What’s really wrong with NAIRU
Earlier this year, Matthew Klein at FT’s Alphaville, made a typically punchy and well-argued case for abandoning a cornerstone of the current macro policy framework – the NAIRU.
This inelegant acronym refers to the rate of unemployment which is consistent with stable inflation. It is one of a set of similar concepts prevalent in macroeconomic policy thinking. The ‘output gap’, a measure of spare capacity in the economy, attempts to play a similar function at the economy-wide level. If the output gap is zero, we might expect inflation to be stable, if there is a shortage of capacity, inflation rises.
The current sensible and pragmatic consensus is that these are useful concepts, very useful pedagogic devices, and a reasonable way to frame policy. Dissonance emerges because the output gap and NAIRU are perceived to be extremely hard to measure – perhaps too hard to be of any use. That position can be accepted by those who still believe in output gaps and NAIRUs.
Simon Wren-Lewis in his response to Klein, perhaps exemplifies this perspective – he believes there is a NAIRU and an output gap, they are extremely difficult to measure, but the framework is broadly accurate and can be useful. He sees intelligent ways of dealing with the measurement problem. For example, he suggests, quite reasonably, that the US Federal Reserve Board should wait until inflation is rising, and the NAIRU has been demonstrably breached, before significantly tightening policy. Jo Michel also outlines a similarly nuanced, and critical perspective – with an useful link to his exchange with Klein on points of agreement and dissent.
I have adhered to this broad framework of thinking about the macroeconomy since I first studied economics at university. One of my earliest personal economic influences was Professor Ken Mayhew at Oxford, one of Britain’s leading labour economists, and a periodic co-author with both Richard Layard and Steven Nickell who first introduced the ‘NAIRU’ in 1986.
The objective of demand-management – either through fiscal or monetary means – is to target the level of aggregate spending consistent with an economy fully utilising its resources. A zero output gap is a proxy for this. Concepts of unemployment consistent with stable inflation are somewhat subtler, because they focus on the institutional features of the labour market – reducing unemployment is not just about aggregate demand, improvements in labour market efficiency, either by reducing search costs or improving the matching of skills – can all reduce the rate of unemployment without negative trade offs – such as higher inflation.
Like much of modern macroeconomics, this entire framework really originates with the problems of stagflation in the 1970s and with the Philiips Curve framework of the 1960s – which Simon and Matthew refer to, and I have discussed before.
That is also its weakness. I am going to suggest a completely different structural description of how the current economy works, in stark contrast to that which prevailed in the 1970s and early 1980s. In the world I will describe, there is no NAIRU – it is not a useful concept, in fact it is highly misleading. Not because it cannot be measured – but because it does not exist. This, I think, is a far strong position than the most vociferous critics have so far made. And I will go further – policy based on the NAIRU may be a costly distraction.
Here goes: Let’s assume that product markets are highly competitive, in the textbook sense that firms are price takers. Let’s assume the same is true of labour markets. To be clear, I am not assuming that prices and wages are symmetrically “flexible” – that is empirically false – I am only requiring that firms and individuals lack market power, and specifically lack the power to determine their prices and wages. This is not an idealised position, in the sense that there will obviously be “negotiations” even in non-unionised worker-employer relationships, where the costs of retraining etc., are reflected in a ‘bargain’ between worker and employer. But that is not my point, I am simply assuming that the broad wage-determining process is determined by demand and supply for specific skill sets.
How do expectations fit into this view of the economy? I am also assuming the Greenspan definition of price stability prevails. In other words, inflation is sufficiently low and has been sufficiently low for long enough that people (‘economic agents’) don’t really pay any attention to it. Put more accurately, I am assuming a world where almost all the information is in relative price variance not aggregate price variance. In other words, when wages for a given skill set fluctuate, or prices of goods and services change materially, participants do not assume anything about aggregate inflation, they draw a signal from the price change about their specific circumstances. Price and wage variance has relative price information and there are no relevant aggregate “inflation expectations”. If people are asked, what do you expect aggregate inflation to be they reply, “don’t know, don’t care.”
In the economy I am describing, people simply do not have “inflation expectations” (nor are they represented by entities, such as trade unions, which have inflation expectations), but they do have knowledge about when demand for their skills (or their products) is strong or weak.
I need to be absolutely clear that this is not some neo-classical nirvana without recessions, or collapses and shortfalls in demand – it absolutely is. I have no problems with the idea that firms and workers can be price-takers and there can be brutal recessions. Accepting that labour and product markets are competitive by no means assumes that they always “clear”. Rigidities, balance sheet problems, coordination problems, errors, animal spirits, credit conditions – none of these potential sources of recession or depression have gone away – they exist in my model.
Ok, so how does the NAIRU fit into the world I am describing? It doesn’t. What happens in the economy that I am describing when unemployment falls to extremely low levels is relatively unpredictable, but not particularly harmful nor requiring any policy response. I am happy to let the system sort it out. Labour shortages may take time to be addressed – there may be a need for more training, there may be a need for more migration, there may be need for firms to substitute capital for labour. I don’t mind – the market system can sort that out, and smart policy-makers can facilitate it with micro-policies. It may be the case that profit margins get squeezed, and a lengthy process occurs of lower return on capital, lower investment, and ultimately lower demand for labour.
This is also a world where variance in certain market prices – commodities being a striking example – and sectoral credit cycles, are the main source of moderate cyclical variance. Something Frances Coppola has astutely suggested. I think it is highly plausible that idiosyncratic sectoral shocks, for example in autos, segments of housing, and the commodity sphere have been far more relevant since the financial crisis to the US economy than monetary policy. And this is the norm.
What I am describing is a world where monetary policy – and specifically the level of overnight interest rates – is really not very significant. Other relative price signals, sectoral shocks and credit cycles, sort themselves out. And credit conditions are determined far more by the behaviour of the financial system and private sector demand for credit. Again overnight interest rates are relatively peripheral. Market shifts in risk premia and the regulatory backdrop dominate the pricing of credit and finance, too.
It is again important to stress that is not a naive Panglossian market-clearing world view of the world. This is definitively a world where sufficiently large, correlating, aggregate shocks occur – the financial crisis and Eurozone austerity being recent examples. And I have long been advocating the need for profound and effective counter-cyclical contingency policies.
Increasingly, these are the lines along which I think we need to rethink our economic structure – macro-theory may itself be regime and structure dependent. The framework of the 1970s – of rational expectations, NAIRUs, and ultimately interest rate-setting central banks managing demand – is a regime of relevance to some parts of the world (perhaps economies such as South Africa, Brazil and Turkey, with higher initial-condition real interest rate structures, and institutionalised inflation expectations), but no longer to the developed world.
I am conscious that, to some extent, I am overstating my case. There is considerable institutional variance across developed countries. But I am increasingly of the view that the standard inflation expectations/output gap/NAIRU framework is more profoundly flawed. The concepts are useful pedagogic devices, but given the current micro-structure of the developed world, they may be significantly redundant, and worse, a hindrance to policy-making.
Central banks may need to accept that changes in overnight interest rates really don’t matter very much. Absent large correlated shocks to demand, there is nothing for them to do, and when they are needed, they will need new tools. Periodically there is a desperate need for aggressive stimulus of demand – such as in the Eurozone currently. Where Simon Wren-Lewis and I are fully in agreement, is that we need independent central banks with proper counter-cyclical policy tools. My concern is that a deeper overhaul of the entire conceptual framework may be needed to get them there.
Nice post, Eric.
I agree that the concept of NAIRU is a hindrance. Policymakers may err on the side of not doing enough because they’re scared that if they do more, inflation may start rising without a limit. Hell, they’re even worried that they won’t even be able to successfully test because once they have expanded fiscal policy, it might be difficult to roll back. Of course this won’t happen as there’s no NAIRU but the fear is still there because theorists have made them fearful.
What I more surprised is that there’s a view that the existence of a non-flat Phillips Curve necessarily implies NAIRU. No. No. No. I mean there could be thousands of different dynamics there — including the opposite i.e., inflation falling long term.
Thanks Ramanan – agreed.
“Central banks may need to accept that changes in overnight interest rates really don’t matter very much.”
Any interest rates?
“The funny thing is: they haven’t. In fact, among the more than 10,000 research articles produced by the major central banks in the two decades prior to the 2008 crisis, none explored the correlation or causation between nominal interest rates and nominal GDP growth. Fortunately, this task is not very demanding, and once we conduct such an examination, we conclude that, in actual fact, there is no evidence to back these assertions whatsoever. To the contrary, empirical evidence shows that the central banking narrative on interest rates is diametrically opposed to the observable facts in two dimensions: instead of the proclaimed negative correlation, interest rates and economic growth are positively correlated. Secondly, the timing shows that interest rates do not move ahead of growth, but instead are either coincidental or even follow it.”