Abstract
This study examines whether state movie production incentives are effective in attracting and/or retaining movie production. The issue is of significant policy interest because of the large amounts spent by states for such subsidies. This study finds that while movie production incentives were effective in increasing the number of film production employment and establishments for a few states such as New York and California from 1998 to 2011, there was no discernable increase across all states. Much of this noneffect appears because of a “crowding out” effect due to the sheer number of states with incentives.
This study examines whether state movie production incentives (MPIs) are effective in attracting and/or retaining motion picture production. After noting tax breaks given by Canada, Louisiana was the first state to initiate state movie production tax incentives in 1992. In 2005, five states had such tax breaks, and the number of states has now increased to 45. MPI supporters note a few studies done by state governments or consulting firms, which seem to support that local job growth occurred, and conjecture that tourism to such states increased after movie exposure. Critics claim that little job growth occurs, or that such growth disappears once production ceases, and that vague statute wordings leave room for abuse.
Despite billions of dollars in incentives given by states, 1 no academic studies have been done to examine whether such incentives are effective. A few state-specific studies have examined changes in economic activity, often using government data on tax benefits awarded. Examples include Pennsylvania, which examined applications and estimated an employment impact of 3,960 jobs (after multipliers) in 2007-2008, with a net revenue loss of $40 million (Economic Research Associates, 2010; Independent Film Office, State of Pennsylvania, 2013) and Louisiana, where 14,011 jobs were estimated to be created in 2012 after multipliers, at a net revenue cost of $168 million (Loren C. Scott & Associates, 2013).
Other examples include Connecticut, where 395 full-time-equivalent (FTE) jobs were estimated to be created at a net cost of $15 million (Connecticut Department of Economic and Community Development, 2008); Massachusetts, which estimated a direct employment increase of 864 FTEs in 2011 at a net state revenue cost of $37 million (Massachusetts Department of Revenue, 2013); Michigan, which estimated 1,542 jobs created after multipliers in 2009-2010 at a net revenue cost of $85 million (Zin, 2010); and New Mexico, 3,829 jobs created in 2007 after multipliers at a net state cost of $25 million (Ernst & Young, 2009). 2 Averaging across these studies, the cost per job created was $27,424. This cost seems in line with costs per job of recent federal programs. For example, Chodorow-Reich, Feiveson, Lisow, and Woolston (2012) estimated a $30,000 cost per job for the American Recovery and Reinvestment Act.
In addition to being state specific and using estimates, these studies may not have examined the counterfactual (or “but for” conditions) by controlling for economic activity that would have occurred in the absence of tax incentives. 3 The purpose of this study is to provide a multistate analysis of the impact of movie production tax incentives on employment, which attempts to control for this counterfactual. This study has a narrow focus insofar as it examines only direct employment in the film production industry, before multipliers. This narrow focus neglects that movie productions can result in other economic benefits, including local spending and longer, more diffuse impacts because of the promotional/branding impacts of movies on other activities such as tourism. 4
The narrow focus here, however, allows for potentially more precise tests. 5 By-state results indicate that, while certain film-intensive states such as California and New York, show postincentive employment growth, other states decreased employment, and there was little employment impact in the aggregate. Econometric results, using a difference-in-differences (DID) method, also indicated no significant net employment gains after incentives. The results suggest the possibility of crowding-out effects across states, which begs the question whether there might be a more cost-efficient solution for state governments. 6
Film Production Incentives Offered by States 7
Forty-six states offered some type of film production incentive from 1998 to 2011. One potentially very lucrative incentive was cash grants, which ranged up to 17% of costs; however, grants were only offered by Tennessee, Texas, and Virginia. Another potentially lucrative incentive was income tax credits. Twenty-seven states gave tax credits that could offset much, if not all, income taxes normally due in that state. Because a number of filming enterprises will generate little or no tax in any year 8 due to being just cost centers with revenues coming in later years, a number of states allowed film companies to either sell their credits to other buyers 9 or to actually have them refunded as if they were a negative income tax. 10 A third, potentially very lucrative incentive, is cash rebates for certain qualifying production costs. The 17 states offering rebates required preapplication, and approval was competitive. A fourth potentially important incentive was preproduction grants, offered by only a few states.
Two incentives were also offered, which a priori appear to be small. Twenty-eight states allowed sales tax exemptions on purchases made by filming companies. Because sales tax rates tend to be less than 8% in most states and apply only to purchases of tangible personal property (not services), how much of a savings this can offer a film production company depends on the magnitude of purchases made locally. Thirty-two states offered lodging tax exemptions (which is a selective tax on hotel guests, often imposed at city levels) if crews stay more than 30 days. Because most states offer such exemptions to transients staying more than 30 days, there may be minimal cost savings from this incentive.
Three general types of restrictions apply to film production incentives. First, most states required preapproval (or at least prenotification before start of filming) to obtain benefits. 11 Second, a number of those states also had a maximum dollar amount of income tax credits that can be awarded per year, resulting in a “first come, first served” award. Finally, a number of states had a minimum required investment if tax credits are to be awarded. 12
It is worthwhile to examine costs of these programs to the states. Table 1 shows annual tax benefit costs by state for those states reporting such data, for the 1998-2000 period. 13 The highest annual cost during that period is shown. Such expenditures varied widely by state; California, Florida, and New York spent in excess of $100 million, whereas some states spent as little as a couple of million dollars per year. Also shown are expenditures such as percent of total state budgets. These averaged .091% and ranged from as little as .0079% in Indiana to 0.7065% in Louisiana.
Prior Estimates of Costs for Film Production Incentives.
Note. States where estimates could not be found are omitted. Data from state budgets and film commission sources. Highest annual amount from 1998-2011 period shown. Amounts are actual tax benefits awarded if published; otherwise, state-budgeted annual amounts are shown. Total state budgets derived from U.S. Census Bureau State Government Finances (2013).
Analysis
There are a number of potential ways to examine the impact of incentives on location choice for movie production. Unfortunately, there is no comprehensive, across all states, data on movies produced. 14 Such data might be difficult to interpret; because movies may be filmed in a number of states, simply showing some production in a state could involve large ranges in actual investment in that state. In fact, the only regularly collected information, by state and over time, for this industry is payroll data collected by state employment tax authorities. This becomes the basis for census data, discussed below. Essentially, an establishment that has “employees” in a state must file payroll tax returns for such employees.
A disadvantage of the census data is that it does not count independent contractors who work in much the same capacity as W-2 employees. For many film productions, such contractors, or “below the line” workers, represent a significant cost. They include set creators, makeup people, movie extras, and other local people who work only temporarily for that production. Since there is no by-state database of so-called “1099” workers (they are issued a federal tax form 1099 by the companies for which they work), examining only W-2 employee data from Census will undercount the true employment impact of the industry. Because W-2 employees are always involved in a production, however, we have a crude measure of film production scope in a state.
Table 2 shows annual employment for the film production industry, by state, for 2000-2011. Both raw employment and raw employment as a percentage of total private-sector employment (times 100) are shown. Data are from the Census Bureau’s County Business Patterns for NAICS codes 512110-512199. Although California’s film production employment is 5% of that state’s employment, the industry, in general, is a relatively modest proportion of total state private employment, averaging approximately 0.17%. 15 Annual employment is reported for each state before and after film production incentives were enacted, with postincentive employment shown in bold. Also shown is the year each state enacted film incentives and the type of incentive offered. 16 The last column of the table shows that aggregate U.S. employment as a percent of total employment steadily increased over time.
Raw (Relative) Annual Employment for Film Production Industry for States.
Note. Posttax employment shown in boldface.
At the national level, industry average employment was approximately 114,000, which translated into about .04% of overall state employment. During this period, relative U.S. annual employment grew by 3.8%, but the aggregate U.S. growth was actually slightly lower in periods where tax incentives were in place, averaging across states; preincentive mean relative employment was .0501 and postincentive relative employment was .0416 (a t test indicated that this difference was significant at .05). Twenty-three states showed increases in mean relative employment after incentives were adopted, 12 showed decreases, and the remaining states either had no incentives in place or there were insufficient data to perform a meaningful comparison. 17 States that increased postincentive employment experienced, on average, a 24.7% growth, with California and New York (easily the two largest producers of films) having relative employment growths of 37.9% and 27.2%, respectively. States decreasing postincentive employment experienced, on average, a 12.7% decline. In general, the results do not provide compelling evidence that incentives increased employment in the aggregate.
The unevenness of postincentive growth by states can be seen by examining some of the states individually. Figures 1A through 1H show relative employment for a sample of eight states from 1998 to 2011. Vertical bars separate pre- and postincentive years. While California, New York, North Carolina, and Pennsylvania appear to have postincentive relative employment increases, Florida, Massachusetts, and Texas do not. For all states, there is no discernable stable pattern pre- or postincentives.
Relative employment before and after tax incentives.
In examining whether MPIs have an employment impact, we of course want to control for other factors that could affect location choice. Unfortunately, there is very sparse academic literature on the location choice decisions by film producers. Texts and professional publications indicate that such locations are a function of the type of setting that fits with the movie script. Location scouts develop lists of potential areas and studios/producers make a final decision. Because movies vary widely in the types of settings that they require (Katz, 2005; Levy, 2000), they are idiosyncratic and not easily empirically specified. The professional literature, however, suggests that there is a type of agglomeration effect 18 insofar as, over time, certain locations begin to develop more support/infrastructure for movie productions; Coe (2001) also conducts an academic study of this effect in Vancouver. 19 In the absence of any existing empirical models, we can attempt to control for such trending and other effects with a DID specification.
Since the work by Ashenfelter and Card (1985), the use of DID methods has become very widespread in policy analysis. The simplest case is one where outcomes are observed for two groups for two time periods. One of the groups is exposed to a treatment in the second period but not in the first period, and the second group is not exposed to the treatment during either period. If the same units within a group are observed in each time period, the average gain in the second (control) group is subtracted from the average gain in the first (treatment) group. This removes biases in second-period comparisons between the treatment and control group that could be the result from permanent differences between those groups. I also remove biases from comparisons over time in the treatment group that could be the result of trends. Using a regression of the form y = β0 + β1dΒ + δ0d2 + δ1d2 * dΒ + e, the regression estimate of d2 is a dummy variable for the second time period. The dummy variable dΒ captures possible differences between the treatment and control groups prior to the policy change. The time period dummy, d2, captures aggregate factors that would cause changes in y, even in the absence of a policy change. The DID coefficient, δ1, multiplies the interaction term, d2 * dΒ, which is the same as a dummy variable equal to 1 for those observations in the treatment group in the second period.
As shown in Woolridge (2002, 2007), the DID method for a panel of data involves time and group dummy variables, with the DID variable being a dummy variable set to 1 for the treatment group during the treatment periods. Accordingly, I specify:
where Yit is film production industry relative employment (or relative number of establishments) for state i in year t,
The dependent variable in Equation (1) is annual film production employment in year t for each state i, divided by total private-sector employment in year t for each state i (i.e., relative employment), and as noted above is taken from County Business Patterns published by the U.S. Census Bureau. An alternative dependent variable in Equation (1) is annual film production establishments in year t for each state i, divided by the total establishments in year t for each state i (i.e., relative establishments), also taken from County Business Patterns.
Table 3 reports mean and standard deviations for the regression variables. DID regression results for relative employment and relative number of establishments from 1998-2011 for all states are reported in Table 4. The regressions in columns 1 and 3 have a single MPI dummy variable, and regressions in columns 2 and 4 use separate dummy variables for each incentive type. 20 Although the relative number of establishments increased after incentives, this increase was only about 1%. In contrast, there was no increase in relative employment postincentives. None of the individual incentive variables are statistically significant in either the relative employment or relative number of establishments regressions; credits, grants, sales tax breaks, and tax-free lodging do not have a significant impact. As discussed previously, there are generally minor amounts of subsidy associated with sales tax waivers and tax-free lodging, so the lack of significance on the coefficients associated with these subsidies is not surprising. Because there were only two states with grant programs, the lack of significance for that incentive should be interpreted with caution as well. Nonetheless, the insignificance of other incentive variables is notable.
Descriptive Statistics for State Panel, 1998-2011.
Difference-inDifferences Regressions for Relative Employment and Relative Number of Establishments by State.
Note. Huber–White heteroscedasticity-consistent standard errors reported in parentheses. DW = Durbin–Watson statistic.
Significant at .05 or better. **Significant at .01 or better. ***Significant at .001 or better.
The high R2 of these models averaging more than 90% is essentially because of the explanatory power of state and year fixed effects. It seems likely there was a crowding-out effect of incentives, as employment switched between states with incentives. As seen in the previous tables and graphs, some states experienced postincentive employment gains and other decreased employment during similar time periods, amounting to a “wash” in the aggregate. Note that the Table 4 results are not influenced by collinearity; Table 5 reports correlations across the variables, and an eigenvalue analysis revealed no collinearity issues.
Correlations Between Study Variables.
Alternative Measures
It is possible that the above results may have been data-driven. Recall that the analysis examines only direct employment in film production, NAICS codes 512110-512199. The census data are drawn from state employment data and includes only people who are classified as employees. If film producers hire many of their local workers as independent contractors, official employment data will undercount the job impact. Indeed, “below the line” people such as set designers and workers, makeup artists, and extras may be hired as independent contractors on a temporary basis.
Larger estimates can be obtained by examining data that considers independent contractors in the employment database, which is what is included in annual surveys of businesses done by Dun & Bradstreet (D&B). 21 A version of the D&B data is available for purchase by researchers through a vendor, referred to as the National Establishment Time Series (or NETS) database. These data have other unique advantages over census data, since it is reported at a disaggregated level for each establishment and has information on births, deaths, and moves (including origin and destination locations). There are more than 51,000 film production establishments in the data set. Employment is generally two to three times larger than what is reported in census data. Table 6 reports DID regressions for relative employment using this database. As can be seen, results are not much different from census data regressions; tax incentives are not significant. 22
Difference-in-Differences Regressions for Relative Employment by State Using Dun & Bradstreet Data.
Note. Yule–Walker–corrected standard errors reported in parentheses. DW = Durbin–Watson statistic.
Significant at .05 or better. **Significant at .01 or better. ***Significant at .001 or better.
That the regression results indicate a washing-out effect in the aggregate is supported by individual state D&B data reported in Appendix Table A1. This table reports relative employment by year, as well as postincentive changes. Such changes are calculated as average annual relative employment after incentives minus average annual relative employment before incentives. Although 21 states increased relative employment after incentives, this increase was relatively minor for most states. Nine states decreased their relative employments, and the remainder of the states either had no incentives or incentives were in effect through the study period, obviating any before and after analysis.
As noted above, an advantage of the D&B data is information on moves by establishment. Although relatively few film establishments move in any 1 year, examining such data may give additional insight insofar as transactions costs are nontrivial and if tax incentives are effective, we may see a pattern of pre- and postincentive moves. 23 Appendix Table A2 reports D&B move data by state, for states having incentives that went into effect during the study period. Employment moves are separated into years during incentives versus other years, and employment moves both into and out of the state are shown. As with all other analyses, results here are mixed. Seventeen states showed a net migration improvement during incentive periods. That is, net moves during incentives (moves into state minus moves out of state) were larger than net moves in nonincentive periods. Nineteen states showed a loss in net employment due to moves during incentive years. Noteworthy positive results include those for New York: preincentives, 234 employees were moved in while 191 moved out. Once 2007 incentives were in place, 552 employees moved into the state and only 92 moved out. Similarly, before incentives were enacted in 2009 in California, 126 establishments with 882 employees moved into California. In contrast, 182 establishments with 1,762 employees moved out of the state. Thus, there was a flight out of California before incentives could be enacted. After incentives the trend reversed, with 140 employees moving into the state and 92 moving out. Similar large, positive, net-move employment effects are reported for Florida, Massachusetts, and Texas. At the other extreme is Maryland, where before incentives 61 employees were moved in and 54 moved out; after incentives, 39 were moved in and 105 were moved out. Similar “worse off” effects are shown for Georgia and Oregon.
The conclusion to be drawn from the D&B data is that, despite its different origin and measurement, it provides results very similar to that of census. That is, we observe very mixed effects of incentives, with some states experiencing growth (most notably, California and New York) and others showing no growth or decline.
Conclusion
This study examined whether state MPIs are effective in attracting and/or retaining movie production employment. The issue is of significant policy interest 24 because of the large amounts spent by states for such subsidies. Although California and New York significantly increased direct employment after incentives were enacted, other states showed moderate growth, no growth, and even negative postincentive growth. When we examine all states in econometric models that control for other factors, the results indicate insignificant increases in direct employment and very minor increases in number of establishments. A different data source that captures independent contractors also found no increase in net employment.
Such “mixed” findings on the general effectiveness of state tax incentives over a variety of types of business decisions seem consistent with surveys of the literature reported in Buss (2001) and McGuire (2003). 25 The results here suggest that the effectiveness of tax incentives were influenced by a “crowding out” effect from numerous states, all offering incentives. This latter result seems consistent with results found by Chirinko and Wilson (2008), who found that the effectiveness of state tax incentives in the manufacturing sector were something of a zero-sum game between states over time.
It is important to reemphasize that some of the analysis here undercounts total employment in the film production industry, since indirect employment was not considered (i.e., industries ancillary to movie production were not examined, via multiplies or otherwise). Furthermore, actual movie production data were not examined, nor were longer and more diffuse positive effects of movies such as state tourism increases considered in this initial study. Future work may consider extending the analysis beyond the 2011 time period examined here and the possible impact of postproduction facilities. 26
Footnotes
Appendix
Employees Moved In and Out of State.
| State | Incentive years | Moves in when there are no incentives | Moves out when there are no incentives | Net change in moves when there are no incentives | Moves in when there are incentives | Moves out when there are incentives | Net change in moves during incentive years |
|---|---|---|---|---|---|---|---|
| AL | 2009-> | 26 | 4 | 22 | 3 | 1 | 2 |
| AR | 2009-> | 17 | 4 | 13 | 5 | 0 | 5 |
| AZ | 2006-9 | 35 | 37 | −2 | 18 | 13 | 5 |
| CA | 2009-> | 882 | 1762 | −880 | 140 | 92 | 48 |
| CO | 2006-> | 20 | 138 | −118 | 40 | 23 | 17 |
| CT | 2006-> | 52 | 23 | 29 | 31 | 40 | −9 |
| FL | 2004-> | 151 | 486 | −335 | 177 | 117 | 60 |
| GA | 2005-> | 76 | 39 | 37 | 29 | 43 | −14 |
| HI | 2008-> | 11 | 9 | 2 | 19 | 7 | 12 |
| ID | 2008-> | 2 | 6 | −4 | 11 | 1 | 10 |
| IL | 2004-> | 20 | 12 | 8 | 124 | 149 | −25 |
| IN | 2008-> | 11 | 13 | −2 | 5 | 19 | −14 |
| IA | 2007-8 | 3 | 18 | −15 | 2 | 0 | 2 |
| KS | 2007-8, 2011 | 29 | 33 | −4 | 2 | 8 | −6 |
| KY | 2008-> | 17 | 13 | 4 | 50 | 11 | 39 |
| ME | 2006-> | 12 | 2 | 10 | 8 | 1 | 7 |
| MD | 2008-> | 61 | 54 | 7 | 39 | 105 | −66 |
| MA | 2006-> | 26 | 298 | −272 | 41 | 41 | 0 |
| MI | 2008-> | 66 | 65 | 1 | 33 | 65 | −32 |
| NJ | 2005-9 | 83 | 189 | −106 | 52 | 59 | −7 |
| NY | 2007-> | 234 | 191 | 43 | 552 | 92 | 460 |
| NE | 2006-> | 81 | 154 | −73 | 73 | 25 | 48 |
| OH | 2009-> | 34 | 72 | −38 | 10 | 51 | −41 |
| OR | 2005-> | 47 | 23 | 24 | 2 | 13 | −11 |
| PA | 2004-> | 6 | 10 | −4 | 67 | 30 | 37 |
| SC | 2004-> | 15 | 3 | 12 | 30 | 34 | −4 |
| SD | 2006-> | 2 | 0 | 2 | 1 | 5 | −4 |
| TN | 2006-> | 128 | 58 | 70 | 125 | 18 | 107 |
| TX | 2009-> | 103 | 300 | −197 | 54 | 26 | 28 |
| UT | 2007-> | 29 | 90 | −61 | 17 | 9 | 8 |
| VA | 2002-> | 49 | 8 | 41 | 119 | 121 | −2 |
| WA | 2007-> | 314 | 14 | 300 | 59 | 28 | 31 |
| WV | 2007-> | 9 | 6 | 3 | 2 | 3 | −1 |
| WI | 2008-> | 31 | 29 | 2 | 3 | 11 | −8 |
| WY | 2007-> | 11 | 1 | 10 | 2 | 12 | −10 |
Acknowledgements
The very helpful comments of Eric Allen are gratefully acknowledged.
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.