Investment, consumption and business cycles

The following is taken from a piece in the FT by Samuel Brittan

Railway buffs will be familiar with trains that have a locomotive behind as well as in front of the carriages. But a train with only a rear locomotive? Maybe in some remote part of the Andes, but hardly a normal phenomenon…

…So far the government has focused mainly on stimulating business investment, equivalent to trying to drive from the rear locomotive. There have been bouts of investment-led growth – such as the US and UK 19th-century railway booms. But the normal process is for consumption to lead. The latest project – the funding for lending scheme –seems better designed than most. But it cannot be large enough to lift the whole economy.

Samuel is spot on. The pattern of recession and recovery is historically led by movements in consumption, both on the way down and the way up. Don’t get me wrong, reducing red tape and making our economies more competitive and addressing structural issues like the UK’s pensions funding shortfall are big issues that will need to be addressed over the medium to long-term. But cutting back on government expenditures at the same time that private expenditures are simultaneously decreasing is like trying to run up an escalator which is going down.

Politicians arguing for public sector spending cuts often use phrases such as ‘we are reducing the deficit so as to encourage private sector investment’ . The argument of the austerity brigades (and austerity light brigades such as the UK and US) usually centers around the fact that for recovery, we need strong private investment to pull us out of the slump with private incomes and thus consumption following, and in order to do so it is usually argued that government must ‘get out of the way’. Again my argument is not with the idea of reducing government expenditures. My argument is that we shouldn’t be reducing them now (at least not with further monetary easing that is effective), and we certainly shouldn’t be doing so in anticipation that private investment will pull us out of the slump and back to better times. Take a look at the following two graphs;


Both plot the annual percentage change in real personal consumption expenditures (multiplied by 6 for scaling) in the US, but each plots an alternative measure of private investment (again – annual percentage change).  In each graph you’ll notice that all the major peaks and troughs (bar one or two) in consumption precede those in investment by approximately 1-3 quarters. So for the US to have an investment driven recovery, it would seem that it would be a major break from previous trends. Furthermore, with extremely similar economic structures, I think it would not be unreasonable to extend such an argument to the UK, the other major developed economy currently undergoing ‘voluntary austerity’.

Remember those two graphs next time you hear a politician use the ‘government needs to get out of the way’ spiel as an argument for current economic austerity programmes.


Monetary policy and income distribution

Discussion about skills biased technical change as a possible explanation for widening income inequality seems to be in vogue at the moment, so I thought I’d touch upon the subject. In a nutshell the theory postulates that technological progress raises the marginal productivity of certain segments of the labour force disproportionately more than others. Furthermore since you earn what you produce (in economics terminology, each worker is paid a real wage equal to his or her marginal product assuming efficient capital markets), those workers who are now producing a higher proportion of overall output get a greater share of total pay. This is as basic an overview as possible as I don’t want to get into too much detail, with many others having already given very succint accounts of the theory elsewhere (see for instance Noah Smith and Mark Thoma).

What I did want to discuss is something else in the data that has always intrigued me. First of all, what’s the general picture in terms of the actual structure of income inequality over the last few decades? Well, lets take the US as an example. Fig.1 shows the percentage of total income going to income bands of 1/5th.



As has been much discussed, the share of the overall pie going to individuals in the top 20% of incomes has been increasing steadily since the early 1970s as everybody else’s incomes have generally stagnated or fallen. What is most worrying is the explosion of the share of GDP going to those in the top 5% of incomes. This is again well documented, but what intrigues me is the precise timing of the increase. As we can see from fig.1, the share of income going to these individuals suddenly takes off in 1980, and increases rapidly up to around 1994, while the share of GDP going to the lowest 80% decreases.

So what’s with the sudden divergence of the top 5%in 1980? Yes, deregulation of the financial sector would have contributed to the take off, but financial services had been undergoing gradual deregulation over a period well before the sudden upward tick. But what about macro stabilization policy at the time? Take a look at fig.2 below which plots the annual percentage change in the CPI.



As is clear, inflation was running well above acceptable levels prior to 1980. However, following the turn of the decade and the ensuing Volcker disinflation, inflation was brought down to around 2% within 3-4 years from a peak of almost 15%. With the huge tightening of monetary policy that the Fed instigated in 1980 coinciding almost perfectly with the sudden upward tick in the share of income going to the top 5% of incomes, is it not beyond the realm of possibility that such a significant tightening of policy impacted certain households and workers disproportionately more?

Here is how I could see such a scenario playing out. First of all, if low skilled labour is in greater abundance than high skilled labour (which it is and was) then in the face of lower firm revenues those with the highest skills have the greater bargaining power when it comes to maintaining wages at their current level than lower skilled labour. Hence, rather than losing a highly skilled worker (who is not easily replaced) due wage reductions in the face of a negative aggregate demand shock, it may well be less costly for employers to cut the wages and the employment of lower skilled workers (cut the wages of or fire 10 cleaners rather than fire/reduce the salary of 2 analysts) and maintain the wages going to those already at the higher end of the pay scale.

Indeed, comparing figures 1 and 2 again, we can see that the next time after the 1980 episode when inflation was reduced by a significant margin over a short time frame, those with the top 5% in terms of incomes experienced a similar jump in the share of overall GDP which they received. That is between 1991 and 1993 when inflation was brought down from approximately 6% to just over 2.5%, we can see from fig.1 that the share of income going to the top 5% jumped significantly again.

Of course, this could all be due to a third factor influencing both variables simultaneously. However I find it hard to reconcile with the fact that despite such a significant reduction of inflation over the 2 entirely separate periods highlighted, the share of GDP going to the top 5%  suddenly increased at a rapid rate.

I certainly don’t see technological factors having such a significant short-term impact.

The sticking point

Gavyn Davies has an interesting post up discussing the possible size of the output gap, and hence whether demand management policies can be used to increase output in the short run. To see why the size of the output gap is important for current policy, consider two extremes. If current output is actually well below potential output, then expansionary monetary or fiscal policy would be capable of increasing output back up to potential without significantly increasing inflation. If on the other hand potential output is in fact much lower than the pre-recession trend (see Gavyn’s excellent group of diagrams for various countries) and output is currently equal to potential, expansionary monetary and fiscal policy would just lead to higher inflation with no change in output or employment. These are the two extremes but they show the basics of the problem for anyone unfamiliar with the theory.

Gavyn discusses an interesting issue relating to the breaking down of this relationship when inflation decreases (i.e. disinflation sets in). Take this quote from Gavyn in which he discusses a previous study;

In 2010, Andre Meier at the IMF published an analysis of persistently large output gaps, which I have previously summarised, in which he concluded that global inflation should be dropping each year by about 0.5 per cent, but that this would stop happening as the rate of wage and price inflation approached zero. The estimated impact of the output gap on inflation would therefore be expected to decline as inflation falls. This prediction seems to have proven accurate.

This is part of a larger body of literature on asymmetric price adjustment originating from such work as this by Greg Mankiw. What I would like to do is explain the theory without getting too wonkish, so here goes.

Consider a starting point where the economy is at full employment, inflation is on target and GDP is equal to potential. Then consider some negative aggregate demand shock. Most of us would instantly assume this aggregate demand shock would cause the prices of goods in the economy to fall, as firms respond to the falling demand for their goods by reducing their price in order to retain market share and sell their output.   If goods in the economy are imperfect substitutes i.e. if consumers place higher value on other aspects of products other than price (think of the brand effect of owning an iPod compared to other mp3 players), then when a firm decreases its price below its competitors it gains only a fraction of the additional market share that it would receive if goods were perfect substitutes.

If the additional market share for a given price decrease is sufficiently small, it is optimal for the firm to simply leave its price as it is and maintain a higher revenue per good sold, but sell fewer goods. When we aggregate this over firms, this provides an explanation of why in the face of negative aggregate demand shocks, output may decrease significantly without any appreciable fall in prices. Thus by extension, if such a situation arises, expansionary fiscal and monetary policy can expand output back to potential without an appreciable increase in inflation.

The beauty of the theory is that we still retain the result that expansionary policy when output is equal to potential leads to increases in inflation rather than increases in output i.e fiscal and/or monetary policy cannot lead to permanent increases in output growth. But why? Given that a firm gains very few customers from cutting prices below its customers, it similarly loses few when raising prices above customers. Hence when a positive aggregate demand shock hits the economy from our starting point of full employment, firms have every incentive to increase prices as the demand for their product increases. Hence on aggregate, output responds very little and inflation increases.

If the theory holds, then it is entirely plausible that the weak output growth that many developed economies are now experiencing could be reversed by expansionary monetary (if you think such a thing is currently possible, which I do) or expansionary fiscal policy, without leading to an appreciable increase in inflation.

Plosser misses the point

Charles Plosser, one of the original developers of the Real Business Cycle group of models and Philadelphia Fed President, argues that the Federal Reserve’s recent promise to keep nominal interest rates at (or near)  0 percent until unemployment falls below 6.5 percent or inflation expectations at the 1 -2 year horizon exceed 2.5 percent will impair the process of deleveraging. Here’s what the Fed President had to say;

“Efforts to drive real rates more negative or promises to keep rates low for a long time may have frustrated households’ efforts to rebuild their balance sheets without stimulating aggregate demand or consumption,” said Plosser, who doesn’t vote on monetary policy this year.

This to me seems totally ludicrous and lies at the heart of the problem with the inflation hawk mindset. At the heart of Charles’ argument is that as interest rates decrease, the rate on savings becomes less and so the incentive to save today diminishes (at least this is what I presume he is getting at), whilst the return on whatever savings they do have decreases and thus reduces the amount of money they have available to pay debts off. However this whole thought process seems to neglect the fact that if an individual is net indebted, his or her net interest earnings are in fact negative i.e. they are paying more interest out on their borrowings then they are earning on their savings,  even if the actual interest rate they pay on each is identical. Thus for such debtors, any decrease in real interest rates is exactly what is needed at the margin.

Secondly, we haven’t even talked about one of the most important, yet one of the most neglected, issues in modern macroeconomics – the idea of debt deflation.  This is the notion first proposed by Irving Fisher (and recently rejuvenated by Gauti Eggertsson and Paul Krugman) that dramatic recessions are often propagated by falling prices, combined with indebted households whose debts are fixed in nominal value, so that the real value of those debts in fact increase as prices fall further.  As these households fight against the tide and cut their spending even further to pay off the increased real value of their debts, prices fall further and the process continues, leading to a greater and greater general contraction.

Now I’m not suggesting that we have as severe a problem with debt deflation today (since inflation in the US is approximately 1.8 percent) as Irving Fisher was confronted with when he cane up with the theory in response to the Great Depression. But when households are attempting to reduce the real value of their outstanding debts, a little extra inflation is certainly a good thing in helping them achieve this.

Also, whilst their may not be such a chronic case of deflation as there was some 80 years ago following the Depression, households have certainly experienced a real increase in the value of their outstanding debts. As nominal incomes have moved to a permanently lower (trend) level – see graph below and David Beckworth’s post for a more in-depth discussion – the nominal value of their outstanding debts remained the same and so the real value of their debts increased.

US Nominla GDP1

All in all, I just don’t see how falling rates of interest will lead to increasing household debts as Charles Plosser suggests.

Stabilising Nominal GDP

Following up on my post yesterday, I decided to delve a little deeper into the properties of the NGDP series for the UK, to see what the characteristics were over different periods and whether they support the notion that an NGDP level target is a good nominal anchor.


Ful NGDP growth


earlier growth


latter growth

Figures 1-3 are simply plots of the annual y.o.y. growth rate of NGDP in the UK split over different time frames for ease of viewing. Oddly enough a quick glance at fig.1 may lead us to believe that nominal GDP for the UK has been more variable over the latter years of the sample range.  Nevertheless, we need to be more precise than just making judgements based on cursory glances of the data, so below are variances calculated over different ranges. I decided to calculate the variance over the period leading up to 1940 and separately for the period after, since I am interested in whether stable nominal GDP  coincides with stronger economic performance. As mentioned in my previous post, since the period after1940 was one of the longest phases of stable economic growth without a major collapse, this division seemed a natural choice.

Variance for 1831-1940=32.115              Variance for 1941-2009=26.4103

As you can see from the values, the variance of the growth rate of nominal GDP after 1940 was some 4 percent lower than the period before 1940.

In my opinion this is another piece of evidence suggesting stable nominal GDP growth is a causal factor in determining stronger economic performance.

Macroeconomic Breakthroughs

In keeping with the nominal GDP theme that I’ve had going at the moment, here’s a little data that has a coupe of intriguing artifacts. Both of the figures below are plots of the log of nominal GDP for the UK (so that we can interpret the gradient of the curve as the growth rate of nominal GDP) over different time frames.


UK Nominal gdp GRAPH


UK nominal GDP graph reduced series

Firstly, taking a longer view, let’s have a look at figure 1. As you can see I’ve highlighted the two points that caught my eye. At the point highlighted by the green arrow, corresponding to approximately 1937 , there is a very clear structural break, where the average growth rate seems to permanently increase. The specific timing of the increase in growth is in all likelihood a combination of the fading out of the Depression and mobilization for war. However, as you can see, there isn’t a mere temporary increase in nominal GDP growth. Rather, the increase in the growth rate is something that carries through all the way to the decline of 2008.

This shows just how important the move towards a more active demand management macro policy –  and away from a completely hands off classical approach – has benefited the United Kingdom. Of course demand management policy wasn’t a unilateral success – we obviously made mistakes when it came to the original miss-specification of the Phillips Curve which led to double-digit inflation in the 1970s and 1980s, see below. But all in all I think this picture should illustrate just how important Keynes’ and Friedman’s insights into stabilization policy were (although they obviously approached the problem from different ends of the spectrum).

Secondly, lets take a look at the circled area. For ease, lets look at the shorter time frame in figure 2. The circled area I think is interesting because of what it says about the character of stabilization policy over the past 30 years. Firstly, it’s quite striking how stable the growth in nominal GDP is up to 1970 – the time series is almost a perfect linear trend over this period. However we can see that between 1970 and 1980 the growth in NGDP takes off – a result of the oil price spike but most significantly the higher inflation which resulted from the overly expansionary policies based on the miss-specified Phillips I mentioned above. Specifically, the belief that there was a permanent trade-off between unemployment and inflation, before the rational expectations revolution and the accelerationist hypothesis sprang forth, and was incorporated into policy making, led to higher and higher inflation. However, from 1980 the higher inflation was curbed very successfully and nominal GDP growth once again returned to a stable linear trend.

To sum, this series shows two very distinct advances in macroeconomic management. The first we discussed saw the introduction of a far more active demand management policy, the second saw the realization that there was no long-run trade-off between inflation and unemployment/output, and thus represented a refinement of Keynes’ and Friedman’s original breakthroughs.  Both changes came after sudden periods of poor economic performance.

The question is, what will the next period look like?

Finally, the notion of adopting a nominal GDP level target is receiving more and more traction all the time. With this in mind its important to remember how stable a period in terms of economic growth the last 50 years have been up to 2008. Then look at how stable nominal GDP growth has been over the same period. Minus the set back in the 1970s, the log of nominal GDP has followed an almost perfect linear trend.

Its by no means enough to immediately change how we conduct monetary policy, but such exploratory analysis suggests a stable growth rate for NGDP levels is potentially an important part of establishing a stable macroeconomic environment.

Some more on Mishkin and Woodford

Scott Sumner has an excellent post also on the subject of the WSJ piece by Mishkin and Woodford, and more generally on the inadequacy of interest rates as a monetary policy tool at the zero lower bound. Here’s a quote from Scott’s post;

We need a new policy instrument, one that doesn’t become mute when nominal rates fall to zero.  We could use the monetary base (QE), but I would prefer pegging the price of NGDP [Nominal GDP]  futures contracts.  And then adjusting the monetary base as required to keep the Fed’s net position in the NGDP futures market close to zero.

I largely agree with Scott’s longstanding view that a nominal GDP target would be a better nominal anchor than a pure inflation target (although I’m probably not as wholeheartedly sold to the idea as Scott is), and although I may not go the whole way to suggesting the launch of NGDP futures contracts and having the monetary authority directly target these assets, when you have a monetary authority target either a price level or an NGDP level rather than an annual change in prices, and crucially a commitment to do so is perceived to be credible, then the zero lower bound in my opinion can indeed be overcome by monetary expansion. The crucial element to either the price level or NGDP level target is that both policies possess reversion qualities. Take the price level target – if a monetary authority were following such a target at a 2% annual rate  and undershot their target, then over some period in the future inflation will have to be greater than 2% so that the price level is stabilized on its prior path. This quality becomes especially important in times of a liquidity trap. For instance, take the below figure from a paper written by Gauti Eggertsson and  Michael Woodford.

Post 21

Although rather complex looking, the graphs are actually very illustrative of what the optimal policy entails (according to the solution of the authors’ DSGE model, and most other models on the subject). In these graphs, each line corresponds to the optimal monetary policy when the liquidity trap persists for different periods of time, with the duration of the liquidity trap set exogenously. Take graph (a). The first spike in the grey line corresponds to the optimal course of inflation when the liquidity trap persists for one period. If the liquidity trap persists for 2 periods then the next grey line to the right corresponds to the optimal path of inflation and so on. The same applies to the graphs for the output gap and the price level. Taking a look at graphs (a) and (c) it is clear that the longer the liquidity trap persists for i.e. the longer the economy finds itself mired in a deflationary spiral, the greater the level of inflation that is needed under the optimal policy. Thus we can see from graph (c) that the longer the liquidity trap persists, the higher the price level at which the economy is eventually stabilized. This is exactly what is meant by price-level reversion, and is exactly what is required to escape a situation characterized by insufficient aggregate demand and interest rates up against the zero bound.


Update – David Beckworth has a nice graph which illustrates the reversion characteristics of an NGDP level or price level target.

Fed Forward Guidance

Mark Thoma points us towards a piece in the WSJ by Michael Woodford and Frederic Mishkin regarding the Federal Reserve’s recent change in communication policy and the significance that the alteration brings. For anyone unaware of the way that the FOMC’s communications strategy has been altered, the Fed recently committed to maintain interest rates at or near 0% so long as unemployment remains above 6.5% and inflation expectations at the 1-2 year horizon are no higher than 2.5%, whereas previous policy announcements had merely stated that federal funds rates of near zero were “likely to be warranted at least through mid-2015.” I think there is often confusion about why “forward guidance” about the future course of nominal policy rates should matter (in particular at the zero lower bound) along with what kind of forward guidance is consistent with optimal monetary policy, so I aim to address both areas here.

As Mishkin and Woodford state;

“A commitment not to raise rates in the future as soon as might have been expected is an obvious way the FOMC can loosen current financial conditions.”

But why should expectations of future interest rates matter for the current stance of monetary policy? Take a look at the below equation taken from a paper by Gauti Eggertsson. If you’re not comfortable with maths don’t be daunted by all the subscripts and variables . The Y_t essentially denotes the growth rate of output in the current period ‘t’, the E_t term denotes current expectations so that the first term on the right denotes current expectations of output growth in some future period ‘T’, but what is of critical importance to our discussion here is what is inside the first circular brackets. The term ‘i_s – π_(s+1)’ is the (expectation) of the real interest rate in some future period ‘s’, with the i’s denoting the nominal policy interest rate and the pi’s representing inflation. Lastly the r^e terms (the last terms in the first bracket) are the `natural rate’ or `efficient’ real rate of interest, which is the real rate of interest required to utilize all available capacity and thus achieve maximum output without leading to higher inflation.

equation 1

We can see that if the Fed were able to alter the nominal policy rate each period so that the real rate of interest were equal to the efficient rate of interest, the term inside the first bracket would be zero (the last circular bracket is related to government spending which we can ignore for the purposes of our discussion here) and output growth would be maximized.

However, the position many central banks currently find themselves in to varying degrees cannot be remedied by simple alteration of the nominal policy rate. Take the Federal Reserve, which has its current policy rate effectively at the zero lower bound and is thus unable to lower it any further, and has fairly low inflation expectations . In our equation above, this is represented by the i’s being set to zero at shorter horizons. Yet from our model, we would expect the efficient rate of interest (from a modelling perspective) to be negative during periods of severe economic contraction. If this efficient rate is less than the real rate of interest the term inside the brackets is positive and we have a sub-optimal outcome. Unfortunately current nominal rates cannot be lowered to compensate this, so we end up in a situation where monetary policy is too tight but we are unable to loosen current monetary policy further. But we can see from the equation that current output growth depends on the entire future path of nominal interest rates, so one way we can loosen current policy is to pledge to keep nominal policy rates at zero for longer than previously expected. If everything else in our equation above remains constant (a simplifying assumption for our discussion) then current output growth must naturally increase.

So what is the difference between good and bad forward guidance? Well there could be many different forms of bad forward guidance, but I want to focus on just one. Let’s take the forward guidance policy that generally characterized that of the Fed before last December when the change was made. Generally, the statements said that the Fed was extending its forecast for how long nominal policy rates should remain at or near zero. Take the statement above that economic conditions meant near zero rates are “likely to be warranted at least through mid-2015.” The problem with these statements is that they don’t tell us whether this means the monetary authority has decided the loosen its own stance, or whether it simply expects economic conditions to be worse than previously expected.

Take our above equation. If markets took such commitments as representing a loosening of policy, the term inside the brackets decreases and current output increases – the desired effect. However if these comments are interpreted as meaning the monetary authority expects economic conditions in the future to be worse than previously expected then either the expected future output term in our equation decreases, or the expectations of future inflation (the π_s) decrease, or both decline. Either way, the extension of near zero interest rates is simply responding to worsening economic conditions and does not constitute a relative loosening of policy. At best, the policy achieves no improvement, at worst it signals to market participants that the monetary authority is losing control of its stabilization mandate by repeatedly undershooting its inflation and/or output stabilization targets.

The new forward guidance, coined by some as the “Evans rule” after Chicago Fed president Charles Evans, whist not perfect nevertheless avoids such confusion by illustrating the criteria the Fed uses to decide on the path of nominal rates, and enables market participants to directly observe how loose policy is and how loose it is likely to remain. Most importantly, it clearly illustrates that the Fed will loosen policy until one of two objectives is achieved – inflation expectations increase to 2.5% or unemployment falls to 6.5%.