What do leading indicators tell us about the future of the real estate market? Quite a lot, as Sotiris Tsolacos reports
The use of data series that signal changes in the fortunes of the real estate market, for both occupier and the investor segments, is gaining ground. And there is more research on this subject area, fostered by the abrupt swings of the last cycle. Have the various data series generated insightful signals into the likely direction of the occupier and investor markets in the past two to three years?
In the abundant amount of research on leading indicators in the sphere of business cycles and the stock market, there is much evidence that some series, most of which have the very attractive feature of not being revised backwards, convey important advance information about changes in the economy and asset prices. An interesting observation is that these series vary by geography, but some consistency can be observed. Typically, series that are found to contain leading information for, say, GDP growth, inflation, or the stock market, are combined into a single series (using alternative weighting schemes), conventionally termed a leading indicator.
Leading indicators exist for the most developed countries, and recently the Conference Board launched a leading economic index for the euro-zone's economic activity. At this time of euro-zone turmoil, this index is undoubtedly closely watched. It is composed of eight economic, monetary, stock market, interest rate, and sentiment variables. All are supposed to contain forward-looking information for the euro-zone's GDP movements. According to the Conference Board, this leading economic index showed weaknesses in the euro-zone economy as early as January 2008; indeed, GDP growth turned out to be 0.5% in 2008 and -4% in 2009. It also predicted the 2000-03 slowdown. Therefore, this index can convey valuable advance information for the coming trends in the occupier office market and near-term rent movements in the euro-zone and can be a valuable tool in the hands of investors.
We examine the capacity of this index to predict the euro-zone's prime office rents and construct a stock-weighted prime office rent index for the euro-zone, based on the largest office centres. Preliminary statistical analysis does establish that the euro-zone leading indicator (EzLI) does precede movements in prime office rents, both in real and nominal terms; further analysis shows that the advance information provided is a six-to-15-month window. This is a truly appealing feature. We exploit the leading properties of EzLI and study what signals it would have generated and still generates, three quarters or nine months in advance for rent movements.
Figure 1 shows the actual quarter-on-quarter office rent growth in the euro-zone. Rent growth slows down from 3Q 07 and eventually declines (negative quarterly growth) in 4Q 08. Using a statistical model with EzLI as a predictor for rents, we obtain implied rent growth estimates. The model begins to signal rent declines from 2Q 08, when it implied a marginal fall in rents. The model generated that signal nine months (three quarters) earlier, that is, in 3Q 07. The subsequent 4Q 07 forecast for rent growth in 3Q 08 also pointed to a fall in rents. Rent declines began in 4Q 08, but EzLI had produced early signs for this change in rent evolution. Throughout 2009, the leading indicator suggested a fall in rents, and these predictions were, of course, obtained from a nine-month lead window.
More recently, prime office rents showed a strong rebound in 1Q 10 when they rose by 3%. Nine months ago (2Q 09), the model would have pointed to a softer rent decline, -1.2% in 1Q10. In 3Q 09 (or more precisely, at the end of October 2009, when both the EzLI and rent data would have been available for 3Q09), the model says that positive rent growth returns in the period April to June 2010 (2Q 10). Again, the changing direction was picked up satisfactorily.
Moving forward with 2Q 10 data available both for EzLI and prime rents (end July 2010), the model paints a good picture, one of positive quarterly growth in prime office rents in the euro-zone as a whole. In the first quarter of 2011, the model predicts a slowdown.
Using statistical models containing leading indicators to predict a reference series is only one way to make predictions. Another approach is to focus on directional forecasts and calculate changing probabilities for possible turning points. In the context of prime office rents in the euro-zone, we ask the question of what the probability is for rents to decline (negative quarterly growth) nine months later and how this probability changes as we roll on through the forecast.
In figure 2, we plot the quarterly rent growth along with probabilities that rent growth will be negative; the probability is estimated three quarters ahead. Using data up to 2Q 07, we estimate a probability of 49% that rent growth will be negative in 1Q 08; rents eventually showed a growth of 1.5%. The probability of negative growth began to rise from 3Q 08 and became slightly higher in 4Q 08 (63%), when rents indeed declined over the previous quarter. This probability was estimated with data up to 1Q 08. Again, this was a good early call, since the analysis was pointing to adverse movements in prime office rents becoming more likely.
The signal for a rebound in rents is registered in 4Q 09 (generated by the model in 1Q 09), but the probability of 74% is still high to imply positive growth. In 1Q 10, when rent growth does rebound strongly, the forecast probability fell to 53% for negative rent growth; it had declined considerably from 81% only two quarters earlier. This forecast was made in 2Q 09. In the forecast obtained in 3Q 09, the probability for negative quarterly rent growth in 2Q 10 fell to 47%. As the graph illustrates, the probabilities estimated nine months in advance for negative rent growth for 1Q 10 and 2Q 10 were the lowest since 3Q 08. This again constitutes good performance in predicting the changing direction in rent growth.
The news is good for the second half of 2010 and first quarter of 2011. With EzLI and rent data available for 2Q10 at the end of July and since we predict nine months ahead, we are able to take a view for rents until 1Q 11 (end March 2011). The leading indicator implies lower probability for negative rent growth over this period and confirms the results for the analysis behind figure 1.
What we should bear in mind is that the success of leading economic indices in predicting real estate markets differs by geography and sector. Indeed, the composition of the indices themselves varies from country to country. Hence, there is no consistency in the specification of forecast models. On the other hand, the extra effort of exploring the usefulness of series proven to lead economies — which have not suffered historical revisions — could really pay off.
Sotiris Tsolacos is director, European research, at PPR