What Are Government Promises Worth - Gold, The SNB, And Sovereign Debt (Part 2)

The Swiss National Bank (SNB) decision to dismantle its financially sustainable currency target against the euro, raises questions about the values of other government promises. I explain herein why I believe that inflation target commitments are fairly credible, while nominal GDP level targets less so. This is the second part of a three-part article; the third part will discuss why commitments to repay debt are strongly credible. (The first part discussed currency pegs, and is found here.)

Inflation And Nominal GDP Targets

By the early 1990s, "inflation targeting" had become the state of the art thinking in central bank circles, replacing the previous failed experiment in Monetarism.

After the crisis, there have been calls for a new target for central bank policy - Nominal Gross Domestic Product level targeting ("NGDPLT"). Amongst the two standard options, I think inflation targeting is more credible, for reasons to be discussed herein.

Complicating the discussion is that the two policies are different on two axes. We have a choice between two possibilities for each axis:

1. We have two variables to target: inflation, or nominal GDP.
2. We can target the rate of change (e.g., the annual inflation rate), or a fixed level that grows over time.

This generates four possible policy target types:

1. The rate of change of the price index - (inflation targeting).
2. The level of the price index (not discussed here).
3. The rate of change of nominal GDP (not discusses here).
4. The level of nominal GDP (NGDPLT).

Basic Targeting Framework

In all of these approaches, the promise of the government is much weaker than the promise behind a currency peg. The government only promises to adjust monetary and fiscal policy so that the target variable will act in a way consistent with the policy target (the rate of change remains near a target, or the level attempts to track a target level).

The government does not promise, however, that the target will be hit exactly. This makes the promise plausible. For example, the U.S. government can do very little about the world traded price of oil, and so it has very little control over near-term realised CPI inflation.

For the purposes of my argument here, the distinction between nominal GDP and CPI is not that important. What matters for my discussion here is how the two regimes handle errors.

• In inflation targeting, the central bank only worries whether the rate of inflation 1-2 years in the future remains within its target range. It has to justify previous misses to its political masters, but nothing is done about them.
• In NGDP level targeting, previous misses affect the return to the target.

The second case is less obvious, so I will explain using charts with recent data.

The Canadian Inflation Targeting Experience

The Canadian experience with inflation targeting is often pointed to as the most obvious success of the policy. Canada announced a policy of keeping the annual rate of inflation between 1% and 3% in 1991, and the policy stance has been periodically renewed. (In the chart above, the shaded area starts in 1992.) Realised inflation has stayed mainly within the band, other than small jumps outside of the band, typically as a result of changing energy prices.

The American inflation experience was similar to Canada's over that interval, even though there was no switch towards an inflation target. The "dual mandate" which also includes labour market conditions, remained the formal policy framework, but the Fed changed its emphasis towards inflation stabilisation. Since the United States policy regime is more vaguely defined, I will discuss the Canadian experience. (Although the similarity of the outcomes does call into question the importance of central bank communication for guiding the economy.)

It is beyond the scope of this article to say whether inflation targeting is inherently good or bad. I will assert that it politically popular; the only common complaints about inflation revolve around prices at the gasoline pump.

However, the policy is credible, and that makes it hard to dislodge it for another framework for central bank policy. The credibility comes from the flexibility of the targeting framework.

Errors tolerated. Firstly, the promise does allow for CPI inflation to leave the target range. This is unlike a currency or gold peg, where the price of the target asset is supposed to remain within a small trading band. This is realistic, as central banks can do little to deflect price shocks (notably oil price spikes).

Room to manoeuvre. The average rate of inflation can deviate from the centre of the target band. This allows for more policy flexibility than mathematical treatments in the DSGE literature imply.

For example, we could imagine that United States policymakers could choose between two broad policy stances:

1. Remain in "secular stagnation", and have "steady state" nominal GDP grow at a rate of 4%, and inflation of about 1.5%. 
2. They could loosen fiscal policy, raising the steady state nominal GDP growth rate to 5-6%, with inflation of about 2.5%.

Both policy options would be compatible with the inflation targeting framework used by the Bank of Canada. Incidentally, this policy flexibility means that the belief that fiscal policy is ineffectual (because it will be offset by an inflation-targeting central bank) is incorrect. The models used to derive that result do not take into account uncertainty about policy effectiveness, and the target flexibility. Moreover, the assumption that money is neutral is dubious.

No suicidal reset after large disturbances.  Gold pegs typically ran into trouble after wars. During wartime. gold pegs were typically suspended, and the price level rose. Afterwards, policymakers crushed the working classes with deflationary policies to force wages back to their pre-war levels. A promise that a similar policy would be attempted in the future is not credible, as the working classes now have the right to vote. In an inflation target, there is no such promise; if the inflation target is breached (in either direction) for a period of time, nothing is done rectify the policy miss. This is very distinct from NGDP level targeting, as discussed below.

This flexibility comes at a cost - the inflation target promise tells citizens much less. If you refer back to the chart of Canadian inflation, it is reasonable to expect that the future average inflation rate will be much lower than it was in 1980. But this does not mean that we can factor in an exact 2% annual rise in the CPI in our retirement/actuarial plans. (And as the statistical agencies repeatedly note, the CPI is not a "cost of living" index.)

Nominal GDP Level Targeting - Less Credible

I do not want to dwell here on the issue of whether targeting nominal GDP or a price index is better; what matters for credibility is the level target component of the policy. Yes, it would be nice to stabilise nominal GDP, but such a policy offers little guidance to entities within the economy. Most prices are administered, which is not a feature of the models used to justify nominal GDP targeting. Firms have some control of the rate of change of prices and wages, for which an inflation target provides useful guidance, but firms have no control over the volume of sales.

The theoretical justification for path level targeting is the belief that policy action will need to be stronger after a shock, in order to return the target variable to its target path. This allegedly would eliminate "secular stagnation" even in the presence of the dreaded zero bound on interest rates. As someone in the post-Keynesian camp, I think neither side of that debate makes any sense.

If we look at real world data, the problem with Nominal GDP level targeting is fairly apparent. The chart below shows how the level target would have evolved if a 5% NGDPLT had been put into place in 2000.

The top panel shows the level of nominal GDP, and a target level, which starts out at the actual level of GDP in 2000, and then grows at 5% per year. Until the Great Recession hit, nominal GDP tracked it closely, but then a considerable gap opened up. The bottom panel shows the deviation from target, as a percentage of GDP. The gap has widened to about 15% of GDP, and it is increasing because the steady state growth rate in this cycle is below 5%.

Proponents of NGDPLT targeting argue that if the United States had switched over to NGDPLT before the crisis, the United States would not have suffered from such a growth shortfall. As the tracking error grew, they would have been more desperate to hit target. Expectations would have stabilised around a higher path of inflation and nominal GDP growth.

I am unconvinced by this argument; investors and policymakers believed for most of the post-crisis period that policy was "unsustainably loose", and inflation and growth would take off any minute. Policymakers were throwing everything at the wall, but nothing stuck. I doubt that the outcome would have been any different within a NGDPLT regime.

But even if the regime shift raised growth rates, it seems unlikely that politicians - and voters - would have accepted an inflation surge that allowed nominal GDP growth to hit 8% or higher. But any growth rate less than that would imply an extremely long time to erase the tracking error, particularly if the economy fell into recession once again.

And going the other way, imagine that there is some episode that approximates the economic effect of a war, with sustained high inflation and nominal GDP growth for a period of time. How are policymakers supposed to work off a situation where actual nominal GDP is 15% above target? The country would be forced to duplicate what happened during the return to pre-war gold parities. The probability that would be acceptable to voters is nearly nil.

In the end, the level target would be revised closer to where actual GDP is, and the system would be reset. In other words, you end up with a system that is similar to targeting of the growth rate of nominal GDP, but it is harder to explain, and the parameters of the regime will be changed at arbitrary times.

Control systems engineers have decades of experience with systems that approximate this behaviour. A central bank reaction function is a direct analogy of the control laws built into control systems.

The gap between nominal GDP and the target level is the integral of the tracking error in growth rates. Many control systems feed the integral of a system's tracking error back into the control action. If errors are sustained for a period of time as a result of the system moving into a regime where it reacts slower than usual  to inputs, the integral of the tracking error overshoots. Engineers describe this tendency to overshoot as "windup". If the controller winds up too much, it may actually destabilise the system. As a result, "anti-windup" corrections need to be built into the control law, which shuts off the integration of the tracking error when the controller has left its normal linear mode of operation.  Shutting off the integration of the error corresponds almost exactly to the political process I describe above. (Strangely enough, my last academic article was on anti-windup design, see Romanchuk, B. G. (1999), Some comments on anti-windup synthesis using LMIs. Int. J. Robust Nonlinear Control, 9: 717–734.)

The need to override the nominal GDP level target if the tracking error gets too large makes a NGDPLT policy regime less credible when compared to inflation targets.

Up Next...

In the third part of this series, I will discuss why I believe that promises to repay government borrowings are inherently credible.

(c) Brian Romanchuk 2015