WSJ suggests abandoning economic models

Simon Nixon has a provocative article in the WSJ where he argues that the current generation of New Keynesian models are useless because of their poor forecast performance. He proposes looking solely at the rate of debt reduction when forecasting economic performance:

[The] dismal science’s [forecasting] record suggests is that there is something profoundly wrong with the mainstream economics profession’s understanding of how modern economies work. The models on which its forecasts are built are clearly badly flawed.

It is true that forecasting performance is poor, but that is largely because forecasting is very, very difficult. The DSGE models Nixon refers to are important in modern macroeconomics but they were originally designed to estimate the impact of monetary policy, not to forecast the future. In fact, until very recently, most forecasting was not done with structural DSGE models but with statistical models that take their structure from the data. They provided better forecast performance and so were preferred by professional forecasters. These days, the best forecasts tend to be made by estimated DSGE models that outperform the best statistical models because they incorporate some of our understanding about how the economy works. No doubt they will be improved over time but it is incorrect to suggest that forecasters are too constrained by theory to forecast accurately. In fact, economists’ understanding of the economy helps them to provide better predictions about the future than simply using statistical relationships.

Nixon then discusses a paper by Claudio Borio at the BIS, which suggests building models that describe not only business cycles but also ‘financial cycles’. Borio’s paper highlights the monetary nature of the current recession and recommends that the next generation of macro models give serious consideration to the slow buildup of disequilibrium forces in financial booms, which then trigger deep recessions. Whether or not you agree with that diagnosis, the question he tackles is crucial: how do crises endogenously develop? Part of the reason forecasting is so difficult is that turning points are hard to pick because we don’t really understand all of the mechanisms that lead to recessions. Nixon uses that paper to claim that…

…[for] investors, the sensible response is surely to disregard all short-term forecasts based on out-dated models. They should focus instead on identifying those economies most likely to deliver a medium-term recovery by aggressively addressing their stock of debt. In the European context, it is the euro zone where the process of debt reduction and restructuring seems likely to proceed most rapidly, not least because the greater independence of the European Central Bank means there is less prospect of loose monetary policy being used to defer tough decisions.

I don’t think that is what Borio’s paper claims. Can Nixon really be advising you to ignore expert forecasters and instead put your money into EU countries, many of whom are currently facing the possibility of further recession in 2013? He has good company in suggesting that current economic models have problems and could be improved; however, the chances that they can be improved upon by following a simple heuristic like ‘less debt equals more growth’ are exceedingly slim. Indeed, there is considerable public debate among economists over the impact of debt on growth. Reinhart and Rogoff’s work has generated a lot of discussion, and there is a paper entitled ‘Macroeconomic Risk and Debt Overhang’ being presented at the ASSA conference. It’s hardly a matter that has been neglected by the profession!

Of course, it’s very difficult to diagnose and fix a problem with a single newspaper article: witness Paul Krugman’s repeated attacks on the current state of macroeconomics and the heated responses that they’ve generated, for instance. When even Nobel prize-winners can’t agree on whether there is a problem it is a sign that we don’t really understand what needs to be done. It’s great that Nixon is bringing interesting papers like Borio’s to public attention and airing the debate that’s going on in the profession. Unfortunately, in this case I don’t think his diagnosis or proposed solution are quite right.

Fiscal multipliers are unhelpful

John Quiggin has re-opened the fiscal multiplier debate to advocate for fiscal stimulus. Quiggin, along with others such as Krugman, Summers and DeLong, and Blanchard claim that the effect of government spending on production will be greater than the government’s initial injection. The empirical evidence they use tends to rely on cross-country regressions, although some calibrated modelling has been done by NIESR.

An important caveat on these studies is that they rely on the unusual macroeconomic circumstances of the current recession. In ‘normal’ times one would expect that a central bank would lean against fiscal policy, resulting in very small multiplier effects. Scott Sumner has discussed this point extensively. To summarise, he claims that:

It sort of implies ‘the’ multiplier is some sort of stable parameter out there, waited to be discovered. Like the cosmological constant. In fact, it is nothing more than an estimate of central bank incompetence, which will vary from one case to the next.

Obviously the illustrious authors of the multiplier studies aren’t unaware of this problem. The general theme of their arguments is that we are currently in a liquidity trap and monetary policy has little traction in these circumstances. Central banks are thus unable to counteract the effects of fiscal policy, which is only doing what the central bank would do itself if it could: boosting demand. Sumner rejects the idea of a liquidity trap, hence the disagreement.

Obviously, these estimates of fiscal multipliers are entirely contingent on the response of the monetary authority. As discussed on Vox, the characteristics of an economy and monetary policy regime can cause the multiplier to vary between zero and 2, which is basically the difference between being in favour of fiscal stimulus and considering it a complete waste of money. Without estimates for each country individually that means the average multiplier is likely to be a poor guide to the multiplier in an individual country. That doesn’t make the estimates ‘wrong’ but it does mean that cross-country estimates are unreliable as a guide to national, fiscal policy. From a policy perspective then, it’s unclear that these estimates provide much guidance on the extent of fiscal stimulus or austerity; at least not without a lot of investigation of country-specific factors.

Update: FT said much the same thing. Simon Wren-Lewis suggests that the theory on this is obvious and people just got too caught up in the empirics.

The case for not cutting

There is a growing call for rate cuts to the OCR in New Zealand given the high unemployment rate, indications that the September quarter was very weak, and the fact people are pissed off that the weakness in the New Zealand economy has been so persistent!

Now I’m not going to go one way or the other on this – after all I don’t really want to second guess the Reserve Bank.  However, the case for a rate cut appears to be weaker now than it was earlier in the year.

How can I say this?  The unemployment rate is undeniably higher.  Well remember that unemployment is a lagging indicator – usually the economy is well into picking up before we see a sustained drop in this.  You may retort (I know I would) with the hours worked figures, which have been very weak.  Hours worked is usually the first thing to pick up (either with or a bit after productivity) during a recovery.  For this all I can say is that hours worked are not as weak as they appear in the HLFS, but we would need to forecast them picking up soon!

Ultimately, we need to ask ourselves what a RBNZ forecast would need to look like to prevent a cut.  We would need them to first forecast no cut, and then to forecast an economy moving back to it’s “potential” level.  This will then be consistent with a forecast of inflation around the target band.

Why might we believe that the economy is heading back to potential (and without a lift in structural unemployment this would imply a swift drop in the unemployment rate in the coming years as well).:

  1. The lift in house sales and (soon to be) house construction – this rebound in durable good spending and investment tends to lead the economic cycle.  Generally households willingness to get involved in these things tells us that demand in the economy is on the up.
  2. Durable good sales to households have risen (although part of this is to builders and plumbers rather than consumers), and business investment has risen … business investment has dropped off in recent months, as part of the prior spike was “rebuild related”.
  3. A similar rebound in the US – with the prospects for the US picking up, underlying demand for a number of our export commodities (dairy, meat, logs) will firm.  Let’s not forget that the US is a big market for our (likely mismeasured) IT services export industry.  Why mismeasured – well if you know anyone who sell services online, you will know that they often avoid tax or business registration 😉
  4. Signs China has found its feet again
  5. Commodity prices are recovering sizably
  6. Easing bank funding costs.  The growing competition between banks in recent months is likely due to easier access to credit – there are reports this is flowing into businesses, albeit not evenly.
  7. The rebuild is now really getting underway.

This isn’t to rule out cuts – I’m avoiding taking a position here, as I want to save that for clients, and generally avoid upsetting people on the internet right now (what can I say, I’m a bit tired).  All I am saying is that, given the time it takes for a lowering of the cash rate right now to flow into the domestic economy, a rate cut when a lot of indicators have turned up in the last couple of months.

Also remember, if the Bank had been able to foresee what occurred through the middle of this year they would have cut earlier on – but they couldn’t foresee it.  This is not a criticism at all, because the Bank does have incredibly good “on average forecasts”, and as a result their actions can minimise the cost of policy mistakes.   But it does indicate that the Bank’s actions aren’t infallible, and that they should publicly explain what happened when we experience a situation of below target band inflation and rising unemployment to the public – instead of leaving all the commentary up to people who want to undermine them.

Fiscal policy and the ZLB

Some links.

Marginal RevolutionMainly Macro.

Summary:  The implied impact of fiscal policy at the ZLB is the same channel as direct monetary policy (such as QE).  Response is, yes but we have more certainty about the impact of fiscal policy.

Sidenote:  As the central bank directly does balance sheet management instead of changing the cash rate, “monetary policy” looks closer to “fiscal policy”.  There are four key differences though:

  1. Central banks have the incentive to reverse out unconventional policies once their job is done – politicians don’t.
  2. Central banks will respond to the data more quickly than politicians.
  3. The “inflation target” gives us the approved mandate of the central bank through the policy process.  As a result, central bank action based on this view are appropriate – as long as the instrument used is assumed to not redistribute over the economic cycle.
  4. Central banks don’t have a direct mandate to redistribute – elected officials do.  For central banks to use a tool that redistributes, they have to be given permission by a democratic government.

Expectations formation, inflation, and a role of inflation targeting

Ever since the Lucas Critique forced economists to recognise that we could not use data for policy in a purely theory free way, the concept of expectations, and how expectations are formed has become important.  The growing interest in behavioural and neuroeconomics are all in part a response to this realisation, and a clearer understanding of these issues will help give economists an idea of what to study, what to measure, and what policy trade0-offs exist.

It is encouraging then to see this study on the formation of inflation expectations.  In it they look at how shocks to food prices impact upon individuals expectations of future general inflation.

Our results suggest that consumers incorporate information about past food prices in forming and updating their own-basket inflation expectations but not their overall inflation expectations. The issue of pass-through to inflation is of particular concern during times of large supply shocks. Our finding that information about past food price inflation has limited pass-through to consumers’ expectations of the “rate of inflation” suggests that the RI question (and not the PP question, which as mentioned above is similar to the one used in the Michigan survey) is a more stable survey question (in the sense that it is less susceptible to volatile price changes), and that it should instead be used to elicit consumer inflation expectations.

This is all well and good, but we need to ask whether we are really gauging inflation here in the way we mean it when we discuss monetary policy.  We are interested in the way prices in general move together that is unrelated to “fundamentals”.  A lift in food prices is a change in those fundamentals, and tells us that there is a change with regards to how scarce food is relative to other goods and services and relative to labour.

This general “comovement” we are talking about is anchored by an inflation target, and as a result we want a description of how a “shock” to the price of one good has an impact on this over the complicated beast that is the economy.  We need to seperate out the bit that is due to households feeling “poorer” or “richer” following a change in the price of one good or service (a relative price, or supply shock) from the impact it has on the general price setting of households and firms (our inflation).

This is easier for us to think about if we assumed the “inflation target” was zero – in that case, as decision makers we know that a rise or drop in prices represents a lift or decline in scarcity relative to other goods and services.  Furthermore, a increase in wages represents productivity – or may represent a change in bargaining power.

However, with an inflation target, which is followed by decision makers, we can say the same thing by just taking off the inflation target from price growth.  This is one of the conceptual reasons why an inflation target is an attractive thing – we are helping to make price signals in the economy a clearer representation of underlying scarcity.

As the gold standard period showed us, inflation/deflation in itself can be very unstable, so an inflation target through inflation expectations can be used to help “co-ordinate” decision makers.

In this sense, it is nice to see expectations formation being discussed here – but I’m not sure the study is truly capturing “inflation” following a relative price shock, and is instead also capturing other factors related to economic fundamentals.

It’s about the “right” counterfactual

In a recent post, Mark Calabria from the Cato Institute took aim at the idea that the Lehman Brothers crisis was the “trigger” for a big crisis.

Now I do not disagree that there was, and would have been, a recession without Lehman Brothers – and even without the uncertainty caused by the lack of clarity around insurance of the shadow banking system which grew post August 2007.  However, these issues, and in turn the failure of Lehman Brothers did make the crisis significantly more severe than it would have been.

He appears to say that the failure of Lehman Brothers was a good thing (Note:  There is nothing wrong with wiping out the company – but the way it was handled, was a big driver of the global slowdown that was to come), and that the US was on the road to recovery post this.  So here we are focusing just on the US, not the contagion to other countries.  His evidence is the following graph:

Employment and consumption stopped declining not long after Lehman Brothers failed, and although the largest declines occurred WHEN Lehman Brothers failed this doesn’t mean the failure caused them – in fact, employment tends to lag the cycle and the drop may well have been the result of prior economic weakness … and the amazingly high fuel prices through the first half of 2008.

Now I agree that there were factors driving a recession prior to the failure of Lehman Brothers – but the impact of Lehman Brothers as an event is captured by asking what would have happened in the absence of the Global Financial Crisis that stemmed from it, and the full blown “bank run” on wholesale financial markets that had been building pressure from the start of 2008.  Going to FRED, grabbing consumption and population, and running a basic time regression in excel no less (so it’s easy to copy) we can get an idea of what the “trend” rate of consumption per capita was during the 1952-2012 period.  Armed with that, we can ask what the percentage difference is between this trend and actual consumption per capita outcomes.  This is:

Something is broken here – I will try to fix that up tonight.  The strange thing is that I can see the graph when editing the post … but it then wont let me do anything with it

Now we can start arguing that consumption per person was too high and a whole bunch of other things if we want to here.  However, this basic analysis clearly shows that the gap between trend consumption and actual consumption, something that should have a tendancy to head back to zero after a recession, actually deterioarted further … and has continued to deteroriate.  We look at business cycles “around trends” not “around levels” given that our counterfactual involves growth – and this makes this post by the Cato institute a bit misleading.

Saying that the downturn, or even the crisis, started with Lehman Brothers is wrong, I agree with the author here – however Lehman Brothers failure started a new dangerous stage of the crisis which, when combined with the persistent institutional failure in Europe, has made sure that the US economy has remained below potential.  A market monetarist would say that the Fed is truly responsible for this in terms of policy action, but even if we were to accept this it is undeniable that it is the “shocks” that have occurred in financial markets are the very things the Fed needs to respond to by “loosening policy”.

It’s failure is indicative of what was underlying the crisis, and the evidence shown in no way suggests that allowing its failure and initially ignoring the quiet, and then full scale, bank runs in wholesale financial markets was good policy.