The Motion Picture Association is refuting a new study by Michael Thom, associate professor at the University of Southern California’s (USC) Sol Price School of Public Policy, for poor data selection, methodology problems, and its authors repeated dismissal of positive effects from production incentives that demonstrate a clear pattern of bias.
“It is unfortunate that a well-respected academic institution that counts some of the greatest filmmakers, directors, and actors among its alumni, is promoting this fundamentally flawed research,” said Vans Stevenson, Senior Vice President for State Government Affairs at the Motion Picture Association – America. Reputable economists have found that Associate Professor Michael Thom’s previous work had serious deficiencies, and this study is no different.”
“In addition to poor data selection and problems with methodology, Thom’s repeated dismissal of positive effects from production incentive programs demonstrates a clear pattern of bias,” Stevenson continued. “It’s no surprise that the research was funded by the Koch Foundation, which has a long history of leading a campaign to eliminate state production incentive programs.”
A previous 2016 study conducted by Prof. Thom that was reviewed by the Motion Picture Association, Oxford Economics, and the Los Angeles Economic Development Corporation (LAEDC) also found similar omissions of positive effects and other serious flaws.
MPA Analysis of Associate Professor Thom’s Study
An initial review of the study conducted by the MPA found the following flaws:
- Although Prof. Thom’s study uses a previously established model (Linden, 2015), it does not provide the necessary evidence that the data and methods used are consistent with those Linden identifies as necessary to use the model correctly. In fact, based on his prior (California) study and given that incentives exist in the “pre-intervention” phase of all states examined, Thom’s data does not appear to meet the specifications needed for the model. This means the model doesn’t accurately measure the effects of the incentive in each state.
- The study focuses on the wrong metric by measuring year-on-year percentage change rather than real employment figures. Percentage point change is related to the size of the value, masking real changes in larger numbers, while producing potentially large jumps in smaller figures that may not be meaningful. In other words, the study over-emphasizes changes in small states. Additionally, the use of percentage point change suggests that only growth (increase in employment) is important, when maintaining production and employment in a state also matters, and in some cases may be the aim of an incentive program.
- The study does not account for how incentives and competition work. The model includes controls for percentage point changes in incentive spending in other states and Canada but cannot account for how production/employment would have been allocated without the presence of incentives, regardless of how spending year-on-year changes. For example, the presence of Canada’s incentives influences the location choices of productions, even if overall incentive spending in Canada is flat.
- Interestingly, even with these issues, the primary explanatory variable (percentage point change in own-state tax expenditures) included in the paper does find positive and significant effects on percentage change in employment in Louisiana, Georgia, and Massachusetts, along with an initial impact effect of incentives in Connecticut and a positive, ongoing effect of incentives in Louisiana. Thom dismisses these results for various reasons, but they suggest that increased spending on incentives can have an effect on employment and that with properly structured data and models, measurable effects would likely be greater than shown by the current analysis.
- This new study looks at several states, including Connecticut, but fails to mention that the Connecticut production incentive program, unlike the others, is restricted to television projects only. The other states considered in this study have a film AND television production program.
