My sense is that non-linearity, partial irreversibility and asymmetric reactions are prevalent and thus shape the very nature of these uncertainties. Let me take a few examples. For the convenience of analysis, I will use Professor Issing’s typology, but, in fact, all three sources will appear deeply inter-linked when focusing on their nature.

Let’s assume first that policy makers have an overly conservative assessment of the potential growth rate. They will then underestimate the output gap and keep real interest rates higher than they should be. Two consequences might arise. First, hysteresis in unemployment, as a result of lower human capital. Second, sub-optimal investment, which, in turn, would translate into slower potential growth. Pushing the analysis to its limits, the underestimation of the output gap might remain concealed, as, over time, the actual (and not observable) output gap would shrink as a result of higher structural unemployment and slower potential growth. In other words, some structural parameters of the economy would have been modified, which is another nice example of quantum mechanics-type interference. Ironically, inflation could be a by-product of this judgement mistake if, for instance, a positive demand side shock occurs.

Now, let’s turn to the apparently symmetric case. If policy makers overestimate potential growth and thus the output gap, as many observers think the Federal Reserve did in the late 1990s, real interest rates will be lower than they should be, all the more so if myopic financial markets make the same error. As a consequence, companies will over-accumulate capital and asset prices will overshoot their fundamental value. A one percentage point of error on the productivity trend might easily generate a 10% to 20% error on the value of stocks. In our world of floating exchange rates, inflation might even remain quite subdued, as a result of capital inflows pushing the currency higher. We all know the end game: at some point, the ex-post return on invested capital does not match expectations and both the real and the asset price bubbles burst violently. Asset price deflation and, possibly, goods and services deflation might be one of the outcomes of this failure to measure the actual state of the economy.

Back to my initial point, the nature of the uncertainty on the state of the economy is by no means symmetric, when the policy feedback is embedded in the analysis. In both cases, the result is a loss in the welfare function, but in the first case, the loss comes from a lower income level and, in the second case, from higher income volatility.

The worst case I can imagine for policy makers is a situation in which the actual unemployment rate would decline at a speed not very different from the unobservable NAIRU/NAWRU.

Here, the non-linearity would affect the short-term (augmented) Phillips curve itself and, by way of consequence, the transmission mechanism of monetary policy decisions based on Phillips curve models. I suspect that, in Europe, we might be in such situation, as a result of the secular increase of flexible job contracts such as temporary positions, fixed-term contracts and part-time jobs, which increases the weight of outsiders in wage bargaining.

However, the reality might be even more non-linear than assumed in this dynamic description. I even wonder if an aggregate NAWRU makes sense in a currency area where the workforce is not mobile. One can easily simulate an asymmetric shock where the aggregate wage rate might accelerate even if the aggregate unemployment rate goes up. This would happen if unemployment went down in one country, where the slope of the Phillips curve is steeper than in all the other countries, where the rise of unemployment would slow down wage rates, but not fast enough to offset the rise of wages in the overheating country. Maybe Professor Issing had such example in mind when he alluded to possible “non-linear aggregated relationships”.

Although economists did not see it coming, consumers quickly realised that they had been massively robbed. But instead of admitting the truth when it started to emerge, monetary authorities preferred to downplay the shock on prices and, instead, insisted on the subjectivity of consumers’ judgement. Wouldn’t have it been better for the sake of price stability and the long-term credibility of the ECB vis-à-vis the wider public to stand up as guardian of price stability and denounce this historical robbery? I wonder whether there was not some sort of conflict of interest in this matter.

I do not share at all the popular view that financial markets did not understand the ECB’s communication. Evidence of the contrary was given by the ex-post analysis of future markets, as, for instance, was done in the ECB Working Paper quoted by Professor Issing [2]. However, there is still scope for improvement, I believe.

The first area of improvement I see is the clarification of the underlying models used by the ECB. Professor Issing gives a very clear economic definition of the two-pillar strategy, as two sets of models, one centred on money, the other on supply demand conditions in labour and product markets. My own experience, taken from the relationship with our global customer base, is that, while the second pillar is understood, the first one remains mysterious and is often seen as a concession to the Bundesbank’s tradition having no practical consequences. This is quite unfortunate, I believe. The global economy is currently struggling to recover from the burst of the biggest asset price bubble since the 1920s, and to avoid being trapped in Fisher-type debt deflation. This should be a reminder of how important it is to watch the indebtedness of private agents for a central bank that has a symmetric loss function. If, as I think it should, the ECB sticks to its two-pillar strategy, then the group of models “that accords an important role to money” should be made much more explicit than it is currently. For what it is worth, I find it easier to explain to clients what could worry the ECB in excessive money supply by focusing on outstanding debt, rather than on M3. I confess that, in this matter, I might be biased here by the reference to the “total domestic indebtedness” the Bank of France used before EMU.

The second one is the “bias assessment” about the nature of short-term uncertainties. Markets took very positively the careful but precise assessment on risks delivered by the ECB Council back in November. As a result, volatility declined substantially. Communicating about a bias was one of the positive innovations made by the US Federal Reserve earlier on. That the other big central bank adopts this strategy is good news for financial stability, inasmuch as it contributes to stabilising expectations. Hopefully, the ECB will prove time consistent and, from now on, will continue to assess its own bias. Although the meaning of “tightening” or “easing” biases is clear, there is a semantic debate about a “neutral” bias, versus no bias at all. Byzantine as it might look, this debate is important: a time consistent communication policy should always make explicit the bias. Think of it as a qualitative variable taking only three values, –1, 0, +1. As the Indians and the Arabs taught us, zero is more than nothing. Therefore, even a neutral bias is a piece of information.

I had in mind many other practical issues related to the uncertainties created by the unsatisfactory state of European statistics, such as flaws in the measurement of unit labour costs because of the structural underestimation of productivity growth, or inadequate volume-price split in CPIs. But coming from a statistical institute, I do not want to appear too severe with my former colleagues, who, I know, are doing their best. Therefore, I will end here my remarks after a short attempt to move on to a more theoretical ground.

Within this framework, I sense that Professor Issing dedicated his thoughtful analysis to the first function, but left aside the second one. Although I must confess that I have not done the maths, my intuition tells me that the Bank should err on the conservative side when risks on price stability are high, but, on the other hand, should act somehow as an insurer against the risk of recession when risks on price stability are low, because of its low-risk aversion to output volatility, which is low in comparison with private economic agents. Maybe this is a topic of choice for another Economic Policy Forum?

E. C.