Grasping Reality on TypePad, by Brad DeLong

In email, lurkers are questioning my claim that:

Forecasting the Obama Economy: ...what happened to the Mankiw CEA over the winter of 2003-2004, when high politics appears to have reached down into the forecast, changed the table for payroll employment (and only payroll employment: the rest of the forecast is not out of line with contemporary professional forecasts), and produced an estimate for December 2004 (a) inconsistent with the rest of the forecast, and (b) high by 2.3 million in its estimate of payroll employment--all because Karl Rove and company thought it important to avoid headlines like "Bush administration forecasts 2004 payroll employment to be less than when Bush took office." White House Media Affairs would have a much harder time pressuring the forecasters to produce a "rosy scenario" if the pressure has to be kept on month after month [as the Troika forecast is revised, updated, and released at a monthly frequency].

I think that the smoking gun is provided by a little forensic table reading--going through the Bush administration's economic forecasts year-by-year as they were published in the successive versions of the Bush-era CEA's Economic Report of the President, the ERP:

• In the 2002 ERP, Table 1.1 shows 3.2% growth expected for the next two years gives you 2.9 million jobs--for a forecast labor productivity growth rate of about 2.1% per year...
• In the 2003 ERP, Table 1.1 shows 3.5% growth expected for the next two years gives you 4.4 million jobs--for a forecast labor productivity growth rate of about 1.8% per year...
• In the 2004 ERP, Table 3.1 shows 3.7% growth expected for the next two years gives you 6.2 million jobs--for a forecast labor productivity growth rate of about 1.3% per year...
• In the 2005 ERP, Table 1.1 shows 3.4% growth expected for the next two years gives you 4.1 million jobs--for a forecast labor productivity growth rate of about 1.8% per year...
• In the 2006 ERP, Table 1.1 shows 3.3% growth expected for the next two years gives you 3.8 million jobs--for a forecast labor productivity growth rate of about 1.9% per year...
• In the 2007 ERP, Table 1.1 shows 3.0% growth expected for the next two years gives you 3.3 million jobs--for a forecast labor productivity growth rate of about 1.8% per year...

The forecast rate of labor productivity growth over the next two years or so is a relatively stable variable. It starts at an annual rate of 2.1% in the first Glenn Hubbard ERP, and then Glenn and company drop it to 1.8% the next year as they become less optimistic about productivity growth in the aftermath of the collapse of the high tech bubble. Thereafter the Bush CEA forecast assumes a labor productivity growth rate of 1.8% - 1.9% in every year save one: the 2004 ERP, issued at the start of 2004, drops the labor productivity growth rate to 1.3% (and the 2005 ERP raises it back up to 1.8%).

Was there anything in the economic data that would make one much more pessimistic about labor productivity growth in early 2004 and only early 2004? No.

But assuming a 1.8% labor productivity growth rate at the start of 2004 would have meant that the forecast average level of employment in Tqble 3.1 for 2004 would have been lower than the level of employment when Bush took office, and that would have created a point of political vulnerability. There were two ways to fix this that would have satisfied White House Media Affairs: (i) reformat the table so that it no longer reports an annual average payroll employment number, or (ii) push assumed labor productivity growth down because if you keep GDP the same but reduce labor productivity arithmetic forces your forecast to produce higher employment.

Why the Bush CEA didn't pick option (i) is something I have never understood...