September 09, 2011
This is an abridged version of a CEPR paper available here.
There are many economists who argue that temporary tax cuts, like those in 2009 stimulus and the ones proposed by President Obama last night, have no impact on the economy. They argue that people will save a temporary tax credit rather than spend it.
Stanford University Professor (and Hoover Fellow) John Taylor is one of the economists making this argument. He purports to show that there was no statistically significant increase in private consumption of goods and services as a result of certain types of government transfers made over the last decade. According to his analysis, it is unclear whether an additional dollar of government transfers led to any additional spending, or, alternatively, whether it raised personal savings by more than one dollar.
However, a closer look reveals something quite different.
If Taylor’s analysis is correct, then such personal transfers are not clearly effective forms of stimulus. People do not necessarily spend such transfers, and with no additional demand for their products businesses have not any incentive to expand. Obviously, if spending desirable it would be far more effective for the government to directly increase its own purchases of goods and services, rather than giving money to the private sector and hoping for the best.
Nevertheless, some of the stimulus money surely to people who were desperate for cash. It is unlikely that none of the money was spent even if many individuals saved their entire share. Teasing out these effects is a tricky business. Obviously, economic stimulus payments are associated with bad economic times, and these times must somehow be accounted for in any analysis. The early $300 Bush rebate checks were almost all sent in the third quarter of 2001. It should hardly require mentioning that determining the effect of this particular transfer of money might be confounded by certain attacks in New York City and Washington, DC which took place in that very quarter.
Taylor’s regression uses oil prices, lagged two quarters as an explanatory variable in his model, but once his data is updated to the latest GDP numbers, the size of the oil effect is implausibly large. The fall in the price of crude oil between the third and fourth quarters of 2008 appears to have added 10.2 percentage points (annualized) to consumption growth in the second quarter of 2009.
In all likelihood, this is a spurious relation. In fact, substitution of the Google trend for worldwide traffic of “obama” for lagged oil prices yields an improved fit. More seriously, there is every indication that Taylor’s model lacks a critical break that distinguishes post-2007 consumption from 2000-07. Whether or not oil prices are included in the model, accounting for the break results in statistically significant effects of stimulus on consumption.
I hesitate to suggest that unlike Taylor’s regression, these structural-shift models are well specified and that the resulting estimates of spending rates are accurate. Indeed, though statistically significant, the lack of precision in the estimates does not make them terribly meaningful economically. However, these regressions do demonstrate that evidence regarding Taylor’s conclusion is sensitive to the specification of the model.
The empirical uncertainty surrounding the personal transfers points to the importance of domestic government expenditures as stimulus. There, the direct impact on GDP is identically equal to 1.0, regardless of any subsequent multiplier effects. When and where crowding-out effects are unlikely given large amounts of unused capacity (both people and machines out of work) the impact may be very large. However, there is very little indication that – based on Taylor’s work – personal transfers from the government fail to stimulate private spending.
