The hypothesis that the Monetary Policy Committee’s “reaction function” changed in 2010 is supported by the “MPC-ometer” forecasting model, which suggests that interest rates would have been raised several times last year and in early 2011 if the Committee had responded to incoming data in the same way as over 1997-2009. The model predicts that the MPC will shift to an easing bias this month because of recent financial market turbulence and more favourable inflation indicators.

Regular readers of these notes may remember previous discussion of an “MPC-ometer” model designed to forecast monthly interest rate decisions based on the latest economic indicators and financial market developments. This model performed well after its introduction in September 2006, correctly signalling the month and direction of 12 out of 13 Bank rate movements – two more than the mean forecast in the monthly Reuters poll of economists.

The MPC-ometer attempts to predict the average interest rate vote of the Committee’s members. For example, if five members recommend a quarter-point Bank rate increase while four prefer no change, the average interest rate vote is +14 basis points (i.e. five-ninths of +25 bp). If it is assumed that votes are either for no change or a move of 25 bp – reasonable under “normal” economic and financial conditions – then the model forecasts an actual rate change when the average prediction is greater than +12.5 or less than -12.5 bp.

The MPC-ometer’s 12 inputs were selected on the basis of statistical analysis and can be grouped into indicators of economic activity, inflation and financial market conditions. The inflation sub-set is largest, comprising the latest headline annual increases in consumer prices and average earnings as well as several measures of expectations. Activity indicators include GDP growth and business / consumer confidence while credit spreads and movements in share prices and the exchange rate are used to gauge financial conditions. Importantly, the model also includes the prior month’s average interest rate vote to capture any revealed bias.

The model was designed to predict the immediate interest rate decision but can also be used to forecast further ahead, based on assumptions about the input variables. In the latter mode, the model’s prediction for the average interest rate vote in a particular month feeds into the forecast for the subsequent month.

This latter approach was used to examine whether the MPC’s decisions since the start of 2010 have been consistent with its “reaction function” between 1997, when the current inflation-targeting regime was instituted, and 2009. Specifically, the MPC-ometer’s coefficients were estimated using data for 1997-2009 and the resulting model used to forecast the monthly interest rate vote based on actual economic and financial data and the previous month’s vote prediction.

This exercise suggests that the “old” MPC would have responded to a rising inflation overshoot and evidence of economic recovery by raising interest rates in four quarter-point stages over the last 18 months, implying a current level of Bank rate of 1.5%. The predicted increases occur in March, May and August 2010 and April 2011. The output of the model is shown in the first chart below – note the four peaks above the “unchanged policy range” since the start of 2010.

The suggestion that the monetary policy “goalposts” were shifted in early 2010 is not new and fits with MPC communications around that time. In particular, an important speech by the Bank’s Governor Mervyn King in January 2010 signalled that the Committee was willing to tolerate a substantial inflation overshoot caused by “temporary price level factors”, which Mr King defined broadly to include the exchange rate – traditionally regarded as a key element of the “transmission mechanism” of monetary policy. Some commentators have speculated that the Governor was preparing the ground for an informal pact with an incoming government, under which the Bank would agree to dilute inflation targeting and maintain super-low interest rates in return for a commitment to sustained fiscal contraction.

Does the divergence of actual from historical behaviour imply that the MPC-ometer should be cast on the statistical scrap heap? The answer is no, for two reasons. First, the model’s coefficients are reestimated every month so adjust – albeit slowly and incompletely – to changing policy priorities. Secondly, its inclusion of the prior month’s actual interest rate vote serves to adjust the current month forecast for any unusual behaviour.

The continued relevance of the MPC-ometer for current month forecasting is illustrated by the second chart, showing the fitted values of a model estimated on the full sample of data (i.e. extending up to July 2011) and incorporating the actual prior month vote. This version has continued to perform respectably over the last 18 months, with only one false tightening signal, in March / April 2011.

The MPC-ometer forecasts that the Committee will shift to an easing bias at this month’s meeting. In June, the predicted average interest rate vote was +7 basis points, consistent with two or three members favouring a quarter-point increase. The result that month was +3 bp, with hike votes by members Weale and Dale offset by Posen’s call for a £50 billion extension of QE (treated as equivalent to a quarter-point cut). The model shifted into neutral in July but the actual vote was unchanged at +3 bp. For August, the predicted average vote is -7 bp – the most negative reading since -13 bp in November 2009, when the MPC last increased QE (by £25 billion). This would be consistent with Weale / Dale retracting their tightening votes and one or two members of the middle grouping moving into Posen’s camp. In reality, any shift is likely to be smaller, partly reflecting the difficulty of changing entrenched positions.

 
The significant swing in the model’s prediction between June and August has been driven by a combination of financial market turbulence – reflected in falling share prices and gilt yields – and declines in consumer confidence, actual inflation and CBI industrial firms’ price-raising plans.

It should be remembered that the MPC-ometer is designed to predict actual decisions rather than indicate the policy stance consistent with the inflation-targeting remit. The view of these notes remains that inflation is unlikely to return sustainably to the 2% target at current interest rates, partly because negative real rates are exerting strong upward pressure on monetary velocity. This view receives indirect support from the suggestion that Bank rate would now be at 1.5% had the MPC had reacted to events in 2010-11 in the same way as over 1997-2009, a period in which CPI inflation averaged 1.8%.