Friday, June 11, 2010

Permanent Versus Temporary Output Loss Revisited

Just over a year ago, I did a blog post on different types of recessions and the policy responses appropriate to both. Thinking about the IPI this morning and overcapacity in electricity generation and manufacturing prompted me to relook this issue.

To recap, a Friedman recession is one which is associated with an increase in the output gap, which is the difference between potential output based on labour and capital capacity to produce. The Hamilton recession is where potential output itself is dropping, which signals a permanent loss in output.

The difference between the two can be illustrated by the following figure:


I thought at the time that this downturn would turn out to be of the Hamiltonian type, which requires an active fiscal policy response beyond automatic stabilizers, even if some of that effort leaks out through higher imports. A Friedman recession can usually be handled by judicious use of monetary policy alone.

Of course, it’s really hard to tell in the heat of the moment which particular situation you happen to be in, particularly with data lags of 2-3 months for some of the critical data. This is especially true since the defining variable – the output gap – is unobservable and has to be estimated.

Past experience and research output helps guide policy of course, so in the spirit of contributing to that, I’m going to take a rough and ready approach to answering this question. We’re still not quite out of the woods yet in terms of both global and Malaysian economic recovery, so this should be taken as just preliminary, not definitive.

In other words, it’s Friday evening, this is a purely academic exercise, and I’m amusing myself.

First, I’m going to take a purely stochastic (probability-based) non-model based approach i.e. I’m not even going to bother with estimating the output gap (that’s beyond my knowledge at the moment).

Many economic series typically display a consistent pattern – you can take advantage of this fact to model them stochastically by just following a mathematical rule (otherwise known as a “data-generation process” or d.g.p.) to define the shocks. Most follow a d.g.p. that approximates to a 1-lag auto-regressive pattern (denoted as AR(1)), which also implies what’s called trend-stationarity. In other words, if you take away the trend of an AR(1) time series, it will look like a random walk – the detrended time series is randomly distributed around a mean of zero.

I’ve taken both nominal and real quarterly GDP for Malaysia for the sample period of 2001:1-2007:4 to illustrate what I mean. The regressions look like this:

rGDP = constant + @trend + d2 +d3 +d4 +AR(1)

nGDP = constant + @trend + d2 +d3 +d4 +AR(1)

…where @trend represents a linear time trend; d2, d3 and d4 are the seasonal dummies for 2Q, 3Q and 4Q of every year; and AR(1) is the autoregressive term.

Running the regressions yields:





All the explanatory variables are almost all statistically significant at the 95% level (Prob. values below 0.05), diagnostics all check out ok, and we have a very, very close representation of the actual time series (very high R2). It’s interesting to note that the error values (residuals) for the rGDP model are an order of magnitude smaller than that of the nGDP model.

Now that I have a baseline model(s), I can use historical values of GDP to forecast future values of GDP. This doesn’t turn out so well as a pure forecasting exercise because we had the commodity boom in 2008, and fell into recession in late-2008 to 2009 (out of sample forecast: 2008:1 to 2010:1):



Now you can partially see why so many forecasters and policymakers got caught out by this recession, and why so many risk models in the financial sector failed. The drop in output far exceeds the potential band of possible outcomes implied by historical data. That argues for a more structurally based approach to forecasting, though to be fair, that hasn’t fared much better in the present crisis.

But I digress. For today’s post, this failure serves my purpose pretty well – the forecasted level of GDP can be taken to represent potential output, and I now have a probability based measure of the output gap. I can’t obviously make a determination on whether there was a permanent drop in potential output, which defines a Hamilton recession, but what I can do with this is to ask (statistically speaking) whether there has been a change in the path of economic growth, which is almost (but not quite) the equivalent.

Specifically, assuming that productivity growth is about constant, then the trough of the recession represents a structural break and the economy would resume on its former trend from the new start point. Mathematically, I’m asking if there is a change in the intercept, while holding the slope of the regression more or less constant.

If the structural break as well as all the other variables are statistically significant, then I have proven my hypothesis that the Malaysian economy has moved to a permanently lower path of economic growth, and there has been a de facto loss in potential output. Bear in mind that we’re only looking at four quarters worth of data to work with i.e. from the point where the economy started growing again.

Defining d5 as the structural break variable, with values of 1 from 2001:1 to 2009:1, and zero thereafter, and rerunning the regressions with a larger sample range of 2001:1 to 2010:1, here are the results:





The results generated generally support my thesis – the economy has moved to a new path for economic growth, which in turn implies that there has indeed been a permanent loss in potential output.

Now for a more normative approach – it’s hard to reconcile my results with the fact that for Malaysia at least, the genesis for the recession was external. There was nothing fundamentally wrong with the economy pre-crisis, pace NEAC and the NEM. If demand was to fully recover, then we should expect to see a “V” shaped recovery back to trend (just back to the pre-crisis level of output is not sufficient). That hasn’t happened here, obviously.

In that light what I think we’re seeing is, potentially, a permanent loss in external demand and not a permanent loss in the potential output capacity of the economy. That in itself will eventually result in the same thing (unused capacity will be depreciate away), but there’s always hope for a stronger recovery though I’m having trouble right now seeing where that might come from.

So what could happen in time is more like a “U” shaped recovery back to trend – I’m crossing my fingers we don’t see a “W”. But in the context of what I’ve done here in this post, we’ll have to wait and see.

Technical Notes:

First graph and intellectual inspiration from:

Cerra, Valerie & Saxena, Sweta Chaman, "Did Output Recover from the Asian Crisis?", IMF Working Paper No. 03/48


  1. Hisham H,
    Interesting post. Well the other issue is the fact that commodity prices are below the historic highs. Perhaps that too may explain the loss in GDP - not directly as it is perhaps a 2nd order effect (i.e money sloshing around being used to finance construction etc).

    By the way - what is your take on the m-o-m contraction in M2? If not for a very strange RM 13 b injection in Mac, it would have been 2 consecutive months of contraction.

    By the way I also declare to be one of the earlies to point out the big elephant in the room called the Public sector debt. I think you can vouch that I had been on the debt story way b4 Jala & Co. proceeded to issue the "bankruptcy" warning.

    This will be the present to the Malaysian people for voting Dr Mahathir in after he had brutalized DSAI. I think its about time you tell the readers on what the debt picture will look like in 2yrs time. Govt revenues are coming in much lower than projected. My projection is RM 600 b direct by 2012. Additionally the amt guaranteed by the Federal Government is increasing as well. Its now RM 87 billion, up from RM 70 billion at the start of the year

    Yes Malaysia can monetize the debt - but lets see the effect on the ringgit when that were to happen.

    Hyper inflation is the gift of Tun Mahathir to the future Malaysians.

  2. Wenger, I'm not putting much store into the contraction in M3 - call it the "ang pow" effect. People take out cash from the system before CNY, and put it back in afterwards.

    I think I'd withhold judgement over government finances until after this year. Tax revenue would be based on household income and corporate profits last year, which obviously wasn't all that good. The real litmus test will be in 2011.

    And hyperinflation is definitely not on the cards. It takes systematic multi-year abuse of the monetary system to unanchor inflationary expectations to that degree. And only if monetary expansion and quantitative easing is transmitted to the real economy via excess government spending or excess bank lending - and neither is really happening.

    The government's deficits are being pushed up by a shortfall in revenue, not overspending.

  3. Dear Hisham H,
    My prediction of hyper inflation is when the
    Central Bank is forced to buy Government debt.

    That is not happening today, nor in 2011. But it is certainly going to happen as the growth in the deficit is larger than the growth in the money supply. When - I do not know as I have yet to build a reliable M3 model. Worse come to worse, I'd just slap several growth rates and work it out.

    Assume it happens in 2016. At that stage, the RM 200 billion banks hold with BNM is depleted. There are no new takers for MGS and the Banks have no room to maneuver. (One should see the crowding effect play a big part here)
    Yet the Govt is still forced to issue more debt as the problem with the deficit is structural. Govt cannot operate w/o RM 150 billion a year, subsidy or no subsidy. At this stage, the CB steps in and issues funds to monetize the debt. The Ringgit gets hammered in the fx-market. Push thru inflation rises.. Dot dot dot.

    At that stage the ringgit will be punished. Similarly at that stage, if one were to think logically enough, surely all subsidies will come to an immediate end as the Government is in severe debt crisis. Hence a depreciating ringgit will result in transmission of inflation from US dollar linked commodities.

    History does not repeat itself, but it certainly rhymes.

  4. Wenger, hyper-inflation does not happen just because the central bank engages in printing money. You have to take into account the transmission mechanism as well - there's no automatic link from an increase in the money supply (which is what happens with quantitative easing) to inflation, especially under current conditions. Check this post for more details.

    And when I mean systematic abuse, just financing the deficit even as it stands will not be sufficient to trigger higher inflation - you'll need wholesale financing of the entire government budget to unanchor inflationary expectations to that extent.

    Another related point: the Ringgit will only fall under your scenario assuming that other countries are relatively more fiscally responsible. I don't see that in any of our major trade partners except Europe, and they have different problems that will also weigh on the valuation of the Euro.

    Also, I don't know if I'd use M3 as a base for determining potential future demand for government securities. The ratio of bank holdings of MGS/GII doubled in 2007/2008 - but we're talking about an increase of 2% of assets to just 4% of assets. While growth government debt has indeed outpaced M3 growth, excess deposits of FIs at BNM has grown even faster - so it's hard to argue for a direct link between M3 growth and the ability of the financial system to finance government borrowing. You might want to look at the national savings rate instead (rate of increase in EPF contributions, insurance assets, banking assets, bond funds).

  5. this is a very good post. it forces me to re read my econometrics and monetary economics. this technical analysis will scare away those chest beaters especially the peddler of rancid curry.
    looks to me, the public economists in blogospere will be domianted by you, hishamh and e theorist. well done guys.

  6. Dato Sak,

    Coming from you, that's a real honor. Thank you