### Thursday, November 29, 2007

## It depends

It's one of these half-full/half-empty stories. The national economic news is considered delightful, as GDP growth in the third quarter was revised up (as expected) to 4.9% from an advanced estimate of 3.9%. Yay! In the same report we find corporate profits fell. Boo! Home sales went up. Yay! Prices went down. Boo! Forecasted growth for 2007 is up. Yay! The forecast for 2008 is down. Boo! Foreclosures are up 94% in the last year, but they've leveled off. Booyay?!? Yayboo?!?

What's a poor economist to do?

Pay attention to the revisions, and the effects of one-time shocks. This third quarter figure appears to be one of these. When I write a forecast within a model, I evaluate it based on whether my guess for X next period (or some periods out in the future) is close to what X turns out to be. The percentage change in X isn't the point of the forecast for the most part.

So suppose I am forecasting the value of X now (call this time t) for two periods from now (t+2). You ask me to represent that as a growth rate, and so I do the necessary calculation. You ask for a forecast for t+1 and I give you that as well, again as a growth rate.

Now let's suppose I receive data on X for next period, and suppose it's higher than that one-period-ahead growth rate I gave you. Do I revise now my estimate for X for period t+2? The answer, as always in economics, is "it depends". It depends on whether there was any information in the one-period-ahead data that causes me to change my mind about what happens two periods ahead. There may or may not be. So, for example, the CEA forecast for GDP was moved up for 2007 and down for 2008 because, for the most part, there's been no change in their medium-term forecast. The jump in 2007:III GDP is moving some of the growth of GDP between mid-2007 and end-2008 into that quarter.

An imperfect analogy: Mario Mendoza is a .200 hitter in baseball (thus the Mendoza Line). After establishing this level of batting crapitude for years, one day he goes four-for-five. Do you revise your opinion of Mario's hitting prowess, or do you think it's just one of those days (which would be expected to occur randomly about once a season?) Of course not -- hitting has a random quality. So too does GDP or about every other phenomema in economics.

Revisions also play a role. Calculated Risk shows how not focusing on the revision can completely change how you view the home sales data. Others have noted as well that these repeated revisions in previous-month home sales is softening the focus on that market's decline.

One last thought thought that might make things more rosy. Cindy discusses the state revenue forecast due Friday her in Minnesota, and fears that it will come in very negative. Revenue forecasts need a base on which to tax, so they forecast output, not percentage change in output. If today's GDP numbers turn out to be true, the additional income in the Minnesota economy is present longer, and is taxed each period. Even if a slowdown occurs next year, higher revenue collections now are already in the bank for the state government. I have no idea what the forecast will be, but a revenue shortfall of less than $500 million is possible, and that would only chew up the reserves and not cause blood on the tracks in St. Paul. Much more than that, though, and it will be one helluva spring at the Legislature.

I'll be on Heading Right Radio at 2:30pm CT to discuss the state of the economy; those reading this post are invited to question me there!

What's a poor economist to do?

Pay attention to the revisions, and the effects of one-time shocks. This third quarter figure appears to be one of these. When I write a forecast within a model, I evaluate it based on whether my guess for X next period (or some periods out in the future) is close to what X turns out to be. The percentage change in X isn't the point of the forecast for the most part.

So suppose I am forecasting the value of X now (call this time t) for two periods from now (t+2). You ask me to represent that as a growth rate, and so I do the necessary calculation. You ask for a forecast for t+1 and I give you that as well, again as a growth rate.

Now let's suppose I receive data on X for next period, and suppose it's higher than that one-period-ahead growth rate I gave you. Do I revise now my estimate for X for period t+2? The answer, as always in economics, is "it depends". It depends on whether there was any information in the one-period-ahead data that causes me to change my mind about what happens two periods ahead. There may or may not be. So, for example, the CEA forecast for GDP was moved up for 2007 and down for 2008 because, for the most part, there's been no change in their medium-term forecast. The jump in 2007:III GDP is moving some of the growth of GDP between mid-2007 and end-2008 into that quarter.

An imperfect analogy: Mario Mendoza is a .200 hitter in baseball (thus the Mendoza Line). After establishing this level of batting crapitude for years, one day he goes four-for-five. Do you revise your opinion of Mario's hitting prowess, or do you think it's just one of those days (which would be expected to occur randomly about once a season?) Of course not -- hitting has a random quality. So too does GDP or about every other phenomema in economics.

Revisions also play a role. Calculated Risk shows how not focusing on the revision can completely change how you view the home sales data. Others have noted as well that these repeated revisions in previous-month home sales is softening the focus on that market's decline.

One last thought thought that might make things more rosy. Cindy discusses the state revenue forecast due Friday her in Minnesota, and fears that it will come in very negative. Revenue forecasts need a base on which to tax, so they forecast output, not percentage change in output. If today's GDP numbers turn out to be true, the additional income in the Minnesota economy is present longer, and is taxed each period. Even if a slowdown occurs next year, higher revenue collections now are already in the bank for the state government. I have no idea what the forecast will be, but a revenue shortfall of less than $500 million is possible, and that would only chew up the reserves and not cause blood on the tracks in St. Paul. Much more than that, though, and it will be one helluva spring at the Legislature.

I'll be on Heading Right Radio at 2:30pm CT to discuss the state of the economy; those reading this post are invited to question me there!