Over the last ten days, statistical institutes have published their early estimates of GDP growth for the third quarter of 2009. Both the US and the Eurozone have surprised the market with very positive readings of 0.9% and 0.4% growth, respectively. The UK, on the other hand, still shows negative growth with a reading of -0.4%.
If these numbers were to be confirmed in revised data, this would imply that the UK and Spain are the only advanced economies that have not yet returned to positive growth. Moreover, the lag with respect to the Eurozone would be very much at odds with historical experience. Figure 1 shows that GDP growth in the Eurozone and the UK have been highly synchronised over the last 25 years, with the exception of the early 1990s recession, which, in the Eurozone, was delayed by the German unification. Over this period, there was no episode in which the UK lagged behind the Eurozone.
Figure 1. Quarterly GDP growth, EU vs. UK
So how accurate has the GDP flash estimate for the UK been historically? Is there other information out there that we can use to form a view on “GDP now”, i.e., in the third and fourth quarter of the current year?
A timely source of information is survey data, which are especially valuable as they are not subject to the long publication lags of statistical series such as GDP. Surveys are typically used to form a view on current economic activity in real time, before final estimates of GDP are published.
Figure 2 shows the economic sentiment indicator for the Eurozone and the UK since 1995. The data are provided by the European Commission (Economic and Financial Affairs DG) and computed by averaging seasonally adjusted balances of opinion (i.e., the indicator is defined as the difference between the percentages of respondents giving positive and negative replies, in each of five sectors – industry, construction, retail trade, services, consumers). The two series are highly correlated, both historically and in the most recent sample, reflecting the correlation already noticed for GDP. The survey data suggest that both the UK and Eurozone turned positive in the second quarter.
Figure 2. Economic sentiment indicators
Should we believe in surveys or in the recent GDP flash release? Although surveys provide timely information, they reflect expectations and are therefore not “hard” data in the way that GDP numbers are. Their value is mostly in their timeliness, but they can be volatile and occasionally give wrong signals. One way to assess how much weight to attribute to surveys is to include them in a statistical model for a “now-cast” of quarterly GDP designed to bridge between the flow of timely information and the quarterly releases of GDP.
These models are now used in many institutions, and in the last few years there have been many new developments in the design of such models. The European Central Bank has been particularly active in model development in this area (see Angelini et al., 2008 and ECB Monthly Bulletin, 2008). Key ideas are described in Giannone, Reichlin, and Small (2008).
Evaluating the UK GDP numbers
Here we use a simple version of this class of models (see Giannone, Reichlin, and Simonelli, 2009), including only surveys and GDP, to evaluate the recent UK release. Our model, using information available up to October 2009, predicts third-quarter UK GDP growth to be +0.15%. We predict that the UK, contrary to the recent flash GDP estimate, is already in positive territory.
To check the accuracy of the model, we with data available the month after the closing of the quarter and forecast the GDP growth for that quarter that would be published after three years (i.e., the final revision). We start at April 1993 to forecast first quarter 1993 and we run the same exercise for all subsequent quarters, always on the basis of only the information available in real time up to one month after the closing of the quarter. Since the flash estimate for GDP is also published by the statistical institute one month after the end of the quarter, this allows us to compare the forecasts based on our model with those produced by the statistical institute. Figure 3 reports the revision errors for our model (ESI-M1) and for the flash (ONS-M1). Although occasionally the errors of the model and the flash are very different, on average they perform similarly, the mean squared errors being, respectively 0.09 and 0.08.
Figure 3. Comparing our model’s GDP growth “now-cast” with the final revision three years later
Let us now focus on the third quarter 2009. Figure 4 below reports how our forecast for that quarter (ESI) has evolved in recent months by estimating the quarterly GDP growth rate using different vintages of data as they became available each month from April up to November 2009. The dotted lines indicate 68% confidence intervals based on historical performance of the model evaluated out of sample, as described above. Notice that, by July, our model identifies signals of a recovery. The red diamond indicates the flash estimate for the third quarter, which is clearly outside the confidence region.
Figure 4. Forecasts of quarterly GDP growth in the third quarter of 2009
Our conclusion is that third-quarter GDP flash estimates for the UK should be used cautiously, as the UK likely has returned to positive growth, like most other advanced economies.
Unlike some commentators (Buiter 2009), we think that an accurate estimate of current economic conditions is important to form a view of longer-term developments, but we should be aware that the numbers produced by statistical agencies are subject to estimation errors and that models that include timely and higher frequency information are valuable tools to form an informed opinion on “where are we now”.
Buiter, Willem (2009) “Another quarter of negative GDP growth in the UK: situation hopeless but not serious”, Financial Times Maverecon, 24 October.
Elena Angelini & Gonzalo Camba-Méndez & Domenico Giannone & Gerhard Rünstler & Lucrezia Reichlin (2008), "Short-term forecasts of Eurozone GDP growth," Working Paper Series 949, European Central Bank.
ECB (2008), “Short-term forecasting in the Eurozone,” Monthly Bulletin, April 2008, pp. 69–74.
Giannone, Domenico & Reichlin, Lucrezia & Simonelli, Saverio (2009), "Nowcasting Eurozone Economic Activity in Real-Time: The Role of Confidence Indicator," National Institute Economic Review, vol. 210, pp. 90 - 97.
Giannone, Domenico & Reichlin, Lucrezia & Small, David (2008), "Nowcasting: The real-time informational content of macroeconomic data," Journal of Monetary Economics, Elsevier, vol. 55(4), pp. 665-676, May.