Pasar al contenido principal

Flawed Study Should Be Given No Credence in Evaluating Jobs and Revenue Impact of California Corporate Tax Break


A coalition of California corporations has released an economic analysis of the job and revenue gains the state supposedly can expect to see if a corporate tax break is allowed to go into effect next year. The study is so flawed, however, that it should not be given any credence in evaluating the potential impact of the tax break on California employment and tax receipts.

The study is based on a statistical analysis of job changes in five other states. It asserts that California will experience job gains and no state tax revenue loss if it implements a “single sales factor apportionment formula” (SSFF) for the state corporate income tax in 2011, as scheduled under 2009 legislation. The formula, a part of the state’s tax code, determines the share of a multistate corporation’s nationwide profit that the state will tax. Proposition 24, on the November 2, 2010 ballot, would block implementation of the new formula along with two other corporate tax breaks if the voters approve the measure; the older, less generous formula, would be retained.

The study, by university professor Charles W. Swenson, makes its predictions based on changes in employment patterns in Georgia, Louisiana, New York, Oregon, and Wisconsin over the 2003-2008 period, a period in which each of those states implemented SSFF. Swenson compares the rate of job creation at each employer, in each state, during each of two periods – 2003-05 and 2006-08. He finds that employers were more likely to add more jobs in the later period if they had a set of characteristics that, in his opinion, are likely associated with benefitting from this particular tax break. He concludes from this that benefitting from the tax break caused employers to accelerate job creation, and estimates a statistical relationship between the jobs and the tax-break-affected characteristics. He then applies this conclusion to California, and predicts modest job growth. Finally, using that estimate of job growth and a related estimate of increased corporate profits, he predicts very large tax revenue growth.

Flaws in this study include the following:

  • Jobs data that diverge dramatically from official government statistics . Official U.S. government data, as well as a broad consensus of economists, holds that 2003-2008 was a period of employment growth across the country. In fact, the U.S. Bureau of Labor Statistics finds that in both the 2003-2005 period and the 2006-2008 period, in all five states studied, there was job growth. But Swenson is not using publicly available, official data, but rather a proprietary dataset. And according to the figures he reports from his dataset, in all five of those states, in both periods, there was a large decline in employment. There are other elements in Swenson’s reported data that also suggest it may contain errors, such as huge, erratic, and inexplicable swings in calculations of total sales. Unfortunately, because it is a private database available only to those able to pay relatively large fees, it is not possible to determine whether the source of these apparent errors is the database itself or the way Swenson has manipulated it. But they might well render this analysis worthless.
  • Misstatement of time periods. Swenson writes on p. 4 of his study that all five states “switched” to SSFF “in 2006,” which if true would validate his approach of contrasting job creation in the 2003-05 and 2006-08 periods in those states. He calls these “natural experiments.” In fact, four of the five states implemented SSFF gradually over this period, and the implementation periods vary greatly. For example, Oregon was well on its way to full implementation of SSFF in 2003; Georgia by contrast did not start it until 2006 and did not fully implement SSFF until 2008. As a result, Swenson’s differentiation of the two periods as a clearcut pre-SSFF period and a post-SSFF period is wrong.
  • Misidentification of which firms benefited from the tax break. Swenson’s conclusions are entirely dependent on the accuracy with which he categorizes firms as being likely to benefit from SSFF or not. But his categorization is extremely rough; his data do not actually allow him to identify employers that have out-of-state sales, employees, or property, so he uses a series of rough proxies. One of his proxies, for instance, is simply whether the employer has subsidiaries, even though SSFF is not limited to corporations with subsidiaries. Another proxy is whether the employer is a publicly traded corporation, again legally unrelated to SSFF eligibility. Yet another is the industrial sector of the business; Swenson contends that the single sales factor formula is inapplicable to retail store chains in “most of the five states examined,” when in fact it applies to such businesses in all of them.

    As it turns out, his strategy for identifying businesses that are likely to be eligible for SSFF is demonstrably incorrect, and not even close. Swenson concludes that there were 3,109 firms in Georgia that benefited from the state’s adoption of SSFF, but the state’s own study found that the actual number was 12,426, almost four times as high. Swenson classified 1,796 Wisconsin firms as likely to be affected by the SSFF, but a state study put the number at 6,404, more than three times as many. In two other states as well, the state estimate of SSFF-affected firms is several times larger than Swenson’s. In other words, Swenson’s technique for measuring the impact of the tax break excludes most of the businesses that were actually eligible for it. This proves that the methodology is well off the mark.
  • What about the other states that didn’t enact SSFF in this time period? Normally, one would expect a study of the connection between state tax policy and economic growth to look at all 50 states if at all possible. But this study completely ignores all the other 45 states, including the 23 states with corporate income taxes that have never implemented the SSFF. Without such a “control group,” we cannot know whether the changing employment patterns that Swenson claims to have discovered were specific to those five states, or whether they were in fact widespread across the U.S. economy — and therefore entirely independent of SSFF implementation. It is totally possible, for instance, that what Swenson’s data actually show is that the 2006-08 period was much better for big, multistate corporations in every state — for some reason unrelated to state tax treatment — than for smaller employers. In fact, there are at least some national data that suggest this is the case.
  • Faulty tax assumptions. Even if Swenson’s conclusions about job creation were right, which for the reasons described above they probably are not, it would not necessarily prove that SSFF implementation is a good idea. Other analyses of the impact of corporate tax cuts on state economic growth — including one based on the state’s own model of the California economy — have found that it would cost California state government a huge amount of tax revenue that it can ill afford to lose. But Swenson finds that, in fact, the new jobs (and associated boost in corporate profits) would produce so much tax revenue that it would offset the cost of the tax break. To make this calculation, he assumes that every dollar of new corporate profits would increase California tax revenue by 19 cents, and the average new worker would pay an additional $5,167 in California personal income and sales taxes. But these are unrealistic assumptions. California business taxes as a share of California corporate profits are probably half as much as Swenson assumes. And a worker earning the average annual wage in California of about $50,000 generally will pay significantly less than $5,167 in combined income and sales taxes.
  • Complete lack of transparency and peer review . Standard practice in economic analysis is to present, for a reader’s inspection, the calculations and data on which the analysis was based. The Swenson report does not make even the most minimal attempts to show how the results of its statistical analyses translate into actual predictions of job and revenue growth in California; policymakers and potential expert reviewers are supposed to take it on faith that these calculations were done properly. Nor is there any evidence that it was submitted to any independent economists for review before publication. Nor has Swenson responded to repeated attempts by one of the authors of this report to contact him for his methodology and results.
  • Math and typographical errors . Finally, it is worth noting that the Swenson report is rife with careless arithmetic and presentation errors; even simple percentage changes between two numbers are repeatedly miscalculated. This further complicates any attempt to understand and critique his methodology, since many of his results cannot be replicated.

These are just a few of the flaws in the Swenson study; they are explained further, and other flaws are described, in the remainder of this report.