Saturday, January 25, 2014

Economic and Jobs Impact of HJR-3, Indiana Constitutional Amendment Banning Same-Sex Marriage

PDF OF THIS RESEARCH--Revised as of 2/4/82014. The new numbers look slightly different from what is presented below

Introduction

In Indiana, and the United States as a whole, the rights of sexual minorities, such as gays/lesbians, asexuals, and transgenders, continue to be debated by legislatures and courts. In Indiana, January, 2014, the bill HJR-3 purports to reinforce Indiana’s current definition of “marriage” as solely between one man and one woman by amending the state constitution to prevent future changes to this definition. Additionally, the proposed amendment is believed by many to limit alternative forms of “marriage” such as domestic partnerships and civil unions. These limitations have been demonstrated in Michigan, Kentucky and Ohio because of similar constitutional amendments, as testified in the January Indiana Assembly hearings on HJR-3.

One of the arguments against the proposed amendments is an economic assertion by Cummins, Lilly, the Indy Chamber of Commerce, and others, that this amendment will make it more difficult for them to recruit candidates into the state. Governor Pence, and the state GOP committee, prior to the last state-wide election in 2012, explicitly dropped the issue of same-sex marriage from their state platform, saying that their focus would be on jobs creation and attracting younger voters to the party. The rationale for the latter is found in all of the polling data that suggests that voters of both parties under 40 years of age are overwhelmingly opposed to marriage discrimination laws, including voters here in Indiana. Testifying at the Indiana Assembly January hearings, were a number of active GOP members, predominantly young people (under 40 years old) who were vigorously opposed to HJR-3, arguing that it violates traditional GOP principles of liberty, and affirming previous testimony, that it would impede the process of attracting young professionals to Indiana.

This study looks at 9 Midwestern states from 2003-2012, and finds that the testimony of business representatives is supported by time-series analysis—that for every anti-gay legislation passed by a state, and supported by court rulings, there is a subsequent loss of professional firms in that state, and negative impact to the state economy, measured by real GDP per capita. These effects are statistically significant on a number of different firms and jobs to the p<0.01 level (99% confidence). Further, not only raw job numbers are impacted, but also annualized rates of growth are impacted, as anti-gay legislation seems to slow the rate of growth of firms and GDP in subsequent years. This study documents that these changes occur at lags of 1-4 years, i.e., the loss of firms and jobs begins to occur the year after anti-gay legislation is enacted (lag=1), and causes increasing economic damage at least as of 4 years afterwards (lag=4).

Theoretically, these effects are supported by urban studies research that indicates that young professionals who move to cities from other areas, or who remain in their cities of origin, often do so because those cities offer a climate of diversity and tolerance. For example, Richard Florida documents these effects in Cities and the Creative Class (2005, Routledge Press), indicating that cities where public policies support racial equality, equality of sexual minorities, and the arts, are growing at faster rates than cities that lag behind on these cultural trends, and that policies of diversity precede and speed growth.

Methods

This study uses time-series analysis of publicly-available data for 9 Midwestern states. Based on claims by several large employers and chambers of commerce in Indiana that anti-gay legislation has a negative impact on candidate recruitment, data was collected from the U.S. Bureau of Economic Analysis, the American Community Survey (U.S. Census), and the County Business Patterns (U.S. Census) to obtain information about firms by sector, and economic growth, in terms of real GDP per capita. Documentation about state-by-state legislation and court cases relating to same-sex marriage or domestic partnerships/civil unions, was obtained from various journalistic, organizational, and government online sources. I specifically looked at firms and GDP, as well as narrowly focused state-wide legislation and court cases relating to marriage or domestic partnerships/civil unions, and did not factor in other types of sexual minority protections, such as adoption law, employment protections, city-specific protections, transgender protections, etc.

Given that each region of the country faces different types of economic influences, this study focuses only on Midwestern/Great Lakes states. While the Census counts states farther west (Kansas, Nebraska, the Dakotas) as Midwestern, they were excluded from this study to focus on states closest to Indiana. States such as Iowa, Minnesota, and Missouri, while not adjacent to Indiana, are close enough to share many of the same economic and cultural variables to warrant inclusion. While arguably, Pennsylvania, West Virginia and Tennessee are similar in distance to Indiana compared to Minnesota, other geographical factors arguably exclude them from this study, since both north-eastern and southern states have widely divergent characteristics from the typical Midwestern state. Table 1 documents the specific state-level events that led to the annualized scores for each state presented here. Scores range from 0-5, with 0 being a score representing the highest level of same-sex marriage equality, and 5 being a score representing the most restrictive legislation. Major anti-gay events, such as passage of a restrictive bill into law, or a higher court ruling opposing the legal recognition of same-sex relationships, count as a +2, while other events, such as the proposal of an anti-gay bill, or lower court ruling, count as a +1. Consecutive events, such as the requirement of multiple-year passage of bills, are often not added, as such state processes would give those states an inadvertently high number compared to states with processes that do not require consecutive passage of bills for the movement into law.

In 1996, following the federal passage of DOMA, most Midwestern states, from 1996-1998, passed a series of state-level laws that prohibited the recognition of any same-sex marriages that may have been recognized in other states, with Iowa and Wisconsin being notable exceptions. After this period, there are relatively few major events until 2004. The subsequent events are documented in Table 2, which lists the specific events that create the state-level scoring system. The existence of DOMA-related laws in most states from the 1990s start them at a score of 2 as of 2003, the starting point for this analysis. In 2004, several state passed constitutional amendments banning same-sex marriage, specifically, Kentucky, Ohio and Michigan. In 2005, the Indiana court of appeals upheld the Indiana statute defining marriage as between one man and one woman—this event increased Indiana’s score from 2 to 3. In 2008, the Michigan Supreme Court decided that their constitutional amendment banning same-sex marriage also bans recognition of domestic partnerships, raising the Michigan score from 4 to 5. In 2009, Iowa’s Supreme Court struck down their anti-gay legislation, lowering their score from 2 to 0, indicating that gay marriage is fully legal in this state. In the same year, Wisconsin passed a law allowing for domestic partnerships, despite their constitutional amendment defining marriage as between one man and one woman, lowering their score from 4 to 2. In 2011, Illinois passed a civil unions law, lowering its score from 2 to 1, the same year that a Minnesota constitutional amendment ban on same sex marriage passed the legislature, raising its score from 4 to 5. However, the following year, Minnesota voters failed to support that amendment ban, lowering its score from 5 to 3. In 2013, Minnesota passed a law allowing for same sex marriage, which would drop their score from 3 to 0, except that this dataset stops at 2012 for the purposes of scoring gay rights by state, since this study looks at subsequent impact to the economy of gay rights legislation—since 2014 economic and firms data is not available, 2013 gay rights events are not pertinent here.

Table 1: Midwestern/Great Lakes Scores by State, Gay-Marriage or Relationship Legislation or Court Decisions. A score of 0 implies that state fully recognizes gay marriage, while a score of 5 implies the most restrictive laws against gay relationships.

2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
IA 2 2 2 2 2 2 0 0 0 0
IL 2 2 2 2 2 2 2 2 1 1
IN 2 2 3 4 4 4 4 4 4 4
KY 2 4 4 4 4 4 4 4 4 4
MI 2 4 4 4 4 5 5 5 5 4
MN 2 3 3 4 4 4 4 4 5 3
MO 1 3 3 3 3 3 3 3 3 3
OH 1 3 3 3 3 3 3 2 2 2
WI 1 2 3 4 4 4 2 3 3 2

Economic, jobs and firms data are publicly available from federal sources. This study looked at firms by industry by state and jobs by sector by state, available from the U.S. Census (factfinder2.gov), and real gross domestic product per capita by state, available from the U.S. Bureau of Economic Analysis (bea.gov), using data collected January 23, 2014, from the years 2003-2013. Given the testimony of corporations about the negative impact of anti-gay legislation on candidate recruitment, four published firm sub-types were analyzed: “Professional, Science, and Technology” firms, “Educational Services” firms, “Health and Social Services” firms, “Finance” firms, as well as “Total” firms. Each of these came from the one-year samples of the County Business Patterns for the years of 2003-2012 (U.S. Census, not all data available for all years, but all available data from these years was used for each variable). In addition to the raw firms numbers, a rate of change variable was calculated for each variable ([current year firms – previous year firms]/previous year firms). Job data is available from the American Community Survey (U.S. Census, not all data available for all years, but all available data from these years was used for each variable). In order to provide a contrast with the “professional” firms analyzed, two of the largest industry categories were chosen for the jobs analysis: manufacturing jobs, and construction jobs. Gross domestic product can be measured in several ways, but in order to compensate for the different population sizes for each state, this study uses real gross domestic product per capita from 2003-2013, as published by the BEA.

Table 2: Major and Minor Events by Midwestern/Great Lakes State, Gay-Marriage or Relationship Legislation or Court Decisions (ssm=same sex marriage) 2004 Approved state ban on ssm: KY, OH, MI
2004 WI constitutional amendment passed both houses
2004 MN ssm ban proposed, failed

2005 IN appeals court upholds ssm ban
2005 WI constitutional amendment passed both houses

2006 WI ban on ssm
2006 MN ssm ban proposed, failed
2006 IN fails to pass ssm ban

2007 MI appeals court says ssm ban also prohibits domestic partnerships
2007 MN ssm ban proposed, failed

2008 MI supreme court says ssm ban also prohibits domestic partnerships

2009 IA court rules ssm ban unconstitutional
2009 WI passes law allowing for domestic partnerships
2009 MN ssm ban proposed, failed

2010 OH lawsuit challenging domestic partnership registries fails

2011 MI bans domestic partnerships
2011 IL passes civil unions approval
2011 MN ssm ban amendment passed legislature
2011 IN constitutional amendment ssm ban passes assembly
2011 WI supreme court upholds ssm ban

2012 MN ssm ban rejected by voters
2012 WI court upholds domestic partnership

To perform this analysis, this study uses the software R, an open source (free and publicly available) statistical analysis project. The specific package used for this time-series, panel data analysis was PLM. All of the data from 2003-2013 were formatted according to the PLM requirements for the 9 Midwestern/Great Lakes states. Because the question at hand is the impact that each individual state’s legislation has on jobs and the economy for each state, a “within” (fixed) approach was used. This approach was confirmed using a Hausman test, which can be used to determine the validity of a random versus fixed approach. In this case, the variable relationships in question achieved statistical significance (p<0.05), with the null being that random effects are likely—in this case, the null was rejected, implying that the most relevant approach uses a fixed effects model. Two types of regression relationships were tested, the first being the bivariate relationship of the outcome variables (firms, jobs and GDP), to the predictor variable alone, the same-sex marriage related legislation/court events, and the second, being a multivariate analysis between either the firms or jobs variables alone, regressed to the combined same-sex marriage legislation/court events and GDP. The rationale for the first approach, is to determine the relationship on either firms or GDP of anti-gay legislation. This approach ignores the potential impact of GDP on firms, or firms on GDP. This establishes a baseline of relationship between these factors and legislative/court events. The rationale for the second approach, is that in reality, GDP does impact subsequent firms—i.e., a downturn of the GDP can itself produce a loss of professional firms, or an upturn of the GDP can subsequently produce a growth in professional firms. The second approach looks at the impact on professional firms of the combined GDP + legislative/court events. The resulting regression equations follow the typical pattern for estimator prediction, where y is the outcome variable, in this case, sector-level firms or GDP, and x is the predictor variable estimator, in this case either GDP, or the legislative/court events. For the combined GDP + legislative/court events analysis, only additive effects are evaluated, not combined effects, and none of the variables are transformed (i.e., the logarithm is not generated, nor are quadratic, or other non-linear effects evaluated).

A) Firms or Jobs or GDP ~ Same sex marriage legislative/court event
B) Firms or Jobs ~ Same sex marriage legislative/court event + GDP

Results

The results of this panel analysis are statistically significant and robust. Take, for example, the regression equation generated for the simple relationship between “Professional, Science, and Technical” firms as impacted solely by the legislative/court events, as seen in Table 3. “Lagged years” represents how length of time between the legislative/court events and the firms measure. For example, for “lagged years”=2, the model suggests with a certainty of p=0.07 (within a 90% confidence that these results are not due to chance), that at two years after an anti-gay legislative/court event, there is a subsequent decrease in the number of residents of that state employed in the professional, scientific, and technical fields at a rate of 108.6 firms for every increase of anti-gay score of “1.” At the third year after the anti-gay legislative event, there is a measurable decrease of 191.1 professional firms at a confidence of p=99.9% that these results are not due to chance. The results are equally strong at the fourth year after the event, and demonstrate an even greater loss of firms. Table 4 shows a similar pattern, but for “Finance” firms. While the impact on the 2nd year after the legislative/court events is not statistically significant, the impact is significant at the 3rd and fourth year, with an even greater impact on raw firms than professional/science/technical firms, at a loss of 309 firms for every anti-gay legislative/court event, with the fourth year events showing confidence at the 99.9% level that these results are not due to chance.

Table 3: Professional, science, and technical establishments, in raw firms, regressed against legislative/court events impacting same-sex marriage or relationships Legislation/Court Event lagged years Firms Estimator (in firms per state per event/score) Statistical Significance (p)
2 -108.6 0.07 *
3 -191.1 0.00 ***
4 -209.0 0.00 ***

Table 4: Finance establishments, in raw firms, regressed against legislative/court events impacting same-sex marriage or relationships Legislation/Court Event lagged years Firms Estimator (in firms per state per event/score) Statistical Significance (p)
2 -17.7 0.82
3 -205.0 0.08 *
4 -309.9 0.00 ***

These impacts are not limited to professional firms. Anti-gay legislative/court events were also compared to total firms numbers. As shown in Table 5, these negative total firms impacts are even stronger than the sector-level approach above. While the sector-level effects are not statistically significant until the 2nd or 3rd year out, total firms are impacted the immediate year after the legislative/court events. In this case, all four years after the anti-gay events have statistically significant effects at the 95-99.9% confidence level, costing each state thousands of firms per year for each anti-gay legislative/court event score. As a frame of reference, in 2011, for these nine states, the average number of total firms was 169,282. So for the 4th year after an anti-gay legislative/court event, the average firm loss would be approximately 2.1% for each event-score.

Table 5: Total establishments, in raw numbers, regressed against legislative/court events impacting same-sex marriage or relationships Legislation/Court Event lagged years Firms Estimator (in firms per state per event/score) Statistical Significance (p)
1 - 1, 425 0.05 **
2 - 2,342 0.00 ***
3 - 3,294 0.00 ***
4 - 3,642 0.00 ***

A second approach used rates of change, rather than raw firm numbers, in some cases, the apparent loss of firms could be attributed to broader historical trends, despite the statistical significance at the 99.9% levels. If these trends have been occurring for over a decade, then such patterns might be masked when looking at raw firm numbers. For example, Indiana, and several Midwestern states, have a persistent “brain drain” as college graduates move to larger cities with more desirable natural amenities, such as mountains and oceans, or jobs that can offer higher rates of pay, advancement or influence. Rates of change can be a way to filter some of those effects. In Table 6, rates of changes are used to look at the impact of the anti-gay legislative/court decisions on health and social services firms in the Midwestern/Great Lakes states. As shown above, these firms are also negatively impacted by anti-gay events. In this case, the immediate year following an anti-gay event leads to a decrease in the rate of health/social services firms by 0.4% from the previous year, with a confidence of greater than 95% that these results are not due to chance. At the fourth year out, firm growth decreases by 3.3% from the previous year, at a confidence level of 99.9%. Table 7 shows these same effects for education firms.

Table 6: Health and Social Services establishments, in rate of change, regressed against legislative/court events impacting same-sex marriage or relationships Legislation/Court Event lagged years Health/Social Services Firms Estimator ( in % rate of change per event/score) Statistical Significance (p)
1 -0.4 % 0.02 **
2 -0.7 % 0.34
3 -0.4 % 0.07 *
4 -3.3 % 0.00 ***

Table 7: Educational services firms, in rate of change, regressed against legislative/court events impacting same-sex marriage or relationships Legislation/Court Event lagged years Educational Services Firms Estimator ( in % rate of change per event/score) Statistical Significance (p)
1 -0.5 % 0.10 *
2 -0.8 % 0.02 **
3 -1.1 % 0.01 ***
4 -1.8 % 0.00 ***

In addition to the negative impact to firms, anti-gay legislative/court events also have a negative impact on the number of jobs in the state. The results of this third approach can be seen in Tables 8 and 9, using calculations similar to those above, but regressed against construction jobs and manufacturing jobs (data from the American Community Survey, U.S. Census). In this case, the loss of jobs in both of these industries is evident. For example, in the case of manufacturing, the job loss is not statistically significant the first year after a major anti-gay legislative/court event. However, by each of the 2nd-4th years, job losses are evident with 99.8-99.99% confidence levels that these results are not from chance. The job losses per year average between 20,000-25,000 manufacturing jobs for each anti-gay legislative/court event-score. Construction job loss follows a similar pattern. As a frame of reference, in 2012, the average number of construction jobs in these nine states was 181,340, and the average number of manufacturing jobs was 511,422. Thus the average job loss per negative event-score the third year after the event, would, respectively, be approximately 8.2% of construction jobs, and 4.9% of manufacturing jobs.

Table 8: Manufacturing jobs, regressed against legislative/court events impacting same-sex marriage or relationships Legislation/Court Event lagged years Manufacturing Jobs Estimator (in jobs per state per event/score) Statistical Significance (p)
1 -9,526 0.31
2 -20,780 0.002 ***
3 -25,080 0.000 ***
4 -23,908 0.000 ***

Table 9: Construction jobs, regressed against legislative/court events impacting same-sex marriage or relationships Legislation/Court Event lagged years Construction Jobs Estimator (in jobs per state per event/score) Statistical Significance (p)
1 -3,424 0.53
2 -11,997 0.002 ***
3 -14,859 0.000 ***
4 -15,023 0.000 ***

A fourth approach looks specifically at the impact of the economy on these anti-gay legislative/court events, by comparing subsequent years of real gross domestic product per capita. As shown in Table 8, these impacts are significant and negative. While this study does not test mechanisms for the impact on GDP, it is reasonable to infer that these negative GDP effects are related to the loss of firms and jobs described above. These effects are statistically significant to the 99.0-99.9% levels through the first three years. As a frame of reference, in 2012, for these nine states, the average real GDP/capita was $39,677.

Table 10: Real gross domestic product, regressed against legislative/court events impacting same-sex marriage or relationships Legislation/Court Event lagged years Real GDP/Capita Estimator (per event/score) Statistical Significance (p)
1 - $384 0.01 ***
2 - $521 0.00 ***
3 - $518 0.00 ***
4 - $318 0.10 *

Finally, a comparison was made between establishments and jobs, separately, regressed against the combination of GDP and anti-gay legislative/court events. The results of this analysis can be seen in Table 9, which looks at total establishments as a combination of these two effects. As expected, real GDP per capita has a strong positive effect on the number of firms in each state. For example, for real per capita GDP, every measured dollar is related to an increase in almost 3 firms the following year, and the second year out, just over 2 firms per dollar. However, by the 3rd year, the impact of GDP/capita drops out of statistical significance. On the other hand, even when taking real GDP/capita into consideration, the competing impact of anti-gay legislative/court events is statistically significant for all four yours after any event. For every major anti-gay event, there is a measurable loss of 963 firms the first year out, a loss of 2,104 firms the second year out, and the 3rd and 4th year have a loss of over 3,000 firms per year for each anti-gay event. These effects are measurable at a 99.9% confidence level, far overpowering the impact of GDP by the 3rd and 4th years after any given anti-gay legislative/court event. While not presented here, these results are evident at the sectoral level as well—i.e., even when taking GDP into account, there is a negative impact on professional job loss from each anti-gay legislative/court event in any given state (results available on request). Similarly, the loss to manufacturing jobs are also robust to the inclusion of GDP into the calculation. The first year after the event-score has no statistical impact on job loss, but each of the subsequent three years show a loss of tens of thousands of jobs per event-score at the 99.9% confidence level, overshadowing any GDP impact.

Table 11: Total firms, regressed against legislative/court events impacting same-sex marriage or relationships combined with real GDP/capita Legislation/Court Event lagged years Anti-Gay Legislative/Court Events (p) Real GDP/Capita Estimator (p)
1 - 963 firms (0.06) * 2.9 (0.00) ***
2 - 2,104 firms (0.00) *** 2.1 (0.00) ***
3 - 3,378 firms (0.00) *** 0.7 (0.37)
4 - 3,415 firms (0.00) *** -0.8 (0.32)

Table 12: Manufacturing jobs, regressed against legislative/court events impacting same-sex marriage or relationships combined with real GDP/capita Legislation/Court Event lagged years Anti-Gay Legislative/Court Events (p) Real GDP/Capita Estimator (p)
1 - 3,985 jobs (0.49) 32.9 (0.00) ***
2 - 18,473 jobs (0.00) *** 14.5 (0.00) ***
3 - 24,882 jobs (0.00) *** 1.7 (0.70)
4 - 23,168 jobs (0.00) *** -6 (0.37

Conclusion

U.S. federal, state, and local public policy is rapidly changing to affirm different family types than the one-man, one-woman ideal that has been dominant in white, middle class culture since WWII. Setting aside religious, moral and social questions of the impact of these changes to legal and cultural recognition of a diversity of family types, the economic impact of states deciding not to affirm diversity can have measurable impacts on both firms and economic growth. This study used a time-series, panel data approach to determine the relationship between state-level legislation and court decisions that impact either same-sex marriage or other types of legal same-sex relationship, such as domestic partnerships and civil unions. Since the 1996 federal DOMA law, various states have taken divergent approaches to these legal affirmations of diversity. Most of the Midwestern/Great lakes states have taken approaches that restrict the rights of same-sex couples. This study looked at nine of these states that are arguably similar in cultural and economic features to Indiana, where the legislature is in the process of evaluating a constitutional amendment banning same-sex marriage and similar legal affirmations of such relationships, to the citizens of Indiana for passage. This study seeks to contribute to that discussion by proposing a negative impact to the Indiana economy because of the passage of such a constitutional amendment ban.

Four different measures were examined in relation to the univariate regression of antigay legislative/court events: sector level and total firms by raw numbers, rates of growth for sector-level firms, jobs and economic growth (real GDP per capita). Further, firms and jobs were examined in relation to the combination measure of anti-gay legislative/court events and economic growth. For all of these measures, a statistically significant finding was that each anti-gay event leads to a negative impact on firms and the economy, often at the 99.9% confidence level, and at the cost of thousands of firms per anti-gay event per state. Given that the current GOP-dominated legislature has previously emphasized their interest in creating jobs versus the disruption caused by social issues debates in their 2012 state platform, it seems unreasonable for the state assembly to pursue this divisive bill, that will undoubtedly hurt jobs growth in Indiana.

One can visualize these effects to produce a more striking representation of the negative impact that this amendment will have on Indiana’s economy. Chart 1 shows the magnitude of this relationship when looking at total firms, with a 3-year lag, comparing three groups of states: States with a score of 0-1 (strongest laws that protect same-sex relationships), states with a score of 2-3 (moderate laws restricting same-sex relationships), and states with a score of 4-5 (strongest laws restricting same-sex relationships). For this chart, growth of total firms is measured for the 3rd year after a given state score. On average, when comparing states with a score of 0-1, those are the only Midwestern/Great Lakes states that show total firms growth. States with moderately restrictive laws show slight negative total firms growth. States with the strongest regulations against same-sex marriage have by far the strongest negative growth rates for total firms. Chart 2 shows the same effect, but measured as a factor of real GDP per capita growth over a 2-year lag. The same progression is evident—the states with the strongest protection for same-sex relationships have the strongest economic growth, measured 2 years out, while the states with the most restrictive laws, not only do poorer, but actually have negative growth. Finally, in Chart 3 one sees the same relative impact on the growth of manufacturing jobs of anti-gay legislative/court events. As states evidence greater respect for the rights of sexual minorities, they see a subsequent impact on jobs growth, in this case, manufacturing jobs at a 3-year lag. On the other hand, states that are the most restrictive in limiting the legal affirmation of the relationships of gays and lesbians, those states evidence a strong jobs decline.

No comments:

Post a Comment