Some ‘inequality’ is good (and other unpopular statements)

We have a tendency as individuals to define things as “inherently good” or “inherently bad”. And when this comes to policy indicators this is dangerous.

We have a tendency as individuals to define things as “inherently good” or “inherently bad”.  

But when this comes to policy indicators this is dangerous thinking.

Bah, inequality is bad – it’s obvious

Yes, yes, the most common response I get – but you’re here now, so let's have a think about what we are doing.

Let us start with a couple of quotes from Dalton (1920) - where Hugh Dalton was trying to push English economists to think about why we may care about inequality, and how to conceptualise this.  

First, when discussing the view that the economist studying the inequality in the distribution of wealth is equivalent to a biologist studying the inequality of any physical characteristic he states:

"But this is clearly wrong.  For the economist is primarily interested, not in the distribution of income as such, but in the effects of the distribution of income upon the distribution and total amount of economic welfare which may be derived from income."

What does this mean for the economist's problem?

"We have to deal, therefore, not merely with one variable, but with two, or possibly more, between which certain functional relationships may be presumed to exist."

Here we go.  We care about social welfare, which we can’t observe.  We can observe some indicator of income and income inequality, and we have to ASSUME some functional relationship between this measure and social welfare.  Given this assumed set of value judgements we can discuss our preference for inequality.

This of course does not mean we should ignore the issue altogether – instead we can ask which potential distributions of income will be favoured given an assumption about the form of social welfare.  In this case we get a clear idea of the set of assumptions that would need to hold for us to favour less or more equality in the distribution of incomes.

Accepting that this functional relationship is unobservable – a limit in the social sciences that does not exist to the same degree in the physical sciences – we can discuss certain principles.  Atkinson (1970) [REPEC] got this underway by comparing distributions with the same mean, but differing degrees of inequality.  Given diminishing marginal utility (or more directly, risk-aversion among individuals), the more equal distribution is preferred.  This concept – extended into the Atkinson-Shorrocks methodology to deal with differing means, and the Atkinson-Bourguignon method for dealing with different assumptions about “need” over household characteristics – provides the base for why many people seem so confident that “less inequality = good”.

Given the focus is solely on measures of “inequality” it is the Atkinson-Shorrocks view that is being leant on.  But using these measures solely as an indication of what is good, and applying them to static measures of inequality, involves assuming specific things – things we may be uncomfortable with.  Say that we introduce a series of value judgements that capture the way we feel income is valued between groups based on household characteristics (an equivalence scale).

By treating the line of ‘complete equality’ as the ‘egalitarian line’ we are at a minimum ignoring:

- The fact that income is not a blob floating around – but the result of co-ordination between those with access to different inputs in a production process. [Note:  Remember much of this literature, given its focus, is treating measured income as endowment]

- Given (1) the differences in the value of marginal value of consumption relative to leisure (the purpose of income) within groups. [Part of the distribution is in fact representative of revealed preferences of people who feel very differently about the relative value of consumption, investment, and leisure]

- Differences, both in characteristics and subjective perceptions of need, based on groups we have not recorded or measured.

- Differences in the individual/households position in their lifecycle.

When we bemoan inequality, we are picturing two individuals who are similar, putting in the same effort, but being rewarded in different ways.  Yes, I can indeed understand how this is seen as unfair, and wanting to reduce this type of inequality is fully understandable!  But this isn’t what static income inequality measures tell us.  Some differences in income at a point in time are due to choices, and more fundamentally our capability to turn income and the cost of work into satisfaction.

A determination to make incomes equal based on these measures ignores the fact that people are inherently different.  At a point in time, an individual who values consumption highly relative to leisure will work more, earn more, and spend more than someone who doesn’t – this factor turns up in our indices as “inequality”, but it is a good thing.  Pushing the person who values leisure to work more and the person who values income to work and earn less will make both individuals worse off.

This is part of the reason why ‘economic mobility’ is so highly focused on by economists and policy makers – people’s ability to shift between bundles of {consumption, leisure} based on their preferences is a great thing.  With mobility, inequality will exist at a point in time, but it will be relatively more to do with “choice” – this is where equality of opportunity comes in.  Of course, we also need to be careful with these measures – sometimes what looks like mobility in a distribution is really just an indication that income is more volatile, which is something individuals and households do not like.

But this is not the whole story

In this discussion we have focused on the idea of inequality for a given level of income.  However, this has ignored the elephant in the room – the trade-off between policies that change the distribution of income and the level of income itself.

Take for example the idea of the government somehow being able to enforce equal renumeration for all jobs.  In this case:

- What happens to the incentive to invest (both in human capital, and in physical capital) – why a large part of it evaporates

- What happens to the choice of work – well people would favour jobs with non-income characteristics.  So “status jobs” and “low effort jobs” would experience an excess supply of workers, while other jobs would experience an excess demand.  If wages are truly fixed, and non-income characteristics that are not part of the role cannot be used, this is very allocatively inefficient.

- What happens to any sum of the absolute gaps between the social value of the work individuals do and its private value – it rises, again more allocative inefficiency.

As a result, these extreme versions of equality WILL lower income – there is an equity-efficiency trade-off along some margin!

This is the kicker, the Atkinson-Shorrock methodology is useful for comparing two distributions when we know what the counterfactual will be given our policies.  Using this we can tell what value judgements have to hold in order to prefer one policy to another.

But the relation of this to a “target level of inequality” is very weak.  We need to know what the impact of the policy is, we need to have a clear idea of the value judgements we hold, and then using these ideas we can choose trade-offs.

Saying  ”let’s just lower this variable” involves ignoring the trade-offs, it involves excessive certainty that what we are doing is “right”, “fair”, and “just”.  This makes me nervous.  It makes me especially nervous when I see the case being based on a book that does not recognise this, and pretends the trade-offs do not exist – the Spirit Level.

This post originally appeared on Sciblogs and TVHE