considering. Certainly, there's a similar argument going on about whether social mobility is more difficult for today's generation, but to the extent that there is social mobility, that makes our long-term decile-watching much less accurate. Not to mention the life-cycle fact I've discussed before: our individual incomes change over our lifetime as we accumulate experience and knowledge, and so changes in demographics or immigration or education can seriously skew our inequality figures one way or another.

In short, there are about a thousand ways inequality statistics could go wrong. Therefore, it always surprises me when somebody makes a strong case one way or another. Considering all the things that can go wrong with any one comparison of inequality, I'm always seriously sceptical when The Guardian or your average leftie unwaveringly claim that inequality is "rising, mounting, exploding" or "out of control" – not to mention well-known scholars such as Piketty. There's an avalanche of statistical mistakes that could have skewed their results one way or another, and I honestly doubt that their data or statistical skills are good enough to accurately account for everything (see Piketty and his countless and over-discussed mistakes).

Enough theorising, let me show some graphs on British Inequality that may run counter to the story of inequality we're constantly bombarded with. The data for the below graph comes from OECD (look up U.K. and scroll down for inequality metrics).

If they are even remotely accurate, British income inequality increased until about the time of Oasis' first album, after which income inequality has been fairly flat, if not falling according to some metrics. That is, for my entire life, British income inequality hasn't increased.

The inequality database SWIID (Standardized World Income Inequality Database) even suggests that income inequality in Britain has fallen during the last twenty-plus years:

Of course, the same caveats apply; there's like a thousand ways this data can go wrong, statistically showing something that isn't. Add to this the more important discussions about economics, philosophy or politics whether it matters at all. No wonder so many researchers find it exciting!

No comments:

Post a Comment