Ready for a Close-Up: The Effect of Tax Incentives on Film Production in California

Abstract

Do motion picture incentives nudge productions to film in an adopting state, increase budgets, or hire more cast and filmmakers? Or do they simply subsidize productions that would have occurred regardless? Using California’s Film and Television Tax Credit Program, the author exploits incentives given out by lottery to answer these questions. The author finds that 19% of films would have filmed in California even without a tax incentive, but that the offering of an incentive increased the probability of a film being made in California by 16 percentage points. Both production budgets and the number of cast and filmmakers increased in response to the offering of a tax incentive, and, in California specifically, they increased budget spent by 267% and the number of cast and filmmakers hired by 123%. The program also had differential impacts based on film type, with only nonindependents increasing budgets and the number of cast and filmmakers in California.

Keywords

economic development policy tax incentives state taxation film industry

In the United States, motion picture incentive (MPI) programs are both widespread and generously funded. Thirty-three states offered a MPI program in 2020 and, although that number has fallen from a peak of 44 in 2009, the amount of funding has only increased. In 2010, an estimated $1.5 billion in MPIs were offered in the United States (Tannenwald, 2010). In 2020, despite a net decline of 11 states, more than $3 billion (author’s calculation¹) were allocated to MPIs. Notably, a handful of states are largely responsible for this increase—Georgia, Illinois, California, Kentucky, and New York. Georgia alone increased its funding from $33.5 million in 2010 to $860 million a decade later.

Clearly, MPI programs are a popular economic tool with substantial financial backing, but are they an effective economic development tool? Many researchers have tried to answer this question using methodologies of varying degrees of sophistication. Cost-benefit studies represent the majority of the literature, though their approaches and assumptions are by no means uniform. Several researchers have supplemented these cost-benefit studies by using regression analysis to approximate the causal impact of MPI programs, frequently using a difference-in-differences (DiD) approach. Unlike cost-benefit studies, where the comparison groups are a state before MPI adoption versus after, these studies typically compare states without MPI programs versus states with them. In these regressions, the dependent variable was a measure of film activity, like motion picture spending, and the explanatory variable was typically expressed as a binary variable to indicate the availability of MPIs.

Not surprisingly, conclusions on the efficacy of MPI programs differed based on both the methodology used and the sophistication of the researcher’s model. The majority of state-conducted cost-benefit studies (Christopherson & Rightor, 2010; McDonald, 2011; Tannenwald, 2010; Thom, 2018b) found negative returns on investment (ROI), though two studies, both funded in part by the Motion Picture Association of America (MPAA), found positive ROIs (Ernst & Young, 2009a, 2009b). Similarly, academic researchers using regression analysis also reached different conclusions, with some having found negative impacts (Lester, 2013; Owens & Rennhoff, 2018), while others noted mixed or small positive effects (Adkisson, 2013; Button, 2018, 2019; Swenson, 2017; Thom, 2018a, 2019).

This article is thus unique from the existing MPI literature in several respects. First, its methodology is experimental by exploiting the fact that version “1.0” of California’s Film and Television Tax Credit used a lottery to assign credits to productions. Second, its research sample is at the production level rather than at the state level. This difference in sampling allows a wider and more granular definition of MPI program efficacy. For example, in addition to quantifying metrics common to the existing literature (e.g., how many more productions get made, how much production spending increased, and how many more cast and filmmakers were hired), production-level sampling allows me to determine how many productions would have filmed in California even without an incentive and which production types are more likely to benefit from incentives. Finally, by focusing on California specifically, this article offers conclusions on MPIs in a state with substantial agglomeration economies in terms of entertainment production. As a result, its conclusions are likely upper-end estimates of MPI program efficacy.

Specifically, to best make use of the experimental data, I employ two measurement strategies. The first offers an estimate of the program’s intention to treat (ITT) effect and the second estimates the treatment effect on the treated (ToT). The first measurement strategy reveals the impact of California’s MPIs on films that were offered an incentive but that did not necessarily receive one. The second estimates its impact on films that ultimately received one. Because the tax incentive lottery was random in the offering of tax credits (productions did not have to take an incentive), the second strategy uses whether a production was initially offered an incentive as an instrument to estimate the program’s effect on productions that ultimately received an incentive.

I find that both the offering and accepting of a tax incentive are associated with positive and statistically significant results. In my preferred specification, I find that simply offering a tax incentive to a film increased the probability that it was made in California by 16 percentage points, increased the production budget spent in California by 267%, and led to 123% more cast and filmmakers hired in the state. For films that accepted a tax incentive, the probability of being made in California increased by 58 percentage points, the production budget spent in California increased by 966%, and 388% more cast and filmmakers were hired in California. Depending on which measurement strategy is used, 19% or 13% of productions would have filmed in California without being offered or receiving an incentive, respectively.

Notably, tax incentives had differential impacts based on a production’s film type. In California, they appear to have much stronger effects for nonindependent films than for independent films. Despite the sample size for nonindependents being roughly half that of independents, budget spent, number of cast and filmmakers, and the percentage of filming locations in California only statistically increased for nonindependent films. Furthermore, the magnitude of these coefficients was several times larger for nonindependents. Given that these productions tended to spend more and hire more cast and filmmakers in general, these represent particularly large effects.

Literature Review

The Logic Behind Tax Incentives

Economic theory holds that tax incentives are justified if there are market failures that lead to underinvestment in a productive sector of the economy (Buss, 2001). Underinvestment can stem from underemployment in a productive labor group or direct underinvestment in a productive economic activity (Noll & Zimbalist, 1997). In both cases, tax incentives can encourage production in a highly productive sector that is currently being underutilized, paying a short-term cost that is more than compensated by a long-term gain.

Tax Incentives in the United States

While tax incentives have been employed since colonial times, their cost was minimal and their usage was regulated to the fringe of economic development policy until the second half of the 20th century, particularly the 1990s, when their cost and usage skyrocketed (Burnett, 2011). Today, approximately $50 billion worth of business incentives are allocated annually, 94% of which are tax incentives (Bartik, 2019). Despite several recent high-profile state attempts to lure individual firms (e.g., Wisconsin’s $3 billion incentive to Foxconn and several states’ incentive proposals for Amazon’s second headquarters), business incentives tripled from 1990 to 2001 but have remained relatively constant since then (Bartik, 2019).

The debate continues on the efficacy of tax incentives, despite their scope and scale. Arauzo-Carod et al. (2010), Buss (2001), Peters and Fisher (2004), and Wasylenko (1997) provided excellent surveys of the literature. Initially, researchers concluded that incentives had ambiguous or very minor impacts at best. Starting with Bartik (1991), however, a new view emerged that incentives did have positive, if marginal, impacts on economic development. More recent research, however, only added to the uncertainty. While some studies found marginal or negative effects (e.g., Hanson & Rohlin, 2013; Lee, 2008; Prillaman & Meier, 2014; Taylor, 2012), others found large positive effects (e.g., Strauss-Kahn & Vives, 2009; Wilson, 2009; Wu, 2008; Zhang et al., 2018). Most of the current literature on tax incentives has focused on manufacturing industries, ignoring other sectors like entertainment, finance, and information technology despite their growing importance to the economy. As a result, the vast majority of the literature’s conclusions applied to industries that were more relocation-averse, likely dampening the effect of incentives. MPIs are thus an interesting case study in the larger tax incentive literature, as they enlarge the literature’s scope to include a service industry and an industry well known for its footloose nature.

The Rise of MPI Programs

The turn of the 20th century saw a fundamental shift in the use of MPI programs in the United States. Up until the 1990s, incentives consisted primarily of in-kind aid—state-run agencies assisted productions by helping identify shooting locations, providing access to public facilities (for example, Christopherson & Rightor, 2010). States then began offering tax breaks in addition to these services. What started as a niche policy used by a handful of states in the early 2000s quickly expanded to more than 40 states a decade later (Tannenwald, 2010). Industry watchers were quick to point to Canada’s rollout of MPIs both at the national and provincial levels in the late 1990s, which resulted in “runaway production” from the United States. Many hypothesized that the explosion of MPIs across the United States in the decade after was an attempt to prevent further runaway production from states with motion picture industry concentrations and, in states without deep concentrations, to emulate the success of their northern neighbors. While there was likely some truth to this competition narrative, several scholars have used event history analysis to quantify the role of competition as well as other factors in MPI adoption. Leiser (2017), for example, found that two factors largely explained the diffusion of MPI programs across the United States: the preexisting concentration of a film industry in a state and a competitive bandwagon effect. Interestingly, Thom and An (2017) found no relationship between a state’s motion picture industry concentration and its enactment of a MPI program. Like Leiser (2017), they found evidence of a bandwagon effect that was primarily the result of national rather than bordering state trends. Thom and An (2017) also found that rising unemployment rates were associated with MPI program enactment, though Leiser (2017) did not find support for this economic distress hypothesis. These analyses showed that MPI adoption did not operate in a vacuum and that a state’s decision to enact a program was, in part, motivated by the size of its film industry, its rate of unemployment, and national adoption trends. Consequently, MPI adoption is far from exogenous and researchers that employ the DiD approach must take these factors into account.

MPI Program Evaluations

The current literature on MPI programs can be divided into cost-benefit evaluations conducted by state agencies and consultants and independent academic evaluations that use regression analysis. While the former category represents the bulk of the literature and is state specific, the latter consists of a handful of studies and tends to look across states. Although many states have completed cost-benefit evaluations of their MPI programs, they differ remarkably in their rigor of analysis. Many simply calculated the difference between total film and television spending and the program’s cost to determine its gross impact, though a few states calculated the net return of their MPI programs by differentiating new spending from spending that was likely to occur anyway.

California and Massachusetts both fall into the latter camp, but they differentiate new spending in different ways. California does so by deducting spending from projects that would have likely filmed in California even without an incentive by using the lottery data that underpin this article. Massachusetts, by contrast, constructed a baseline measure of production spending by totaling spending from advertising, TV series, and documentaries prior to the state’s introduction of a MPI. Notably, no feature films were included in this baseline measure of production.

By no means exhaustive, the states that have made evaluations public include the following: Alabama, Arizona, California, Connecticut, Florida, Georgia, Louisiana, Maryland, Massachusetts, Michigan, Minnesota, Mississippi, Missouri, New Mexico, New York, North Carolina, Oklahoma, Pennsylvania, Rhode Island, South Carolina, Virginia, and Washington. See the online supplementary appendix for references to each state’s report.

According to these state evaluations, nearly all MPI programs resulted in negative ROI (Christopherson & Rightor, 2010; McDonald, 2011; Tannenwald, 2010; Thom, 2018b). Christopherson and Rightor (2010) reviewed 14 state evaluations and concluded that MPIs have a negative impact on state revenues for most states. They noted that only two studies found positive returns—one in New Mexico and one in New York. Both studies were conducted by Ernst and Young and were funded in part by the MPAA. Tannenwald (2010) reached the same conclusion by examining eight state evaluations and calculating revenue gained from feedback effects per dollar of film subsidy claimed. Again, positive returns are only found in the Ernst and Young studies for New Mexico and New York. The remaining states all reported losses, ranging from 0.07 to 0.28, meaning that, for every dollar spent on MPIs, states recouped between 0.07 to 0.28 cents. Massachusetts, the sole state evaluation among Tannenwald’s sample to recognize that some production would have occurred without incentives, calculated an ROI of 0.16. Thom (2018b) provided a more recent summary of state ROI for nine states. Like previous summaries, Thom concluded that all states lost money on MPIs, ranging from 0.07 to 0.46 cents per dollar of MPIs.

In addition to the multitude of state cost-benefit evaluations, a few independent academic studies have sought to quantify the effects of MPI programs by looking across, rather than within, adopting states. Like their cost-benefit counterparts, different conclusions were reached within this regression analysis group, with some having found negative impacts (Lester, 2013; Owens & Rennhoff, 2018), while others noted mixed or small positive effects (Adkisson, 2013; Button, 2018, 2019; Swenson, 2017; Thom, 2018a, 2019). Lester (2013) used a benefit-cost analysis of Canada’s MPI program and focused on real income per capita. He found that real incomes in Canada decreased by the equivalent of 96% of the tax revenue forgone by offering credits. Owens and Rennhoff (2018) used a discrete choice model to understand a production’s filming location decision. This approach essentially used propensity score matching on film locations to create comparison groups that were similar to each other on modeled characteristics, except for MPI status. They found that incentives attracted film productions to adopting states but that all states lost revenue on tax credits. Based on an analysis of 29 states, they estimated that states were only able to recoup between 0.03 and 0.77 cents per dollar of tax credit. Button (2018) evaluated efficacy using the Abadie et al. (2010) synthetic control case study method. He compared Louisiana and New Mexico’s MPI programs to their “synthetic” counterparts (i.e., a mix of states with similar characteristics except that they did not adopt a MPI program) using data from IMDb, Studio System, and the Quarterly Census of Employment and Wages (QCEW) between 1988 and 2008. He concluded that MPI programs increased the production of films, but not TV series, employment, or business establishments. Thom (2019) used an interrupted time-series model that created a counterfactual outcome by dividing the time analyzed into a pre-MPI intervention period and a postintervention period. He focused exclusively on five high-spending states—New York, Louisiana, Georgia, Connecticut, and Massachusetts—and used the QCEW data on motion picture industry employment. He found statistically significant gains in immediate motion picture employment in Connecticut and gains over time in Louisiana, though he considered both gains in employment to be modest in scope. In Connecticut, immediate motion picture employment increased 91 percentage points following the program’s introduction, but this was from a relatively small base of 500 workers. In Louisiana, annual employment gains were 11 percentage points. In the other three states, Thom found no statistically significant effects.

Swenson (2017), Thom (2018a), and Button (2019) all assessed the impact of MPI programs using a panel DiD approach that exploited state-level variation in MPI programs. DiD is a widely used econometric technique that acknowledges that the control group is inherently different from the treatment group and builds this difference into the model. In DiD analyses, the difference between control and treatment groups is calculated twice, once before MPI adoption, then again after adoption. This produces a DiD that is used to estimate the causal impact of a MPI program, on average. DiD analyses require a host of assumptions to be credible, including testing for parallel paths, endogeneity in treatment status, and violations of the Stable Unit Treatment Value Assumption (SUTVA). Researchers can employ a variety of controls to meet these assumptions, but estimates are highly sensitive to them. As a result, some DiD estimates are much more credible than others and even the most credible estimates are susceptible to critiques.

Swenson’s model focused on movie production employment between 1998 and 2011, using data from the U.S. Census Bureau’s County Business Patterns (CBP). Swenson concluded that there were no significant increases in direct employment and minor increases in the number of business establishments overall. By contrast, Thom (2018a) used state-level MPI data between 1998 and 2013, with data on motion picture employment and spending from the U.S. Bureau of Economic Analysis, and found that tax credits have heterogeneous effects on entertainment employment and wages based on their type. Specifically, he noted that refundable credits had no effect on employment, but wages increased an average of 4.92 percentage points. States with transferable credits, by contrast, increased annual employment gains by 0.58 percentage points for each year the credit was available but had no effect on wages.

Button (2019) is both the most recent analysis and the most comprehensive in terms of how he tackled potential violations to his DiD model. In the case of MPI programs, the researcher must show that the trend in the outcome variable is the same for both adopting and nonadopting states. This is called the parallel paths assumption. A second assumption is that treatment status—MPI adoption—is exogenous. If, for example, a state’s decision to adopt a MPI program is related to falling state unemployment, the model’s estimates will be biased. Finally, the DiD approach must meet the SUTVA, which assumes that one state’s treatment status does not affect the outcome of interest in other states. In other words, Georgia’s adoption of a MPI program does not affect entertainment activity in other states. Button (2019) tested and employed a variety of controls for parallel paths, endogeneity in treatment status, and violation of the SUTVA. His various model specifications are easily the most robust to account for violations of the DiD key assumptions.

Button (2019) also used state-level MPI data, though his data set extended from 1976 to 2017. Unlike Swenson and Thom, Button linked his MPI data to filming data from IMDb, Studio System, CBP, and the QCEW. He concluded that incentives had a large, though “not entirely robust” effect on TV series filming between 6.4% and 55.4%, translating to 0.67 to 1.50 additional TV series. Button also noted a small but nonrobust effect on feature filming of 13.5% at the upper bound (translating to 0.03 additional feature films) and 18.2% for employment (equivalent to 314 jobs). He found no effect on motion picture establishments or wages. Finally, for states with MPI programs that also have states nearby with MPI programs, Button found evidence of positive spillovers for TV series and feature films. This suggests that the statistically significant results he initially found may be negatively biased.

Button (2019) was thus an interesting motivation for this study. He pointed to the “non-robust” nature of the increased production of TV series and feature films and the small increase in entertainment employment, but his results did find statistically significant effects even though he used an aggregated data set of all states. The existence of positive spillovers may suggest that MPI programs are effective for some states and not others. Maybe states with a history of motion picture production and agglomeration see larger effects, for example. Button explicitly tested for this by decomposing the United States into three groups—small, medium, and large—in terms of average motion picture employment between 1978 and 1985. He found that states with medium and large levels of motion picture employment appeared to drive increased TV series filming, lending support to the idea that MPIs operate differently for some states than others.

Evaluations of California’s MPI Program

Given the historic importance of film to California’s economy, it’s not surprising that there have been four evaluations of California’s Film and Production Tax Credit: a state-mandated evaluation by the Legislative Analyst’s Office (LAO) of California, two studies by the Los Angeles Economic Development Corporation (LAEDC), and an independent academic evaluation conducted by Thom (2018b). After making several adjustments to differentiate new spending from old, the LAO’s evaluation of California’s first MPI program found that $800 million in incentives over the program’s lifetime generated about $4.5 billion in increased motion picture production spending, and would result in reduced general fund revenues of approximately $100 million in California’s 2018/2019 fiscal year before converging toward $0 by fiscal year 2024/2025 (Weatherford, 2016). These adjustments included an estimate that 25% of production spending would have occurred in California even without the incentive by exploiting some experimental data. While the LAO’s evaluation was relatively unique among state evaluations in using estimates of windfall tax benefits, the “net amount of new economic activity is [still] uncertain” (Weatherford, 2016).

In addition to the LAO’s evaluation, the LAEDC conducted two separate evaluations, which found positive ROIs in terms of tax receipts. The first, sponsored by the Headway Project, reviewed by the UCLA Institute for Research on Labor and Employment, and funded in part by the MPAA (Hiltzik, 2011), found that the tax credit returned $1.04 for every tax credit dollar (Appelbaum et al., 2010). The second study, commissioned by the Southern California Association of Governments and conducted by the LAEDC’s Economic and Policy Analysis Group, estimated a larger ROI of $1.11 (Cooper et al., 2010), translating to an increased ROI of nearly 7%. Both studies used IMPLAN analysis, an input-output economic modeling program, to determine the program’s impact that is not causal in nature. Importantly, like the LAO’s analysis, the Appelbaum et al. (2010) study assumed that some production spending (albeit a much lower 8.4%) would have occurred in California even without the incentives. The Cooper et al. (2010) study, however, assumed that 100% of production spending would not have occurred in California without the program.

Finally, Thom’s (2018b) evaluation offered a fourth conclusion on the program’s efficacy. Whereas the LAEDC’s analyses utilized IMPLAN analyses, Thom’s analysis used an interrupted time-series model that created a counterfactual outcome by dividing the time period analyzed into a pre-MPI intervention period (1991-2009) and a postintervention period (2010-2016). Rather than focusing on expenditures, his analysis focused exclusively on motion picture industry employment. Thom found that California’s MPI program had, at best, a very small effect on motion picture industry employment in California. Specifically, he noted that a one percentage point increase in the size of annual film and production tax credits increased occupational employment by 0.09 percentage points.

Motion Picture Production Agglomeration and California

California has long been synonymous with motion picture production. Los Angeles County alone, home to Hollywood, employed an average of 95,277 workers in the motion picture industry in 2018 according to the QCEW, translating to 42% of national motion picture employment. In total, 103,583 motion picture workers were employed in California in 2018 or 45% of national employment in the sector. The heavy concentration of motion picture employment in California (and Los Angeles, specifically) drives and stems from the area’s entertainment agglomeration economies.

The motion picture industry’s tendency for clustering and agglomerating is one of its defining characteristics and has been well documented in the literature (Christopherson, 2008; Florida et al., 2011; Scott, 2002, 2005; Storper & Christopherson, 1987). Marshall (1920) pointed to three key gains of agglomeration: (a) proximity to customers and/or suppliers, (b) labor market pooling, and (c) intellectual spillovers. Marshall’s proximity to customers is less applicable in the motion picture industry, but readers need to look no further than Scott’s (2002, 2004) maps, which showed the clustering of motion picture production companies and television production companies, respectively, in specific parts of Los Angeles to see establishments’ decisions to locate next to other motion picture establishments. The close proximity creates “thicker” markets with more buyers and sellers, reducing costs and increasing scale and specialization (Button, 2018). Labor market pooling is evident from Los Angeles’ dominant share of motion picture employment, and allows labor, like establishments, to further specialize. Finally, in the context of the motion picture industry, Marshall’s spillover effect refers to the transfer of knowledge and ideas that is more likely to occur when workers and establishments are closely knit. This could occur casually—over food and drink at a restaurant—or more formally through workers moving to a new establishment and bringing new knowledge.

In summary, while a significant body of work exists about the impact of tax incentives in general and a growing body of work applies these techniques to MPI programs specifically, the efficacy of tax incentives is still unclear. Researchers have shown that MPI program adoption was far from exogenous, making even the most credible DiD estimates circumspect. Button (2019), the most credible DiD study in the literature, suggested that MPI programs likely operate differently for states with preexisting entertainment clusters. California is the quintessential entertainment cluster and likely benefits from agglomeration gains as a result. This article thus extends Button (2019) by examining the effect of MPIs in California specifically and determining if they achieve different effects in a state with motion picture production agglomeration. Furthermore, this study stands apart from the existing literature by its use of experimental data, potentially offering the most credible estimates of MPI program efficacy to date.

Program Background

California’s Film and Television Tax Credit Program

California’s Film and Television Tax Credit Program was created on February 20, 2009. This version of the program is usually referred to as “1.0” or the “old credit.” Compared with other states, California’s MPI program was modest in scope. While many states chose not to cap incentives and to provide transferable credits, California initially chose to cap its program at $100 million per year and to provide nontransferable credits. Credits were issued once a year (starting on July 1, 2009) and could be used to offset tax liability in California beginning on January 1, 2011. Several types of productions could apply for tax credits, including independent and nonindependent films, movies of the week, miniseries, and 1-hour TV series that were produced for basic cable. Budget caps for nonindependent films and independent films were $75 million and $10 million, respectively. Of the $100 million in annual funding, $10 million was explicitly reserved for independent productions. To apply for a credit, each application required the following: an application form, financial documentation evidencing at least 60% of the financing of the production budget, production budget, shooting schedule, project synopsis, screenplay, and relocation statement, if applicable.

The Lottery

California’s legislature delegated application processing and tax credit allocation duties to the California Film Commission (CFC). The CFC allocated $100 million worth of credits per year on a first-come, first-served basis. If any single fiscal year’s credit was used up, credits could be used from the following year to meet excess demand. If more than one applicant applied for credit on the same day, the CFC was required to hold a lottery.

The CFC’s lottery procedure involved a California Highway Patrol officer drawing queue numbers from a bowl or bingo-style tumbler device. Credits were then offered according to a production’s queue number. Since the CFC was required to offer $10 million in credits to independent productions, the productions offered credits were not always those with the lowest queue numbers. Typically, the cap on the amount of credit offered to nonindependents was met before reaching a minimum of $10 million in independent credits. When this occurred, the CFC offered a credit to the independent production with the lowest queue number.

Fiscal Years Used for Analysis

Table 1 summarizes available data for each application year. Demand for credits did not exceed availability for the program’s first two fiscal years. As a result, a lottery was not used. The CFC used a lottery for California’s fiscal year 2011/2012 (July 2011 to June 2012), but over two separate application days. Fortunately, both a lottery and a single application day were used for fiscal years 2012/2013 through 2016/2017. Since application data were incomplete for fiscal years 2015/2016 and 2016/2017, however, this article’s analysis is restricted to three fiscal years: 2012/2013, 2013/2014, and 2014/2015.

Table 1.

Comparison of Application Periods.

Letter	Application date(s)	Fiscal year	Lottery used	Number of applicants	Number of tax credits issued (final)
Letter	Application date(s)	Fiscal year	Lottery used	Total	Total	Films only
A	July 1-31, 2009; August 1, 2009—December 31, 2009	2009/2010 and 2010/2011	No	^*	67	46
B	June 1, 2010; June 2—December 31, 2010	2011/2012	Yes, but two periods	119	46	21
C	June 1, 2011	2012/2013	Yes	176	46	25
D	June 1, 2012	2013/2014	Yes	322	51	25
E	June 3, 2013	2014/2015	Yes	389	40	19
F	June 2, 2014	2015/2016	Yes	502	32	10
G	April 1, 2015	2016/2017	Yes	246	24	6

Source. Amy Lemisch (2015, August). Progress Report. California Film Commission. Amy Lemisch (2016, October). Progress Report. California Film Commission. Amy Lemisch (2017, December 11). Program 1.0 CFC Approved Projects List. California Film Commission.

The number of applicants was not published for Fiscal Years 2009/2010 and 2010/2011.

Program Data

Data Set

While California’s MPI program extended credits to several types of production, this article focuses on films because they are the most likely to have the supplemental data necessary for analysis. This article’s data set combines data from the LAO, various industry publications, the CFC, and IMDb to construct a data set with (a) films that applied for tax credits, (b) films that initially received credits, and (c) those that ultimately received credits. The LAO’s data, though incomplete, provided information on which films applied for tax credits. Since the tax credit program was random only in the offering of tax credits (not in terms of compliance), articles from various industry publications (e.g., The Hollywood Reporter and Variety) that listed the lottery winners each year were used to determine which productions initially received tax credits. Although these industry publications are likely the best sources for determining which productions initially received credits, inconsistencies may exist between the productions reported in these publications and those that truly were initially offered credits. A list of productions that ultimately received tax credits is publicly available on the CFC’s website. Finally, a wealth of data (if available) were pulled from IMDb for each film. Table 2 provides descriptive statistics of the key variables used for analysis.

Table 2.^a

Descriptive Statistics.

	Mean	Standard deviation
	A. LAO, CFC, and tax credit data
Film is an independent production	0.68	0.47
Estimated tax credit ($ millions)	2.19	2.36
Offered tax credit (initial)	0.06	0.24
Received tax credit (final)	0.12	0.33
	B. IMDb data
Film was made	0.52	0.50
Film made in CA (>74% of filming locations in CA)	0.20	0.40
Number of cast	28.65	41.82
Number of filmmakers	114.08	166.69
Number of cast and filmmakers	142.73	202.06
Inverse hyperbolic sine of number of cast and filmmakers	3.58	2.72
Cast and filmmakers in CA (cast and filmmakers × percentage of filming locations in CA)	67.54	137.70
Inverse hyperbolic sine of cast and filmmakers in CA	1.64	2.68
Number of filming locations	1.91	3.81
Inverse hyperbolic sine of number of filming locations	0.80	1.04
Number of filming locations in CA	1.03	2.73
Percentage of filming locations in CA	0.25	0.41
Budget ($ millions)	4.92	11.90
Inverse hyperbolic sine of budget	5.16	7.69
Budget spent in CA (Budget × percentage of filming locations in CA; $ millions)	2.04	6.08
Inverse hyperbolic sine of budget spent in CA	3.14	6.47
IMDb match has high certainty	0.67	0.47
IMDb match has some certainty	0.05	0.22

Note. The sample includes n = 573 films that applied for a tax credit to the California Film Commission between fiscal years 2012 and 2014. Panel A: Data compiled from the Legislative Analyst’s Office of California, the California Film Commission, and various industry publications. Panel B: Data from IMDb. A film is considered made in CA if 75% or more of listed filming locations were in CA. High certainty matches contain the exact title originally listed on the production’s application for a tax credit or list the exact title under IMDb’s “Also Known As” section. Some certainty matches include films that closely match the title listed on the application or appear likely matches based on the production’s working title.

Information courtesy of IMDb (http://www.imdb.com). Used with permission. Data extracted on 3/13/18.

Note that, since productions could reapply for credits if they lost the lottery, there are n = 573 observations but n = 501 unique titles. Of the 573 observations, 434 applied for a tax credit once, 126 twice, and 12 applied three times. All observations were used in the empirical analysis, but fixed effects were included to account for reentry into the three lotteries (2012, 2013, and 2014). In total, 36 productions were initially offered a tax credit and 69 productions ultimately received a credit. Of the 36 productions initially offered a credit, 14 (39%) complied with treatment.

Percentage of Filming Locations in California

To isolate the effect of key dependent measures on California specifically, filming location data were pulled for each production from IMDb. A percentage of filming locations in California variable was created by calculating the number of filming locations in California and dividing it by the total number of filming locations for each production. Button (2019) used a similar approach, though instead of using the actual percentage of filming, he split production equally between states if more than one state was listed. Both are imperfect measures, as some locations surely represent most of a film’s production, while others represent a small fraction. Ideally, a measure that uses the actual percentage of filming in each location would be used.

Figure 1 shows the distribution of the percentage of filming locations in California for all productions and for productions that received a tax credit. The distribution among all productions indicates that a majority (72%) did no filming in California, a minority (19%) filmed entirely in California, and about 1 in 10 filmed somewhere in between these two extremes. Since California’s program required productions to film 75% of a production or spend 75% of budget in California to receive a credit, Figure 1 also plots the distribution for productions that ultimately received a credit. Among this group, 17% did no filming in California, 57% filmed entirely in California, and 26% filmed between 47% and 94% in California. This suggests that the percentage of filming locations in California variable is a strong predictor of actual filming or budget spent in California.

Figure 1.

Kernel densities of percentage of filming locations in California.

Film Titles

Since titles frequently change throughout the life of a film, the title provided to the CFC in a film’s application was not necessarily its final title. As a result, for all the films in the data, a web search was done to determine the film’s final title and to find its supplemental data on IMDb. Some films were clear matches, some possible matches, and others had no match. To account for these differences, variables were created for each match type. High certainty matches contain the title originally listed on the production’s application for a tax credit or list the title under IMDb’s “Also Known As” section. Some certainty matches include films that closely match the title listed on the application or appear likely matches based on the production’s working title. Sixty-seven percent of films were matched with high certainty, 5% with some certainty, and 28% had no match. As a robustness check, an additional regression is presented in the Robustness Checks section of this article that excludes films with only “some certainty.” The results of this additional regression suggest that the default model’s estimates are lower bounds and are approximately 5% lower than the estimates that exclude some certainty films.

Data Manipulations

Manipulations were made to the number of cast and crew variables (grouped and in California), the number of filming locations (in total), and the budget spent variables (in total and in California). Since many productions were never made and because IMDb does not always list a production’s information, there are a sizable number of 0s in the data set. Specifically, 28% of films have no cast and filmmakers, 54% have no film location information, and 69% have no budgets. Due to the large number of 0s, these variables are not normally distributed. To correct for this nonnormality, additional variables were created that incorporate an inverse hyperbolic sine (IHS) transformation. This approach mirrors Button (2019) and allows log-linear regression interpretation with variables that contain zeros (Bellemare & Wichman, 2020; Burbidge et al., 1988; Mackinnon & Magee, 1990).

Estimation Framework

If all films that were initially offered a tax credit ultimately received one and if all films that were not initially offered a tax credit did not ultimately receive one, this study’s posttreatment outcome of interest (Y) would be causal evidence of the impact of a tax incentive. There were two forms of noncompliance in this study, however. First, some films that were offered a credit ultimately did not receive one. Second, some films that were not offered a credit did ultimately receive one. Following Abdulkadiroğlu et al. (2011), this study exploits two distinct measurement strategies to account for these two forms of noncompliance. The first is a measure of the ITT effect and the second involves the use of an instrumental variable to measure the ToT. The former measures the effect of offering a film a tax credit, regardless of whether it ultimately received a tax credit, and the latter measures the effect of actually receiving a tax credit by using whether a film was initially offered a tax credit as an instrument for whether it ultimately received one.

Intention to Treat Effect

A cohort is defined as a group of films that applied for a particular fiscal year’s lottery, and a film’s type is defined as either independent or nonindependent. The posttreatment outcome of interest ( $Y_{i f c}$ ) is given for each film i, film type f, and cohort c.

Y_{i f c} = α_{c} \times α_{f} + δ T a x C r e d i t_{i f c} + ε_{i f c}

In this equation, $α_{c}$ and $α_{f}$ indicate fixed effects for the three cohorts (lotteries) and the two film types, respectively. $T a x C r e d i t_{i f c}$ signifies whether a film was offered a tax credit for either film type f during lottery c, and $ε_{i f c}$ is the residual for film i, film type f, and cohort c. Note that the interaction of the two fixed effects produces six groups in total. The coefficient $δ$ is the effect of being offered a tax credit on the posttreatment outcome of interest.

Treatment Effect on the Treated

Since the initial assignment of tax credits was random, $Y_{i f c}$ is unbiased in the absence of random chance. Final assignment of tax credits was not random, however, as productions ultimately were not forced to comply with the tax credit program’s requirements and, in turn, receive a tax credit. To factor in these differing compliance rates, I use whether a film was initially offered a tax credit as instrumental variable Z to predict whether the film ultimately received a tax credit (complied) D using a two-stage least squares (2SLS) regression:

D_{i f c} = α_{c} + γ Z_{i c} + μ_{i f c}

With predicted compliance $({\hat{D}}_{i f c})$ in hand, I proceed to the second stage using the same equation for the ITT effect, except that $T a x C r e d i t_{i f c}$ is now replaced by predicted compliance ${\hat{D}}_{i f c}$ :

Y_{i f c} = α_{c} \times α_{f} + δ_{D} {\hat{D}}_{i f c} + ε_{i f c}

Randomization Check and Balance Test

Two things must be true to use the estimation framework outlined above. First, the lottery used by the CFC must be random and, second, films that were offered a tax credit must not be statistically different from those that were not offered one. To test that the lottery was random, an OLS regression was run to predict the estimated tax credit a film would receive (the only pretreatment variable available) using the production’s fiscal year queue number. Based on the regression, estimated tax credit size does not significantly change based on fiscal year queue number. Specifically, a one-unit increase in fiscal year queue number is associated with a .04% increase in estimated tax credit (standard error of 0.000713) and is not significant at the 10% level. As a result, the lottery appears to be random. To test that films offered a tax credit were not statistically different from those that were not offered a tax credit, the ITT regression from the estimation framework was run to predict estimated tax credit size. Here, there is no statistically significant difference in the size of the estimated tax credit between films offered a tax credit and films not offered one. Specifically, films that were offered a tax credit are associated with a 6.3% increase (standard error of 0.142) in estimated tax credit size and is not significant at the 10% level. The results of these tests, combined with the evidence on how the lottery was done, suggest that the lottery was truly randomized, minimizing bias in my estimates.

First-Stage Effect of Offering a Tax Credit on Receiving a Tax Credit

First-stage estimates of the effect of being offered a tax credit on ultimately receiving a tax credit are shown in Table 3. An initial offer of a tax credit is a powerful predictor of ultimately receiving one—productions initially offered a tax credit were 27 percentage points more likely to ultimately receive one. Additionally, the coefficient is statistically significant at the 99% level and the F statistic is 24.80, well above the minimum threshold of 10 typically used. Furthermore, these results hold whether a 2SLS regression or a limited information maximum likelihood is used.

Table 3.^a

First Stage Regression.

Regression number	(1)
Dependent variable	Received tax credit (final)
Offered tax credit (initial)	0.274*** (0.0550)
Constant	0.103*** (0.0137)
Observations	573
R ²	.042
F-statistic	24.8
Number of groups	6

Note. Standard errors in parentheses.

Results are robust to two-stage least squares (2SLS) and limited information maximum likelihood (LIML).

p < .1. ^**p < .05. ^***p < .01.

Main Empirical Results

Intention to Treat Effect

Although the primary intent of this analysis is to uncover how being offered a tax credit affects key indicators of film production as they apply in California, it is insightful to begin with an overview of a tax credit’s effect, in general. Table 4 presents four key indicators: whether the film was made, production budget (IHS), number of cast and filmmakers (IHS), and number of filming locations (IHS). Offering a tax credit increased the probability of a film being made by 22 percentage points, statistically significant at the 5% level. The coefficient on the production budget (IHS) variable is 2.61 and is also statistically significant at the 5% level. This is an average increase in production budget after being offered a tax credit of 254%. The coefficient on the number of cast and filmmakers (IHS) represents an average increase in cast and filmmakers of 109% and is similarly statistically significant at the 5% level. The offering of a tax credit had no statistically significant effect on the number of filming locations.

Table 4.^a

Intention to Treat Effect, Overall.

Regression number	(1)	(2)	(3)	(4)
Dependent variable	Film was made	Budget (inverse hyperbolic sine)	Cast and filmmakers (inverse hyperbolic sine)	Percentage of filming locations (inverse hyperbolic sine)
Offered tax credit (initial)	0.215** (0.0859)	2.606** (1.297)	1.129** (0.459)	0.225 (0.176)
Constant	0.507*** (0.0214)	4.998*** (0.322)	3.510*** (0.114)	0.790*** (0.0437)
Observations	573	573	573	573
R ²	.011	.007	.011	.003
Number of groups	6	6	6	6

Note. Standard errors in parentheses.

Linear probability model regression (Film was made) is robust to probit model, remaining statistically significant at the 5% level.

p < 0.1. ^**p < .05. ^***p < .01.

Next, I reapply the ITT framework to the same key indicators except that I multiply each by the proportion of filming activity that was done in California to understand the effect of California’s tax credit, specifically. The results are presented in Table 5 for the following key California indicators: whether the film was made in California, budget spent in California, number of cast and filmmakers in California, and percentage of filming locations in California. The probability of a film being made in California increased by 16 percentage points simply by offering a production a tax credit. Notably, the constant term from column 1 reveals that 19% of films were made in the state even without an incentive. This figure is similar to the LAO’s 25% estimate but is considerably higher than the LAEDC’s estimates of 8.4% and 0%, respectively. Columns 2 and 3 show that the budget spent increased by an average of 267% and the number of cast and filmmakers increased by an average of 123%. Additionally, the offering of a tax credit increased the percentage of filming locations in California by 15 percentage points. All four coefficients are statistically significant at the 5% level.

Table 5.^a

Intention to Treat Effect, California.

Regression number	(1)	(2)	(3)	(4)
Dependent variable	Made in CA	Budget spent in CA (inverse hyperbolic sine)	Cast and filmmakers in CA (inverse hyperbolic sine)	Percentage of filming locations in CA (inverse hyperbolic sine)
Offered tax credit (initial)	0.159** (0.0692)	2.615** (1.104)	1.171** (0.460)	0.149** (0.0709)
Constant	0.191*** (0.0172)	2.977*** (0.274)	1.566*** (0.114)	0.236*** (0.0176)
Observations	573	573	573	573
R ²	.009	.010	.011	.008
Number of groups	6	6	6	6

Note. Standard errors in parentheses.

Linear probability model regression (Film was made in CA) is robust to probit model, remaining statistically significant at the 5% level.

p < .1. ^**p < .05. ^***p < .01.

Treatment Effect on the Treated

The ToT framework is incorporated to uncover what effect, if any, stems from a film receiving (complying with) a tax credit. The general results are presented in Table 6 and the California-specific results in Table 7. In contrast to the ITT regressions, the magnitude of the coefficients are now several times larger. For example, when a film received a tax credit, the probability of its being made increased by 79 percentage points, its production budget increased by 920%, and the number of cast and filmmakers increased by 462%. A film being made and the number of cast and filmmakers are statistically significant at the 1% level, production budget at the 5% level, and the number of filming locations remains insignificant.

Table 6.

Treatment Effect on the Treated, Overall.

Regression number	(1)	(2)	(3)	(4)
Dependent variable	Film was made	Budget (inverse hyperbolic sine)	Cast and filmmakers (inverse hyperbolic sine)	Percentage of filming locations (inverse hyperbolic sine)
Received tax credit (final) [Offered tax credit (initial) as instrument]	0.785*** (0.303)	9.516** (4.632)	4.121*** (1.593)	0.822 (0.617)
Constant	0.426*** (0.0415)	4.016*** (0.636)	3.085*** (0.219)	0.705*** (0.0846)
Observations	573	573	573	573
Number of groups	6	6	6	6

Note. Standard errors in parentheses.

p < .1. ^**p < .05. ^***p < .01.

Table 7.

Treatment Effect on the Treated, California.

Regression number	(1)	(2)	(3)	(4)
Dependent variable	Made in CA	Budget spent in CA (inverse hyperbolic sine)	Cast and filmmakers in CA (inverse hyperbolic sine)	Percentage of filming locations in CA (inverse hyperbolic sine)
Received tax credit (final) [offered tax credit (initial) as instrument]	0.580** (0.227)	9.549*** (3.697)	4.275*** (1.477)	0.544** (0.231)
Constant	0.131*** (0.0311)	1.991*** (0.507)	1.125*** (0.203)	0.180*** (0.0318)
Observations	573	573	573	573
Number of groups	6	6	6	6

Note. Standard errors in parentheses.

p < .1. ^**p < .05. ^***p < .01.

Turning to the California-specific results in Table 7, the magnitude of the coefficient is also much larger. Now, the probability of a film being made in California increases by 58 percentage points. Films that received a tax credit also increased budget spent in California by 966% and increased the number of cast and filmmakers in the state by 388%. Additionally, receiving a tax credit increased the percentage of filming locations in California by 54 percentage points. Production budget and the number of cast and filmmakers are statistically significant at the 1% level, and a film being made and the number of filming locations are statistically significant at the 5% level.

Subgroup Analysis

In both measurement strategies, tax credits have pronounced effects. A secondary, but perhaps equally important question, is whether tax credits are equally as effective for different film types. Tables 8 and 9 present the ITT results in general and for California, respectively, for independent films (Panel A) and nonindependent films (Panel B).

Table 8.

Intention to Treat Effect, Independent Versus Nonindependent Films, Overall.

Regression number	(1)	(2)	(3)	(4)
Dependent variable	Film was made	Budget (inverse hyperbolic sine)	Cast and filmmakers (inverse hyperbolic sine)	Percentage of filming locations (inverse hyperbolic sine)
	A. Independent films
Offered tax credit (initial)	0.202** (0.0930)	1.822 (1.301)	1.134** (0.497)	0.0687 (0.171)
Constant	0.489*** (0.0261)	3.950*** (0.365)	3.122*** (0.140)	0.643*** (0.0479)
Observations	390	390	390	390
R ²	.012	.005	.013	.000
Number of groups	3	3	3	3
	B. Nonindependent films
Offered tax credit (initial)	0.292 (0.226)	7.207* (3.872)	1.097 (1.200)	1.144** (0.546)
Constant	0.544*** (0.0371)	7.239*** (0.638)	4.339*** (0.198)	1.106*** (0.0898)
Observations	183	183	183	183
R ²	.009	.019	.005	.024
Number of groups	3	3	3	3

Note. Standard errors in parentheses.

p < .1. ^**p< .05. ^***p < .01.

Table 9.

Intention to Treat Effect, Independent Versus Nonindependent Films, California.

Regression number	(1)	(2)	(3)	(4)
Dependent variable	Made in CA	Budget spent in CA (inverse hyperbolic sine)	Cast and filmmakers in CA (inverse hyperbolic sine)	Percentage of filming locations in CA (inverse hyperbolic sine)
	A. Independent films
Offered tax credit (initial)	0.145^* (0.0762)	1.422 (1.093)	0.833* (0.479)	0.121 (0.0780)
Constant	0.199*** (0.0214)	2.455*** (0.307)	1.446*** (0.135)	0.237*** (0.0219)
Observations	390	390	390	390
R ²	.009	.004	.008	.006
Number of groups	3	3	3	3
	B. Nonindependent films
Offered tax credit (initial)	0.242 (0.175)	9.616*** (3.321)	3.153** (1.295)	0.315* (0.179)
Constant	0.174*** (0.0288)	4.100*** (0.547)	1.824*** (0.213)	0.234*** (0.0295)
Observations	183	183	183	183
R ²	.011	.045	.032	.017
Number of groups	3	3	3	3

Note. Standard errors in parentheses.

p < .1. ^**p < .05. ^***p < .01.

Compared with nonindependent films, only independent films are statistically more likely to be made. This appears driven primarily by the smaller sample size of nonindependent films (n = 183) compared with independent films (n = 390). The coefficient for nonindependent films, 0.292, is nearly 50% larger than the coefficient for independent films, but the standard error is much higher for nonindependent films. Despite the smaller sample size for nonindependents, production budget is significantly higher (at the 10% level) only for nonindependents, and the magnitude of the coefficient is four times larger than independents. For nonindependent films, the offering of a tax credit increased production budget by 684%. Interestingly, the magnitude of the coefficient for the number of cast and filmmakers (IHS) variable is similar between the two groups. Again, due to the larger standard error for nonindependents, the coefficient is only statistically significant (at the 5% level) for independents. For independent films, the offering of a tax credit increased the number of cast and filmmakers by 111%. Finally, the coefficient for number of filming locations is only statistically significant (at the 5% level) for nonindependents and is much higher for nonindependents. The offering of a tax credit increased the number of filming locations by 114% for nonindependent films.

Turning to the California-specific results, a similar pattern emerges. Only independent films are statistically more likely to be made in California, specifically 15 percentage points, and the coefficient for nonindependent films is not statistically significant due to the larger standard error. Budget spent in California (IHS) is only significant for nonindependents and the coefficient is more than six times larger. For nonindependents, the offering of a tax credit increased budget spent in California by 991%. The number of cast and filmmakers is now statistically significant for both film types, but the magnitude of the coefficient is nearly four times larger for nonindependents. The offering of a tax credit increased the number of cast and filmmakers by 341% for nonindependents and by 130% for independents. Finally, the percentage of filming locations in California remains statistically significant only for nonindependents and is more than twice the magnitude of independents. The offering of a tax credit increased the percentage of filming locations in California by 32% for nonindependents.

In summary, while the results are somewhat imprecise due to the smaller sample size for nonindependents, tax credits appear to have much stronger effects in California for nonindependent films than for independent films. In California, budget spent, number of cast and filmmakers, and the percentage of filming locations are only statistically significant for nonindependent films despite this group’s smaller sample size. In contrast, only a film being made and the number of cast and filmmakers in California are statistically significant for independents. Statistical significance aside, the magnitude of the coefficients for nonindependent films is 1.7, 6.8, 3.8, and 2.6 times, respectively, larger than the coefficients for independent films. Given that nonindependents tend to spend more on productions and hire more cast and filmmakers, in general, these are particularly large effects.

Robustness Checks

As discussed in the Program Data section, 29 films (5%) were matched to IMDb profiles with some certainty. A more conservative estimate would exclude these films and treat them as if they had not been made. With the exception of whether a film was made, the magnitude and the statistical significance of all coefficients are unchanged. After the correction, the offering of a tax credit increases the probability of a film being made by 24 percentage points (vs. 22 points before). The coefficient is also now statistically significant at the 1% level (vs. 5% previously). Turning to the California-specific indicators, the magnitude of the coefficients for films made in California, budget spent in California (IHS), the number of cast and filmmakers in California (IHS), and the percentage of filming locations in the state increase by 6%, 2%, 6%, and 7%, respectively. Furthermore, the number of cast and filmmakers in California (IHS) is now statistically significant at the 1% level (vs. 5% previously).

Discussion

Cost-Benefit Analysis

In its simplest form, the net benefit of California’s Film and Television Tax Credit Program is the total additional spending in the state that occurred less the cost of providing the incentives. Among films not offered an incentive, the average California budget was $1,991,859. Since films offered an incentive increased their California budget by 267%, additional spending in California per film is estimated to be $3,326,405, on average. There were 36 films that were initially offered a tax credit, so aggregate additional spending is $119,750,563. The estimated tax credit size for these 36 films totaled $53,759,291.² Thus, the net benefit of California’s Film and Television Tax Credit Program is $65,991,272, which translates to a 1.23 ROI.

Based on this ITT estimate, for every dollar that California provided to a film production, a $1.23 was spent in California on film production. Unfortunately, without film-level data on tax expenditures, this study cannot estimate whether Program 1.0 was a net gain or a net loss to Californian taxpayers. Using new spending as a result of incentives and the cost of tax incentives, however, we can estimate how much the state of California would need to recoup to break even. We know that $119,750,563 in new spending occurred due to the offering of incentives at a cost of $53,759,291. This means that California would have to recoup 45% of new spending to pay for the program’s incentives. This estimate implies an extremely high level of recovery, suggesting that the MPI program is likely a net cost to California taxpayers and that MPIs do not pay for themselves, even in California.

Implications for Policy

In January 2015, legislators approved significant changes to California’s Film and Television Tax Credit Program, creating Program 2.0. These changes included increased annual funding from $100 million to $330 million; quotas on the amount of tax credits given to each production type; removal of the budget cap for feature films, although qualified expenditures are limited to $100 million for nonindependent films and $10 million for independent films; created multiple application periods; and replaced the use of a lottery with a jobs ratio ranking to assign credits. Per the CFC, the jobs ratio is primarily determined by calculating qualified wages and dividing by the amount of tax credits to be allocated. Other factors can also be included, such as qualified spending for vendors and equipment. Furthermore, the base score can increase by up to 25% for in-state spending on visual effects, filming outside the Los Angeles 30-mile zone, and filming at approved production facilities.

In June 2018, the program was further revised and dubbed Program 3.0. Much of the previous program’s structure remains intact under Program 3.0, though several minor changes were implemented, including the creation of a training program for underserved communities, a reduced credit for relocating TV shows, an increased credit for wages to individuals outside of Los Angeles’ 30-mile zone, the elimination of sound stage from bonus point consideration, the removal of the additional 5% credit for music scoring wages, and an extension of the date that principal photography must begin from 180 days to 240 days for productions with budgets over $100 million in qualified spending.

While the replacement of the lottery with the jobs ratio likely increases the economic benefit of the tax credit program to California, it means that California loses the ability to say with certainty that economic benefits increased after implementing the changes. Given the results of this study, legislators’ decision to remove the budget cap for nonindependent films appears wise. It seems plausible that, without the budget cap in place, more nonindependent films would seek tax credits. Given that nonindependent films were the only film type to spend significantly more budget and hire more cast and filmmakers in California when offered a tax credit, this should pay extra economic dividends for the state.

Conclusion

MPI programs now consume more than $3 billion of public subsidies, but their efficacy as an economic development tool has been inconclusive. Though positive ROIs have been noted in studies partly funded by the industry, many cost-benefit studies have found negative ROIs. On the academic side, some studies found negative impacts, while others note mixed or small positive effects.

In addition to adding experimental evidence to the efficacy of MPI programs, this article applies it in the context of California, a state with substantial agglomeration economies in entertainment production, to reveal novel causal estimates of MPI efficacy. I find that nearly one-in-five productions would have filmed in California even without an incentive. This is slightly lower than the LAO’s estimate of 25% and much higher than the LAEDC’s 8.4%, though both estimates include films and television shows. Furthermore, my estimate is significantly lower than the 75% “but for” in terms of economic development incentives that Bartik (2018) calculated and likely speaks to the footloose nature of film production.

I also find that simply offering a tax incentive to a film increased the probability that it was made in California by 16 percentage points, increased production budget spent in California by 267%, and led to 123% more cast and filmmakers hired in the state. These estimates imply an ROI of $1.23, far higher than the $0.07 to $0.47 range in the current literature but below the $1.50 and $1.90 ROI estimated by the MPAA-funded studies in New Mexico and New York. The $1.23 ROI is also surprisingly larger than the LAECD’s two California-specific studies, which calculated ROI of $1.04 and $1.11. This larger ROI is likely due to this article’s unique experimental data, its focus on films rather than films and television shows, and the fact that the MPI program in question is in California. Generalizing California’s ROI to other states, particularly those without a history of entertainment production, could prove to be a poor comparison. Note that, even though California obtained a positive ROI, that does not mean that tax incentives paid for themselves. My estimates imply that the state of California would need to recoup 45% of new spending in order pay for the incentives.

Notably, in California MPIs appear to have much stronger effects for nonindependent films than for independent films. Budget spent, number of cast and filmmakers, and the percentage of filming locations in California only statistically increased for nonindependent films and the magnitude of these coefficients was several times larger for nonindependents. Since only nonindependent films with production budgets less than $75 million could apply for California’s MPIs, it could be the case that even larger effects exist for blockbuster or tent-pole films.

Future research should seek to estimate the effect of tax incentives on other forms of production, like television production, as well as the impact of tax incentives outside of the state of California. Given the vast sums of public investment poured into MPI programs across the United States, additional research is needed to understand how much of the difference between California’s experience with tax incentives and other states’ experiences is due to California’s agglomeration in film production versus tax incentives alone.

Supplemental Material

sj-pdf-1-edq-10.1177_08912424211000127 – Supplemental material for Ready for a Close-Up: The Effect of Tax Incentives on Film Production in California

Supplemental material, sj-pdf-1-edq-10.1177_08912424211000127 for Ready for a Close-Up: The Effect of Tax Incentives on Film Production in California by Alec Workman in Economic Development Quarterly

Footnotes

Acknowledgements

I would like to thank Craig McIntosh, Jennifer Burney, Gordon Hanson, and anonymous referees for their invaluable comments and suggestions. I thank Brian Weatherford for sharing the raw data that underpin this article.

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.

ORCID iD

Alec Workman

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biography

Alec Workman is the founder of WM Analytics. His background combines a wide range of research interests, including political campaigning, market research and consulting, and finance. Alec currently focuses on key issues confronting the entertainment industry, such as the impact of new distribution strategies, tax incentives, and piracy.

References

Abadie

Diamond

Hainmueller

(2010). Synthetic control methods for comparative case studies: Estimating the effect of California’s tobacco control program. Journal of the American Statistical Association, 105(490), 493-505. https://doi.org/10.1198/jasa.2009.ap08746

Abdulkadiroğlu

Angrist

J. D.

Dynarski

S. M.

Kane

T. J.

Pathak

P. A.

(2011). Accountability and flexibility in public schools: Evidence from Boston’s charters and pilots. Quarterly Journal of Economics, 126(2), 699-748. https://doi.org/10.1093/qje/qjr017

Adkisson

R. V.

(2013). Policy convergence, state film-production incentives, and employment: A brief case study. Journal of Economic Issues, 47(2), 445-454. https://doi.org/10.2753/JEI0021-3624460218

Appelbaum

L. D.

Tilly

Huang

(2010). Economic and production impacts of the 2009 California Film and Television Tax Credit. UCLA Institute for Research on Labor and Employment. https://escholarship.org/uc/item/0qh4h6jw

Arauzo-Carod

J. M.

Liviano-Solis

Manjón-Antolín

(2010). Empirical studies in industrial location: An assessment of their methods and results. Journal of Regional Science, 50(3), 685-711. https://doi.org/10.1111/j.14679787.2009.00625.x

Bartik

(1991). Who benefits from state and local economic development policy? W.E. Upjohn Institute for Employment Research Press.

Bartik

T. J.

(2018). “But for” percentages for economic development incentives: What percentage estimates are plausible based on the research literature? (Upjohn Institute Working Paper 18-289). W.E. Upjohn Institute for Employment Research. https://doi.org/10.17848/wp18-289

Bartik

T. J.

(2019). Making sense of incentives: Taming business incentives to promote prosperity. W.E. Upjohn Institute for Employment Research.

Bellemare

M. F.

Wichman

C. J.

(2020). Elasticities and the inverse hyperbolic sine transformation. Oxford Bulletin of Economics and Statistics, 82(1), 50-61. https://doi.org/10.1111/obes.12325

10.

Burbidge

J. B.

Magee

Robb

A. L.

(1988). Alternative transformations to handle extreme values of the dependent variable. Journal of the American Statistical Association, 83(401), 123-127. https://doi.org/10.1080/01621459.1988.10478575

11.

Burnett

(2011). State business incentives: Trends and options for the future. The Council of State Governments. http://knowledgecenter.csg.org/kc/system/files/StateBusinessIncentives2011.pdf

12.

Buss

T. F.

(2001). The effect of state tax incentives on economic growth and firm location decisions: An overview of the literature. Economic Development Quarterly, 15(1), 90-105. https://doi.org/10.1177/089124240101500108

13.

Button

(2018). Can tax incentives create a local film industry? Evidence from Louisiana and New Mexico. Journal of Urban Affairs. Advance online publication. https://doi.org/10.1080/07352166.2018.1530570

14.

Button

(2019). Do tax incentives affect business location and economic development? Evidence from state film incentives. Regional Science and Urban Economics, 77(July), 315-339. https://doi.org/10.1016/j.regsciurbeco.2019.06.002

15.

Christopherson

(2008). Labor: The effects of media concentration on the film and television workforce. In McDonald

Wasko

(Eds.), The contemporary Hollywood film industry (pp. 155-166). Blackwell.

16.

Christopherson

Rightor

(2010). The creative economy as “big business”: Evaluating State strategies to lure filmmakers. Journal of Planning Education and Research, 29(3), 336-352. https://doi.org/10.1177/0739456X09354381

17.

Cooper

Sedgwick

S. M.

Mitra

(2010). California’s Film and Television Tax Credit Program: Assessing its impact. Southern California Association of Governments. https://laedc.org/wp-content/uploads/2014/03/SCAGFilmReport_FINAL.pdf

18.

Ernst & Young. (2009a). Economic and fiscal impacts of the New Mexico tax credit. New Mexico State Film Office.

19.

Ernst & Young. (2009b). Estimated impacts of the New York State film credit. Motion Picture Association of America.

20.

Florida

Mellander

Stolarick

(2011). Geographies of scope: An empirical analysis of entertainment, 1970-2000. Journal of Economic Geography, 12(1), 1-22. https://doi.org/10.1093/jeg/lbq056

21.

Hanson

Rohlin

(2013). Do spatially targeted redevelopment programs spillover? Regional Science and Urban Economics, 43(1), 86-100. https://doi.org/10.1016/j.regsciurbeco.2012.05.002

22.

Hiltzik

(2011, July 20). Upon review, shine comes off glowing report on film tax credit. Los Angeles Times. https://www.latimes.com/business/la-xpm-2011-jul-20-la-fi-hiltzik-20110720-story.html

23.

Lee

(2008). Geographic redistribution of US manufacturing and the role of state development policy. Journal of Urban Economics, 64(2), 436-450. https://doi.org/10.1016/j.jue.2008.04.001

24.

Leiser

(2017). The diffusion of state film incentives: A mixed-methods case study. Economic Development Quarterly, 31(3), 255-267. https://doi.org/10.1177/0891242417710715

25.

Lemisch

(2015). Progress report. California Film Commission.

26.

Lemisch

(2016). Progress report. California Film Commission.

27.

Lemisch

(2017). Program 1.0 CFC Approved Projects List. California Film Commission.

28.

Lester

(2013). Tax credits for foreign location shooting of films: No net benefit for Canada. Canadian Public Policy, 39(3), 451-472. https://doi.org/10.3138/CPP.39.3.451

29.

Mackinnon

J. G.

Magee

(1990). Transforming the dependent variable in regression models. International Economic Review, 31(2), 315-339. https://doi.org/10.2307/2526842

30.

Marshall

(1920). Principles of economics. MacMillan.

31.

McDonald

(2011). Down the rabbit hole: The madness of state film incentives as a solution to runaway production. University of Pennsylvania Journal of Business Law, 14(1), 101-181.

32.

Noll

R. G.

Zimbalist

(1997). Sports, jobs, and taxes. Brookings Institution.

33.

Owens

M. F.

Rennhoff

A. D.

(2018). Motion picture production incentives and filming location decisions: A discrete choice approach. Journal of Economic Geography, 20(3), 679-709. https://doi.org/10.1093/jeg/lby054

34.

Peters

Fisher

(2004). The failures of economic development incentives. Journal of the American Planning Association, 70(1), 27-37. https://doi.org/10.1080/01944360408976336

35.

Prillaman

S. A.

Meier

K. J.

(2014). Taxes, incentives, and economic growth: Assessing the impact of pro-business taxes on U.S. state economies. Journal of Politics, 76(2), 364-379. https://doi.org/10.1017/S0022381613001345

36.

Scott

A. J.

(2002). A new map of Hollywood: The production and distribution of American motion pictures. Regional Studies, 36(9), 957-975. https://doi.org/10.1080/0034340022000022215

37.

Scott

A. J.

(2004). The other Hollywood: The organizational and geographic bases of television program production. Media, Culture & Society, 26(2), 183-205. https://doi.org/10.1177/0163443704039498

38.

Scott

A. J.

(2005). Hollywood: The place, the industry. Princeton University Press.

39.

Storper

Christopherson

(1987). Flexible specialization and regional industrial agglomerations: The case of the U.S. motion picture industry. Annals of the Association of American Geographers, 77(1), 104-117. https://doi.org/10.1111/j.1467-8306.1987.tb00148.x

40.

Strauss-Kahn

Vives

(2009). Why and where do headquarters move? Regional Science and Urban Economics, 39(2), 168-186. https://doi.org/10.1016/j.regsciurbeco.2008.07.001

41.

Swenson

C. W.

(2017). Preliminary evidence on film production and state incentives. Economic Development Quarterly, 31(1), 65-80. https://doi.org/10.1177/0891242416674807

42.

Tannenwald

(2010). State film subsidies: Not much bang for too many bucks. Center on Budget and Policy Priorities. https://arev.assembly.ca.gov/sites/arev.assembly.ca.gov/files/hearings/CBPP_Report.pdf

43.

Taylor

C. D.

(2012). Conservation easement tax credits: A viable policy for stimulating donations? State and Local Government Review, 44(2), 101-112. https://doi.org/10.1177/0160323X12445392

44.

Thom

(2018a). Lights, camera, but no action? Tax and economic development lessons from state motion picture incentive programs. American Review of Public Administration, 48(1), 33-51. https://doi.org/10.1177/0275074016651958

45.

Thom

(2018b). Time to yell “cut?” An evaluation of the California Film and Production Tax Credit for the motion picture industry. California Journal of Politics and Policy, 10(1). https://dx-doi-org.web.bisu.edu.cn/10.5070/P2cjpp10138993

46.

Thom

(2019). Do state corporate tax incentives create jobs? Quasi-experimental evidence from the entertainment industry. State and Local Government Review, 51(2), 92-103. https://doi.org/10.1177/0160323X19877232

47.

Thom

(2017). Fade to black? Exploring policy enactment and termination 12 through the rise and fall of state tax incentives for the motion picture industry. American Politics Research, 45(1), 85-108. https://doi.org/10.1177/1532673X16661819

48.

Wasylenko

M. J.

(1997). Taxation and economic development: The state of the economic literature. New England Economic Review, 1997(March), 37-52. https://surface.syr.edu/cgi/viewcontent.cgi?article=1001&context=ecn

49.

Weatherford

(2016). California’s first film tax credit program. Legislative Analyst’s Office. https://lao.ca.gov/Publications/Report/3502

50.

Wilson

D. J.

(2009). Beggar thy neighbor? The in-state, out-of-state, and aggregate effects of R &; D tax credits. Review of Economics and Statistics, 91(2), 431-36. https://doi.org/10.1162/rest.91.2.431

51.

(2008). State R and D tax credits and high technology. Economic Development Quarterly, 22(2), 136-148. https://doi.org/10.1177/0891242408316728

52.

Zhang

Chen

(2018). The effect of investment tax incentives: Evidence from China’s value-added tax reform. International Tax and Public Finance, 25(4), 913-945. https://doi.org/10.1007/s10797-017-9475-y

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.15 MB