Sunday, August 15, 2010

Separable Preferences, Review of My Forecasts

When, in early 2009, I began modeling this recession and the recovery, I considered whether my model should reflect the fact that non-work time and market consumption are substitutes. Professors Aguiar and Hurst have done a lot of work on this topic, showing convincingly that people cut their market spending when they retire from their jobs, because they have more time to do tasks like cooking and shopping, and because they no longer need work related expenses.

In theory, the same could be true when people temporarily(?) loose their jobs, as in this recession. However, Professor Hurst advised me that unemployed people do not use their spare time to replace market consumption expenditures to the degree that retired people do. For that reason, and for ease of computation, the model I constructed last year assumed separable preferences.

However, even if Aguiar and Hurst are correct about the time allocation of unemployed people (and those patterns have repeated themselves in this recession), it is clear that unemployed people do not make work related expenditures like working people do. So for now I have decided to consider a measure of consumption that excludes personal expenditures on transportation, in addition to the measured I had considered before, without making any changes to my model and the forecasts derived from it.

The figures below show the data (through mid-2010) and my forecast scenarios. Recall that I have always given most weight to the "partial recovery scenario" shown in green in each of the Figures.

When consumption is measured to include transportation, and includes the BEA's latest revisions (see the black circles in Figure 5c), it seems to conform to the no-recovery scenario! But, for the reasons cited above, that is a mismatch between the model and data. The model refers to a consumption drop that derives solely from a wealth effect, whereas I think that transportation spending also falls because that spending is associated with going to work.

The consumption series without transportation spending (shown as black triangles in Figure 5c) conforms more closely to the "partial recovery scenario." So that scenario continues to be the likely one, although with the BEA revision I now see some of the first suggestions that it might be a little bit too optimistic.

As shown in Figures 5b, 5d, and 5e, the partial recovery scenario continues to fit the labor usage data very well, and fits the investment data well during the past couple of quarters. The model fits the general productivity trend (Fig 5d), although not all of the quarter-to-quarter changes.

No comments: