With the news that GDP grew 3.5 percent in the third quarter of 2009, it seemed apparent that economic recovery was under way. How much of this was a result of government programs? To evaluate this, it is important to understand what constitutes a recovery. There are three developments needed to restore the economy to its prior vibrancy.
The first development, bank stabilization, began in late autumn of 2008. The source of the recession was financial-sector turmoil that commenced in August 2007 and peaked in early autumn of the next year. Although we did not know it at the time, by the end of 2008 the financial crisis had passed. Financial markets were far from normal, but the panics and major collapses that characterized September 2008 were behind us, and no others arose. This financial-sector stabilization created the environment that is allowing our economy to heal.
Last January, at the end of my term as chairman of the President’s Council of Economic Advisers, my agency released the White House economic forecast. At that time, I said that I foresaw a couple of bad quarters but expected that the second half of 2009 would be positive, with perhaps very strong growth in 2010. These forecasts assumed no stimulus; the projected turnaround was instead based on the natural rebound of the economy that would come after the financial crisis had eased. The resumption of GDP growth, which is the second development on the road to full recovery, probably began in the late spring of 2009.
The third recovery factor—job growth—will be slower to develop. In a shallower recession that ended in late 2001, job growth did not become positive until 2003. Historically, recoveries have a consistent pattern: productivity grows first, then jobs are created, and finally wages rise.
A crucial question at this point is, how important were government programs in bringing about positive economic growth? There is good evidence that a portion of third-quarter growth can be attributed specifically to the cash-for-clunkers program. The disproportionate increase in purchases of motor vehicles relative to other consumption components is evidence that the program altered expenditures.
It is less likely that the housing program, which consisted primarily of a tax credit for first-time home buyers, was responsible for much of the turnaround in housing investment that occurred during the third quarter. Housing investment depends on longer-term demand shifts. The program’s limited duration would be expected to reduce existing inventories more than it would be expected to stimulate additional building and investment.
Direct government expenditures also increased during the third quarter, but most of this was in defense spending, which was not a primary target of the stimulus programs.
Third-quarter growth was 4.2 percentage points higher than second-quarter growth, which was negative 0.7 percent. My estimate is that less than half of this growth can be attributed to government programs. Of course, cash for clunkers has ended, and the sharp decline in auto sales in September means that there will be some payback for third-quarter growth, but proponents of the program argue that the shift in timing was a decent trade-off.
After reporting GDP, the government released new numbers claiming that the stimulus programs have “created or saved” more than a million jobs. These data were collected from responses by government agencies that received federal funds under the American Recovery and Reinvestment Act of 2009. Agencies were required to report “an estimate of the number of jobs created and the number of jobs retained by the project or activity.” This report is required of all recipients (generally private contractors) of agency funds.
Unfortunately, these data are not reliable indicators of job creation or of the even vaguer notion of job retention. There are two major problems. The first and most obvious is reporting bias. Recipients have strong incentives to inflate their reported numbers. In a race for federal dollars, contractors may assume that the programs that show the most job creation may be favored by the government when it allocates additional stimulus funds.
No dishonesty on the part of recipients is implied or required. But when a hire conceivably can be classified as resulting from the stimulus money, recipients have every incentive to classify the hire as such. Classification as stimulus-induced is even more likely if a respondent must only say that except for the money, an employee would have been fired. In this case, no hiring need occur at all.
Another subtle but important point: during the cash-for-clunkers program, sales of small cars like Honda Civics went up while sales of large SUVs like the Ford Expedition went down. To discern the net effect of the program on auto sales, it is necessary to take into account not only the addition to sales of subsidized models but also the reduction in sales of discouraged models. Reporting the positives without the negatives would be misleading and would complicate attempts to objectively evaluate the success of the program.
Yet this type of gains-only reporting is precisely what the government is doing with respect to the figures on stimulus-induced new hires. When the government reports this figure, it wants us to believe that the new hires came from the pool of the unemployed and that they are net additions to the stock of employed workers. But the data do not speak to the number of workers who left their current jobs to fill government-sponsored jobs.
If the vacancies that are created as these workers move from their old firms to government-sponsored projects go unfilled, then these job-to-job transitions are negatives that must be subtracted from the positives. And in an economy that is growing on the basis of productivity improvements rather than more workers employed, firms may wait to fill vacated jobs.
Because these data do not tell us where the workers come from and what happens to the slots they leave, the numbers cannot answer the ultimate question: how many net jobs were created? A similar point holds for those who are reported as retained, only in this case the issue is that some of those retained would have moved to different jobs rather than to the pool of unemployed had they been let go. The government is reporting the gross positive figures, not the relevant net figures.
Each month, the Department of Labor reports hires and layoffs from the Job Openings and Labor Turnover Survey (JOLTS). The JOLTS data revealed that in August 2009 more than 4 million workers were hired. But, unlike the administration’s new jobs-created-or-saved data, the JOLTS data also tell us that in the same month about 4.3 million workers lost their jobs.
As a result, the labor market lost more than a couple hundred thousand jobs. It would be misleading to examine only the hires without also looking at job losses when evaluating the conditions of the labor market, but that is exactly what the stimulus job accounting does.
Net labor market figures do exist. Administrations have always been held to the time-tested and well-understood monthly job numbers put out by the Bureau of Labor Statistics, which reports the unemployment rate and the net job gain or loss for the economy as a whole. It is important to use reliable, accurate, and well-understood numbers to determine the true causes of recovery. The national unemployment rate, at 10 percent as of December, has stayed high, and job losses also have remained at high levels throughout the stimulus period. Few will be comforted by the good-news-only claim that the stimulus “created or saved” more than 1 million jobs.