In August 2007, when the euro area was first hit by the worst financial crisis after the Second World War, its economy already displayed signs of cooling from the peaks reached in 2006.
€-coin is a real-time, monthly estimate of area-wide GDP growth, computed each month by the staff of the Banca d’Italia. In this column we address two related questions:
- How did €-coin track the loss in momentum in euro area growth over the last 2 years?
- Is a conjunctural indicator like €-coin reliable in a situation like the current one, in many respects markedly different from the more recent historical experience?
€-coin tracking of the current cyclical downturn in the euro area
Synthetic cyclical indicators – like €-coin – are designed to portray in a single figure the economic situation by “properly” weighting the available information. According to €-coini, euro area growth cycle peaked in 2006, when the indicator estimated the underlying growth to be around 3.5% on an annual basis (Figure 1).
Figure 1. €-coin
In the second quarter of 2007, euro area growth began to lose momentum; however in July, right before the effects of the financial crisis stemming from the US subprime mortgage market began to spread, euro area trend growth was still estimated close to 3%. From August 2007, financial and survey data – that account for roughly one-third of the data series used to produce €-coin – quickly reflected the deterioration in the financial markets and the growth outlook worsened further. €-coin dropped to 2.0% (on an annual basis) by the end of last year and bottomed at zero in October 2008, its lowest level since 1993. The only exception to €-coin’s steady decline since the start of 2007 was a small rebound last spring due to a spurt in industrial productions and a slight improvement in the survey data.ii
Figure 2. Euro area
However, as in all economic downturns, the current one did not occur simultaneously across macroeconomic variables. Quantitative data, like industrial production or unemployment, showed clear signs of a slowdown in the euro area only in the first quarter of this year when the unemployment rate began to rise and the industrial production index flattened (Figure 2, upper left panel), several months after the sudden jump in money market spreads had signalled the seriousness of the crisis (see bottom right panel). Survey and stock market data pointed at a deterioration of the economic outlook much earlier than quantitative data (second and third panel). Consistently, since August 2007, business opinion and financial data have been the major driving force of €-coin, accounting for the largest part of its decline, while real variables played a minor and, at times, counterbalancing role (as in the second quarter of 2008).
This occurred also in the US, where the financial crisis originated and where its consequences should have been felt sooner and more sharply. Many key macroeconomic time series like employees on non-agricultural payrolls or industrial production showed signs of a weakening in real activity only at the beginning of 2008 (fig. 3).iii Notably though, unlike €-coin that declined earlier, the Conference Board coincident index for the US business cycle, whose indications are often in line with the NBER dating committee decisions, did not mark a clear fall until the beginning of 2008.iv
Figure 3. United States
Comparing €-coin with the index published by the Conference Board helps clarify the different nature of the two indicators – while the latter is a “pure” coincident index, €-coin combines leading, coincident and lagging variables to obtain a better approximation of the unobservable underlying growth rate. To this end, the €-coin database includes also financial and stock markets variables which are used to provide a more timely, reliable and smoother signal of the economic conditions with respect to synthetic indexes based solely on coincident macroeconomic time series. At the same time it retains the character of a cyclical indicator, since it discounts the vagaries contained in financial and stock market variables (essentially what the statisticians define as noise) retaining only the information valuable for the macroeconomic outlook: by doing so it strikes a balance between coincident and leading indexes.v
The main conclusion is therefore that by appropriately weighing all the information, including that contained in financial data, €-coin delivered a timely and smooth signal that the underlying growth in the euro area was falling. This fact emerges even more clearly if we compare the signal given by €-coin with that derived by other indicators (that enter the €-coin information set) in real-time (see video): (1) €-coin summarises the sometimes divergent patterns of the various indicators (2) €-coin, smoother than its components, does not overreact to specific events unlike the individual indicators.
The financial crisis and the expert opinion: Too little, too late?
A closer look at market perceptions and expert evaluations of the effects of the financial crisis on the real economy can help us to better assess the reliability of a synthetic indicator for the euro area in current economic circumstances, as well as from a quantitative point of view (i.e. €-coin seen not only as a signal of underlying trend, but also as an quantitative estimate of growth).
The current financial crisis has rapidly affected the entire financial system, particularly the banking sector – bank runs and major bank failures, events that had been almost unheard of for a long time in advanced economies, are now deemed a possible outcome against which governments and central banks have taken extraordinary steps. Credit tightening and scares of a credit crunch have made their way to the newspapers headlines. A recent analysis by the IMF found that episodes of financial turmoil accompanied by banking distress tend to be followed by more pronounced downturnsvi. Moreover, the magnitude of the negative impact of the financial crisis on the real side of the economy appears to be correlated with the rise in house prices and the growth of credit in the periods preceding the outburst of the financial tensions.
The concurrence of all these aspects in the present crisis suggests that its unravelling might differ from the typical cyclical downturns of the last decadesvii. This makes one wonder whether we are well equipped in terms of statistical tools to gauge the evolution of the real economy in a timely fashion. In particular, one might ask whether synthetic indicators, like €-coin, built with the “usual cyclical fluctuations” in mind (and in the data) are really able to take appropriately into account all the relevant information in the present conjuncture.
To answer this question, we compare the evolution of €-coin with assessments of the economic outlook by the business community. Compared to a simple statistical tool like €-coin, the main advantage of business opinions and experts’ forecasts is that they can make judgmental corrections or factor in information and “animal spirits” which are not reflected in official statistics until their effects have become all too visible.viii
We hence compare the evolution of €-coin with that of GDP growth forecasts by leading experts (sampled by Consensus Forecasts) over the last two years. The main message is that experts reacted to the incoming news, but only gradually adjusted their views. Prior to August 2007, forecasters were anticipating a growth rate of 2.3% in the euro area in 2008 and the most pessimistic evaluations still set growth at 2.2%. Three months later average growth was down to 2.0% according to the same source and the lower quartile set it at 1.8%. The downward revision continued in 2008 and after a rebound to 1.7% in June it fell by six-tenths of a percentage point in the following four months, bottoming at 1.1% at the end of last October. Growth prospects for 2009 fell from 2.0% to 0.4% from January 2008 to October 2008. The size of these corrections reveals that it took some time for experts to fully realise the extent of the impact of this financial crisis on the real economy. This is even clearer if one considers the forecasts for 2009: most of the downward correction took place after last summer. In contrast, €-coin clearly signalled that the real economy was suffering the effects of the crisis, possibly to a larger extent than what envisaged by the available projections.
As a last check, the fact that €-coin has provided more accurate signals about the depth of the downturn than those suggested by professional forecasters does not prevent the possibility of a weakening in its “quantitative” ability to track the medium-term GDP growth. To check for this possibility we analysed the errors of €-coin, with respect to year-on-year GDP growth one quarter ahead (i.e. we compare €-coin in – say – March 2008, with year-on-year GDP growth in 2008Q2)ix. The size of the errors is roughly the same before and after the start of the crisis in August 2007 (i.e., across the potentially distinct regimes).x
In conclusion, €-coin has suggested a downturn in the euro-area economy since the beginning of 2007, promptly recognised a stronger deceleration after August 2007, and signalled that the outlook had become much gloomier after last summer. The evidence presented in this column supports the view that, in the current downturn, €-coin has continued to perform as good a “growth tracker” for the euro area in comparison to both professional forecasters and in historical perspective.
Altissimo F., R. Cristadoro, M. Forni, M. Lippi and G. Veronese (2007), “Eurocoin: tracking economic growth in real time” in Banca d’Italia, Temi di discussione n. 631 and CEPR WP n. 5633 (forthcoming in Review of Economics and Statistics).
Cristadoro, R. and G. Veronese, (29 October 2007) “€-coin and the euro area economic outlook: Where do we stand?” VoxEU
Forni, M., M. Hallin, M. Lippi, and L. Reichlin (2005), “The generalised dynamic factor model: One-sided estimation and forecasting”, Journal of the American Statistical Association 100 830-40.
Leamer, E.E.(2008) “What's a recession, anyway?”, NBER WP 14221.
Stock, J.H. and M.W. Watson (2002), “Forecasting using principal components from a large number of predictors”. Journal of the American Statistical Association 97, 1167-79.
Zarnowitz, V (2008) “Present conditions of the US economy in an historical perspective”
i A short description of €-coin can be found in Cristadoro and Veronese (29 October 2007). A detailed account of the methods applied and the data used is in Altissimo et al. (2007). The indicator is published in eurocoin.cepr.org or at www.bacanditalia.it.
ii Which was also embedded in the slightly less pessimistic view of the economic outlook among market watchers.
iii Many observers claimed that, at least until June 2008, there were no clear signs of a serious slowdown in the US (see among others Zarnowitz, 2008 and Leamer, 2008).
iv See the October 2008 press release of the Conference Board US business cycle indicators available on the web at http://www.conference-board.org/pdf_free/economics/bci/LEI1008.pdf.
v The leading index of the Conference Board for the US economy showed signs of an incoming downturn much earlier than the coincident index. See again http://www.conference-board.org/pdf_free/economics/bci/LEI1008.pdf.
vi See Chapter 4 of the IMF World Economic Outlook, October 2008.
vii To quote the most recent ones: the currency crisis in 1998 and the dotcom bubble in 2001.
viii For instance being aware of the longer history of past crises that extends beyond the sample span on which models and indicators are typically estimated (like, say, the 1929 depression) can be an important competitive edge in a time of “unusual” turmoil.
ix It might be at first sight surprising that we choose the y-o-y growth with lead 1 to assess the performance of €-coin. As stressed in paper (see Altissimo et al., 2007) €-coin tracks growth in the euro area trying to estimate its medium to long run component, i.e. a component that is free from short run and statistical noise. This component is clearly non observable, but, when we have enough “future” data we can estimate it with a bilateral bandpass filter or, in simpler terms, by taking a centred moving average of quarterly growth rates with appropriate weights. The lack of data at the end of the sample is what makes this operation unfeasible and €-coin useful. This bilateral weighted moving average can be crudely proxied as a simple moving average of quarterly rates, an y-o-y growth rate is an example of this and taking a lead of one serves to “center” it with respect to the running quarter.
x The mean absolute error from 1999Q1 to 2007Q2 was 0.283 if we consider the most recent period (2007Q3-2008Q1) the mean absolute error is only slightly bigger (0.298).