Nominal wage rigidity in macro: an example of methodological failure

This post develops a point made by Bryan Caplan (HT MT). I have two stock complaints about the dominance of the microfoundations approach in macro. Neither imply that the microfoundations approach is ‘fundamentally flawed’ or should be abandoned: I still learn useful things from building DSGE models. My first complaint is that too many economists follow what I call the microfoundations purist position: if it cannot be microfounded, it should not be in your model. Perhaps a better way of putting it is that they only model what they can microfound, not what they see. This corresponds to a standard method of rejecting an innovative macro paper: the innovation is ‘ad hoc’.

My second complaint is that the microfoundations used by macroeconomists is so out of date. Behavioural economics just does not get a look in. A good and very important example comes from the reluctance of firms to cut nominal wages. There is overwhelming empirical evidence for this phenomenon (see for example here (HT Timothy Taylor) or the work of Jennifer Smith at Warwick). The behavioural reasons for this are explored in detail in this book by Truman Bewley, which Bryan Caplan discusses here. Both money illusion and the importance of workforce morale are now well accepted ideas in behavioural economics.

Yet debates among macroeconomists about whether and why wages are sticky go on. As this excellent example (I’ve been wanting to link to it for some time, just because of its quality) shows, they are not just debates between Keynesians and anti-Keynesians, so I do not think you can put this all down to some kind of ideological divide. I suspect nearly all economists are naturally reluctant to embrace cases where agents appear to miss opportunities for Pareto improvement - I give another example related to wage setting here. However in most other areas of the discipline overwhelming evidence is now able to trump these suspicions. But not, it seems, in macro.

While we can debate why this is at the level of general methodology, the importance of this particular example to current policy is huge. Many have argued that the failure of inflation to fall further in the recession is evidence that the output gap is not that large. As Paul Krugman in particular has repeatedly suggested, the reluctance of workers or firms to cut nominal wages may mean that inflation could be much more sticky at very low levels, so the current behaviour of inflation is not inconsistent with a large output gap. Work by the IMF supports this idea. Yet this is hardly a new discovery, so why is macro having to rediscover these basic empirical truths?

There may be an even more concrete example of the price paid for failing to allow for this non-linearity in wage behaviour. For all it inadequacies, the Eurozone Fiscal Compact does at least include a measure of the cyclically adjusted budget deficit among its many indicators that are meant to guide/proscribe fiscal policy. However, as Jeremie Cohen-Setton discusses here, the Commission now think they have been underestimating the output gap. As he suggests, the reason is pretty obvious: they have overestimated how much the natural rate of unemployment has risen in this recession. Here is the example he gives for Spain.

Actual unemployment and European Commission estimates of the NAWRU for Spain, from European Commission, 2013 spring forecast exercise. Source Jeremie Cohen-Setton

How could the Commission have been so foolish as to believe the natural rate had risen from 10% to 27% in a few years? Might it be because they looked at nominal wages in Spain, and inferred from the fact that nominal wages were not falling that therefore actual unemployment must be close to its natural rate? If empirical macromodels as a matter of course allowed for the absence of nominal wage cuts, would they have made such an obvious (to anyone who is not a macroeconomist) mistake?

I think this example illustrates why it can be dangerous to rely on DSGE models to guide policy. Yet the influence of DSGE models in policy making institutions is strong and growing. The Bank of England’s core forecasting model (pdf) is a fairly basic DSGE construct, and as far as I can see its wage equation is a standard New Keynesian specification, with no non-linearity when wage inflation approaches zero. Now I know the Bank have many other models they look at, and they will undoubtedly have looked at the implications of a reluctance to cut nominal wages. (I discuss the Bank’s ‘new’ model in more detail here.). However default positions are important, as the examples I discussed earlier show. Focusing on models where consistency with fairly simplistic microfoundations is all important, and consistency with empirical evidence is less of a concern, can distort the way macroeconomists think.   

Unemployment, the output gap and wage flexibility

This post is about the impact of nominal and real wage flexibility on unemployment and the output gap. It starts in an academic, abstract sort of way, but the policy implications do follow. I try and make the analysis as accessible as I can to non-economists.

Start with an economy with a zero output gap (defined below) and no involuntary unemployment. Everything in the economy is just fine, which is a non-technical way of saying it is efficient. Then a ‘crisis’ happens that leads consumers to consume less and save more, so aggregate demand falls. Normally in these situations the central bank cuts nominal and real interest rates sufficiently to restore aggregate demand. Once this has happened, call everything in this economy ‘natural’, so the real interest rate that restores demand is the natural rate of interest. The natural level of output may not be the same as the pre-crisis level, because for example the new natural rate of interest can have knock on effects on how much people want to work. [1] However the natural level is the level of output that policymakers should aim for. [2]

In the Great Recession this mechanism did not work because nominal interest rates hit zero, and maybe also because monetary policy put a cap on inflation expectations. As a result, actual real interest rates are above the natural level. In addition, fiscal policy is in the hands of people who know nothing about macroeconomics, so there is no help from there. However monetary policymakers still think they could do something ‘unconventional’, so they want to know what to aim for. The answer is that, as long as what they do does not seriously distort the economy, they should try to get to the natural level of output, because that produces an efficient economy.

The difference between the actual level of output and the hypothetical natural level is called the output gap. The traditional way of defining the output gap was the difference between actual output and ‘productive potential’, which was the amount that could be produced if all factors of production were fully utilised. That is still how the gap is often measured in practice, although the measurement problems can still be huge, as Paul Krugman notes here. The problem at a conceptual level is that this approach downplays considerations of optimality, so nowadays theoretical macroeconomics uses the natural level of output to define the output gap. This has the advantage that we know what policy should be aiming to do: achieving the natural level of output.

Now imagine three almost identical economies where an output gap exists because nominal interest rates have hit zero. The level of real interest rates that would eliminate the output gap is the same in all three economies (i.e. they have the same natural levels of output). In the first economy, workers resist nominal wage cuts, so this puts a floor on how much unemployment reduces real wages. (Equally firms may be reluctant to impose wage cuts, as this research suggests - HT Kevin O’Rourke.) If nominal wages stop falling, at some point firms will stop cutting prices to protect their profits. We settle down to a new lower level of demand deficient output, high unemployment, but stable wages and prices. There is plenty for unconventional monetary policy to do, even though inflation is not falling.

In the two other economies nominal wages carry on falling. In the second economy prices get cut pari passu, so real wages remain unchanged, while in the third they do not, so real wages fall. So in the second economy inflation is lower than in the first, but real wages are the same. Does this lower rate of inflation increase or decrease the output gap? That depends only on whether actual output falls or increases because of lower inflation: the natural level of output involves a hypothetical economy which is unaffected by whether nominal wages fall or not in the actual economy [3]. Actual output may fall if negative inflation makes debtors spend a lot less but creditors not much more - this and other mechanisms are discussed in Mark Thoma’s post here. However, if monetary policymakers have been inhibited from doing much because inflation was not falling (which would be one interpretation of UK policy, for example), then as David Beckworth says, lower inflation may raise actual output by encouraging expansionary unconventional monetary policy.

How about the third economy, where real wages have fallen? Suppose firms respond to lower real wages by substituting labour for capital, and this process continues until all those who want to work can find a job. So in the third economy involuntary unemployment goes away. But is the output gap any lower? Once again, the natural level of output has not changed. (It was set in our hypothetical economy where real interest rates fell to their natural level.) So the key question becomes whether lower real wages and lower unemployment reduces or increases aggregate demand, and therefore actual output. It could go either way. So it is perfectly possible that both actual output and therefore the output gap is exactly the same in all three economies, even though unemployment has returned to its natural rate in one, and the other two have very different inflation rates. 

This comparison suggests that those who say unemployment in the first two economies is caused by wage inflexibility kind of miss the point. The basic problem is lack of aggregate demand. You could argue (I would) that the third economy is better off than the other two, because the pain of deficient demand is evenly spread (everyone has lower real wages), rather than being concentrated among the unemployed. But the first best solution is to raise aggregate demand, because that gets rid of the pain.

I started writing this post because of a recent study by Pessoa and van Reenan, who argue that the mysterious decline in UK labour productivity that I have talked about before can in large part be explained by unusually slow growth in UK real wages. The mechanism they have in mind is entirely traditional: if real wages are low firms substitute labour for capital. This in turn may explain (see Neil Irwin here for example) why UK unemployment originally rose by less than in the US (see first chart), even though the UK’s output performance was worse. On this issue looking at consumer price based measures of real wages will be misleading, so below is a very simple measure of real product wage growth in the two countries: compensation per employee less the GDP deflator. Real wage growth in the UK has noticeably fallen since the recession, whereas the fall has at least been less abrupt in the US (2013 is a forecast).

Unemployment in the US and UK: Source ONS and BLS

Growth in compensation per employee less GDP deflator: OECD Economic Outlook

In terms of just the UK economy, whether Pessoa and van Reenan are right is debatable. When I discussed this in an earlier post I referenced a Bank of England paper by MPC member Ben Broadbent, which argued that for the factor substitution story to explain most of what we have seen in the UK, investment should have completely collapsed, which it has not. This difference in view reflects a number of nitty gritty issues, like how you measure the capital stock, and whether the substitution elasticity is one (as implied by the Cobb Douglas production function), or nearer one half.

However most seem to agree that some of this factor substitution is going on in the UK. So my hypothetical discussion above suggests that, by spreading the pain of deficient aggregate demand further, this ‘real wage flexibility’ in the UK has been a good thing, but it does not mean the aggregate demand problem has decreased. If anything, it suggests that looking at unemployment underestimates the size of the output gap. Monetary policy makers please note.

[1] New Keynesian economists sometimes call the natural economy the outcome when all prices are completely flexible. That is OK, as long as we note that flexible prices here has to include the possibility that nominal interest rate can go negative, which in the real world it cannot.

[2] Opinions may differ on whether the crisis itself is a necessary correction for past errors, or whether it is itself a distortion. For example, was risk undervalued before the crisis, or is it overvalued now. In other words, was the pre-crisis economy efficient, or would there be a distortion in the post-crisis economy even without an aggregate demand problem? These are important complications compared to the story I tell here, but they will have to wait for another post.

[3] The idea is that the economies are identical except for the extent of nominal inertia in goods and labour markets. In economy 1 wages are sticky, in economy 3 prices are sticky but wages are flexible, and in economy 2 the degree of wage and price stickiness is such that real wages do not change.