I have spend my entire career as an economist doing forecasting – both of macroeconomic numbers and of financial markets. First as a government economist and then later as a financial sector economist. I think I have done quite well, but I also know that I only rarely am able to beat the market “consensus”. If I beat the market 51% of the time then I think I am worth my money. This probably is a surprise to most none-economists, but it is common knowledge to economists that we really can’t beat the markets consistently.

My point is that the “average” forecast of the market often is a better forecast than the forecast of the individual forecaster. Furthermore, I know of no macroeconomic forecaster who has consistently over long periods been better than the “consensus” expectation. If my readers know of any such super forecaster I will be happy to know about them.

I truly believe in the wisdom of the crowd as manifested in free markets. So-called behavioural economists have another view than I have. They think that the “average” is often wrong and that different biases distort market pricing. I agree that the market is far from perfect. In fact market participants are often wrong, but they are not systematically wrong and markets tend to be unbiased. The profit motive after all is the best incentive to ensure objectivity.

Unlike the market where the profit motive rules central banks and governments are not guided by an objective profit motive but rather than by political motives – that might or might not be noble and objective.

It is well known among academic economists and market participants that the forecasts of government institutions are biased. For example Karl Brunner and Allan Meltzer have demonstrated that the IMF consistently are biased in a too optimistic direction in their forecasts.

I remember once talking to a top central banker in a Central and Eastern European central bank about forecasting. He complained to me that he frankly was tired of the research department in the central bank in which he was in the top management. The reason for his dissatisfaction was that the research department in his view was too optimistic that the central bank would be able to fulfil its inflation target in the near term. He on the other hand had the view that monetary policy needed to be tightened so the research department’s forecast was “inconvenient” for him. Said in another way he was basically unhappy that the research department was not biased enough.

Luckily that particular central bank has maintained a relatively objective and unbiased research department, but the example illustrates that central bank forecasts in no are guaranteed to be unbiased. In fact some banks are open about the fact that their forecasts are biased. Hence, today some central bank assumes in their “forecast” that their target (normally an inflation target) is reached within a given period typically in 2-3 years.

When central banks publish forecasts in which they assume the reach their targets within a given timeframe they at the same time have to say how the will be able to reach this target. This has lead some central banks to publish what is called the “interest rate path” – meaning how interest rates should be expected to be changed in the forecasting period to ensure that particular target. This is problematic in many ways. One is that it normally the research department in the central bank making the forecasts, while it is the management in the central bank (for example the FOMC in the Federal Reserve or the MPC in the Bank of England) that makes the decisions on monetary policy. Furthermore, we all know that monetary policy is exactly not about interest rates. Interest rates do not tell us much about whether monetary policy is tight or loose. Any Market Monetarists will tell you that.

Instead of relying on in-house forecasts central banks could consult the market about the outlook for the economy and markets. Scott Sumner has for example argued that monetary policy should be conducted by targeting NGDP futures. I think that is an excellent idea. However, first of all it could be hard to set-up a genuine NGDP futures markets. Second, the experience with inflation linked bonds shows that the prices on these bonds often are distorted by for example lack of liquidity in the particular markets.

I believe that these problems can be solved and I think Scott’s suggestion ideally is the right one. However, there is a more simple solution, which in principle is the same thing, but which would be much less costly and complicated to operate. My suggestion is the central bank simply set-up a prediction market for key macroeconomic variables – including of the variables that the central bank targets (or could target) such as NGDP level and growth, inflation, the price level.

So how do prediction markets work? Prediction markets are basically betting on the outcome of different events – for example presidential elections in the US or macroeconomic data.

Lets say the Federal Reserve organised a prediction market for the nominal GDP level (NGDP). It would organise “bet” on the level of NGDP for every for example for the next decade. Then market participants buy and sell the NGDP “future” for any given year and then the market pricing would tell the Fed what was the market expectation for NGDP at any given time. If market pricing of NGDP was lower than the targeted level of NGDP then monetary policy is too tight and need to be ease and if market expectation for NGDP above the targeted level then monetary policy is too loose. It really pretty simple, but I am convinced it would work.

The experience with prediction markets is quite good and prediction markets have been used to forecast everything from the outcome of elections to how much a movie will bring in at the box office. A clear advantage with prediction markets is that they are quite easy to set-up and run. Furthermore, it has been shown that even relatively small size bets give good and reliable predictions. This mean that if a central bank set up a prediction market then the average citizen in the country could easily participate in the “monetary policy market”.

I hence believe that prediction markets could be a very useful tool for central banks – both as a forecasting tool but also as a communication tool. A truly credible central bank would have no problem relying on market forecasts rather than on internal forecast.

I of course understand that central banks for all kind of reasons would be very reluctant to base monetary policy on market predictions, but imagine that the Federal Reserve had had a prediction market for NGDP (or inflation for that matter) in 2007-8. Then there is no doubt that it would have had a real-time indication of how much monetary conditions had tightened and that likely would caused the Fed into action much earlier than was actually the case. A problem with traditional macroeconomic forecasts is that they take time to do and hence are not available to policy makers before sometime has gone by.

This might all seem a little bit too farfetched but central banks already to some extent rely on market forecasts. Hence, it is normal that central banks do survey of professional forecasters and most central banks use for example futures prices to predict oil prices when they do their inflation forecasts. Using prediction markets would just take this praxis to a new level.

So I challenge central banks that want to strengthen their credibility to introduce prediction markets on key macroeconomic variables including the variables they target and to communicate clearly about the implications for monetary policy of the forecasts from these predictions markets.

—-

See my earlier comment on prediction markets and monetary policy here.

Update: If you are interested in predictions markets you should have a look at Robin Hanson’s blog Overcoming Bias and Chris Masse’s blog Midas Oracle.