Economists Forecast Modest 3rd Quarter Rebound

.
A panel of 45 U.S. economists surveyed by the National Association for Business Economics (NABE) expect a modest rebound in the 2nd half of 2009 and then the economy strengthening from there in 2010. Quote from the report:

The panel ... [predicts] positive, albeit modest, growth in the third quarter, followed by steady improvement thereafter,... On the whole, however, the economic rebound lacks luster. Real GDP growth over the second half of 2009 is expected to average a well-below-trend 1.2 percent pace.

So I was curious how well NABE economists surveys had forecasted in the past. A year ago back in May of 2008 their forecast said:

[Expect the] current economic downturn is likely to be short and shallow with a
slow recovery.

Not quite right on the short and shallow but correct on the slow recovery. Going back and looking at past forecasts by the economists surveyed by NABE it looks like they tend to underestimate the possibilities of the extremes both good and bad. But that is just a generalization, it is hard to really do any good statistical analysis on this type of inconsisent and somewhat qualitative forecast.

Some GDP Charts to Look At.

"Real" (adjusted for inflation, in 2000 dollars) GDP by quarter on an annualized basis:

The current recession is the hook downward at the end. But it is always good to put things in a long term perspective which in this case the overall trend is unquestionably up with relatively constant slope on log scaling which means a relatively constant compounded growth rate.

Real GDP % Change Each Quarter:

The zero line is shown so you can see easily when the GDP was growing or contracting. The current recession is at the end of the plot on the right. Also notice the considerable reduction in variability since about 1985.

An ARIMA Forecast Model for GDP

So I wasn't really going to even try but I couldn't help it. I took the data on monthly % GDP changes since 1985 since that was since the drop in volatility. When fitting many statistical models, relatively constant variation is important. It fit a fairly simple ARIMA model which had statistical significance. It was a 2 autoregressive term model. Here is what the forecast says going forward:

Interesting recovery curve shown in blue with confidence intervals in red. So my forecasting model suggest the rest of 2009 quarterly GDP %change to be negative and the first postive GDP change to come in the 1st quarter of 2010. Now I would take this model in general terms, not in exact terms. The general shape of the forecast curve actually looks reasonable to me as an armchair economist but there is alot of obvious quarter to quarter variability here so the forecast line ought to be viewed as an average forecast going forward with quarters potentially being above and below that line. So with that noise, even if this forecast was as good as it could be, you very well could see the GDP going positive a quarter early or a quarter late.

In the interest of full disclosure, in case you were wondering, a few months back this ARIMA model would not have forecasted such a steep decline in GDP like we have seen in the last couple of quarters. And I would not hold this ARIMA model as high as I would my Case Shiller ARIMA forecasting model.

Will be interesting to come back in a year and see how this time series forecast did. Although I am only an armchair economist and have no real experience to make a knowledge-based GDP forecast, I really hold alot of value in the high level of objectivity a statistical forecasting model has. Sometimes, expert knowledge about a subject does not lend one to make good forecasts because of the extreme bias that expert knowledge imparts.
...