Week 1 – Assignment 1: Assess Political Decision Making and Week 2 - Assignment: Examine the Impact of Economic Policies on Society

JOURNAL OF FINANCIAL AND QUANTITATIVE ANALYSIS Vol. 53, No. 1, Feb. 2018, pp. 109–136 COPYRIGHT 2018, MICHAEL G. FOSTER SCHOOL OF BUSINESS, UNIVERSITY OF WASHINGTON, SEATTLE, WA 98195 doi:10.1017/S0022109017000977

Fiscal Policy, Consumption Risk, and Stock Returns: Evidence from U.S. States

Zhi Da, Mitch Warachka, and Hayong Yun*

Abstract We find that consumption risk is lower in states that implement countercyclical fiscal poli- cies. Moreover, firms with an investor base that is concentrated in countercyclical states have lower stock returns, along with firms that relocate their headquarters to a countercycli- cal state. Therefore, countercyclical fiscal policies lower the consumption risk of investors and, consequently, their required equity return premium. This conclusion is confirmed by smaller declines in market participation during recessions in countercyclical states. Over- all, the location of a firm’s investor base enables state-level fiscal policy to influence stock returns.

I. Introduction The controversies surrounding fiscal policy usually involve its macroeco-

nomic implications for unemployment or investment. In contrast, we investigate the impact of fiscal policy at the state level on stock returns. Our study pro- vides two novel empirical findings regarding consumption-based asset pricing in a market with imperfect risk sharing among investors. First, countercyclical fiscal policies lower consumption risk. Second, firms with an investor base that is

*Da, [email protected], University of Notre Dame Mendoza College of Business; Warachka (corre- sponding author), [email protected], University of San Diego School of Business; and Yun, [email protected], Michigan State University Broad College of Business. We thank Jennifer Con- rad (the editor) and Mariano Croce (the referee) for helpful comments. We also thank Bo Becker, Frederico Belo, Effi Benmelech, Michael Brennan, Cristina Cella, Henrik Cronqvist, Phillip Dybvig, Antonio Fatas, Wayne Ferson, Francisco Gomes, Klaus Grobys, Harrison Hong, Eric Hughson, Chris- tian Julliard, Andrew Karolyi, Roger Loh, Dong Lou, Joel Peress, Jeff Pontiff, Lucio Sarno, Ivan Shaliastovich, Richard Smith, Jiang Wang, Tracy Wang, Scott Weisbenner, Shu Yan, and Fan Yu for their comments as well as participants at the 2013 Sun Trust Beach Conference, the 2013 Financial Intermediation Research Society (FIRS) conference, the 2013 Rothschild Caesarea Center Confer- ence, the 2013 China International Conference in Finance, the 2013 Auckland Finance Meeting, the 2012 European Finance Association Meeting, the 2011 Financial Research Association Meeting, and the 2011 Center for Economic Policy Research (CEPR) Summer Symposium in Gerzensee. We thank George Korniotis for providing us with state-level retail sales data and Diego Garcia for providing us with data on the state-level operations of firms. We also thank Hojong Shin for his excellent research assistance. Some of the data used in our analysis are derived from the Restricted Data Files of the Panel Study of Income Dynamics, obtained under special contractual arrangements designed to protect the anonymity of respondents. These data are not available from the authors.

109

110 Journal of Financial and Quantitative Analysis

concentrated in countercyclical states have lower stock returns. Intuitively, state- level fiscal policy impacts the consumption risk of investors and, consequently, their required equity return premium.

Our results are consistent with state governments financing consumption during recessions through their budget deficits. For example, tax reductions dur- ing a recession can finance consumption that otherwise would have required in- vestors to sell a larger portion of their equity portfolio or invest less in the equity market. Without this countercyclical fiscal policy, investors may require a higher equity return premium as compensation for their higher consumption risk. Ricardian equivalence (Barro–Ricardo equivalence theorem) disputes the ability of government budget deficits to stabilize consumption because government debt issuance increases expected taxes. However, household borrowing constraints and difficulties hedging income shocks allow fiscal policy to influence consumption.1

The ability of state governments to issue debt also differs from that of the fed- eral government. Because state governments are constrained by their respec- tive balanced-budget amendments, many states have adopted budget stabilization funds that accumulate precautionary savings to facilitate countercyclical fiscal policies.

A state’s fiscal policy is identified by the sensitivity of its budget deficit (sur- plus) to state-level economic growth. Specifically, we estimate state-level fiscal policy betas to identify variation in fiscal policy across states. We also estimate a predicted fiscal policy beta for each state based on the deposit rules of its budget stabilization fund and its frequency of having a Democratic governor. Counter- cyclical states are more likely to accumulate precautionary savings through strin- gent deposit rules and are more likely to elect a Democratic governor.2

We find that consumption declines less during bust periods in countercyclical states. A bust period for a state is defined as negative economic growth relative to the prior year. A 1-standard-deviation fluctuation in the fiscal policy beta results in consumption declining 1.47% less during bust periods in the countercyclical state. Thus, consumption risk is lower in countercyclical states. Moreover, consumption in the 5 most procyclical states is 34% more volatile than consumption in the 5 most countercyclical states. This cross-sectional variation in consumption is consistent with imperfect risk sharing within the United States.

The local-investment-bias literature (Coval and Moskowitz (1999)) doc- uments the tendency of investors to overweight nearby firms. Pirinsky and Wang (2006) find evidence of return comovement attributable to correlated trad- ing among investors in the same location. Korniotis and Kumar (2013) and Kumar, Page, and Spalt (2012) confirm that the trades of local investors are

1 Heaton and Lucas (1992), (1996) demonstrate the importance of borrowing constraints and in- complete insurance. Lucas (1994) notes that investors self-insure against transitory shocks to labor income through precautionary savings. However, Brav, Constantinides, and Geczy (2002) find evi- dence that consumption is not completely insured, whereas Storesletten, Telmer, and Yaron (2004) find that precautionary savings provide inadequate insurance against a prolonged negative income shock (unemployment).

2 Tuzel and Zhang (2017) estimate local betas for metropolitan areas based on their industrial com- position. However, these local risk factors are motivated by immobile assets such as real estate whose markets clear at the metropolitan level, not differences in the fiscal policy of metropolitan areas.

Da, Warachka, and Yun 111

correlated and, consequently, that a firm’s cost of equity depends on its head- quarters location.3

We initially compute a firm-level local-investment-bias measure that cap- tures the sensitivity of a firm’s investor base to the fiscal policy of the state in which its headquartered. We find that the combination of local investor bias and a countercyclical fiscal policy lowers equity returns. State-level political uncer- tainty cannot explain this finding. In terms of economic significance, a 0.70% annual difference in their cost of equity is attributable to a fluctuation in the fiscal policy beta of 1 standard deviation. The difference in the average fiscal policy beta of the 5 most countercyclical states versus the 5 most procyclical states produces a return difference of 2.59% per year. Headquarters relocations confirm the im- portance of state-level fiscal policy to returns because firms relocating to a more countercyclical state subsequently have lower stock returns.

We then aggregate the state-level fiscal policy betas across investors to esti- mate the fiscal policy of each firm’s investor base. This aggregate fiscal policy beta is computed using the dollar-denominated holdings of all institutional investors in a firm and the fiscal policy beta associated with their respective state locations. Using the aggregate fiscal policy betas, we find that firms have lower stock returns if their investor base is concentrated in countercyclical states. Predicted fiscal pol- icy betas, which condition on state-level budget-stabilization fund-deposit rules and the political affiliation of state governors, confirm that countercyclical fiscal policies lower stock returns.

Two channels enable countercyclical fiscal policies to influence stock re- turns: a discount-rate channel and a cash-flow channel. Both of these channels originate from geographic segmentation. The discount-rate channel arises from the location of investors, whereas the cash-flow channel arises from the location of customers. The importance of investor location to the relation between fiscal policy and stock returns supports the discount-rate channel. State-level market participation lends further support to the discount-rate channel. Specifically, par- ticipation in the equity market declines less during recessions in countercyclical states and exhibits less variation over time.

Nonetheless, by smoothing the consumption of households (including non- investors), countercyclical fiscal policies can stabilize firm-level cash flow. However, unlike consumption, cash flow does not decline less during bust periods for firms headquartered in more countercyclical states. Using the data of Garcia and Norli (2012), we construct firm-level cash-flow betas by weighting the state- level fiscal policy betas by the fraction of a firm’s operations in each state. We find that the ability of these cash-flow betas to explain returns is limited to firms that operate in a single state. Intuitively, the influence of state-level fiscal policy on cash flow is mitigated by firms having diversified operations across several states. Korniotis and Kumar (2013) also conclude that a firm’s cost of capital depends on the discount-rate channel instead of the cash-flow channel.4

3 The economic justification for the local investment bias has been attributed to informational ad- vantages (Ivković and Weisbenner (2005)), familiarity (Huberman (2001)), and social interactions (Hong, Kubik, and Stein (2005), Ivković and Weisbenner (2007)).

4 Weak empirical support for the cash flow channel in our sample of public firms does not nec- essarily imply its irrelevance. With access to the public equity market, firms can expand and lower

112 Journal of Financial and Quantitative Analysis

Although Korniotis and Kumar (2013) do not examine the role of fiscal pol- icy, an extensive literature on fiscal policy precedes our empirical study. Poterba (1994) examines cross-sectional variation in fiscal policy across states. However, this study does not examine its implications for consumption risk or stock returns. Moreover, the existing literature that studies the consumption implications of fis- cal policy typically focuses on specific stimulus programs initiated by the federal government during recessions. Parker, Souleles, Johnson, and McClelland (2013) report that stimulus payments during the 2008 financial crisis increased house- hold consumption, and Johnson, Parker, and Souleles (2006) document increased consumption following the 2001 tax rebates. In contrast, we focus on the cross- sectional relation between state-level fiscal policy and consumption risk, along with its implications for stock returns.

Our study of state-level fiscal policy mitigates the policy and tax uncertainty induced by intervention in the economy by the federal government. Pastor and Veronesi (2012) examine the impact of government policy uncertainty on asset prices. Pastor and Veronesi (2013) estimate a risk premium for government pol- icy uncertainty, and Kelly, Pastor, and Veronesi (2016) find empirical evidence that political uncertainty increases the implied volatility of options. However, instead of uncertainty arising from regulatory and trade policies, the impact of countercyclical fiscal policies on tax uncertainty is more relevant to our study. Croce, Nguyen, and Schmid (2012) and Gordon and Leeper (2005) highlight the long-term cost of countercyclical fiscal policies, with Croce, Kung, Nguyen, and Schmid (2012) concluding that tax uncertainty is as important to the cost of equity as the level of taxation.

However, state-level budget stabilization funds accumulate savings before the occurrence of poor economic conditions to offset debt issuance that is con- strained by balanced-budget amendments. Therefore, the countercyclical fiscal policies of state governments induce less uncertainty regarding future taxation. On average, countercyclical states have nearly identical levels of outstanding debt as procyclical states (normalized by economic output). Moreover, outstanding debt at the state level is typically below 10% of annual economic output, which is far below the federal government’s ratio of debt to gross domestic product (debt–GDP ratio).

The remainder of the article is organized as follows: Section II details the estimation of fiscal policy at the state level. Section III provides our results re- garding the relation between fiscal policy and consumption risk, and Section IV provides our results regarding the relation between fiscal policy and stock returns. Section V then offers our conclusions.

II. State-Level Fiscal Policy Korniotis (2008) highlights three advantages of using state-level data to ex-

amine consumption-based asset pricing models. First, state-level data have less measurement error than individual data. Second, income shocks are not fully

their exposure to the fiscal policy of any individual state. In contrast, the sensitivity of consumption to state-level fiscal policy may reflect purchases from smaller firms whose operations are concentrated in a few states.

Da, Warachka, and Yun 113

diversified across states. Third, there is considerable cross-sectional variation in consumption across states. Our analysis identifies another important property of using state-level data: variation in fiscal policy.

Although our cross-sectional tests are limited to 50 states, international stud- ies often involve fewer countries, and their conclusions are complicated by dif- ferences in labor markets as well as legal, political, and monetary institutions (Acemoglu, Johnson, Robinson, and Thaicharoen (2003)). In contrast, state-level data enable us to condition on deposit rules and political affiliations, whose defi- nitions are comparable across states.

Nonetheless, in an international setting, Julio and Yook (2012) conclude that investment declines around national elections, and Durnev (2010) reports that investment is less sensitive to stock prices during election periods. In ad- dition, Brogaard and Detzel (2015) construct a country-specific proxy for eco- nomic policy uncertainty and report that greater uncertainty reduces investment. Within the United States, Santa-Clara and Valkanov (2003) find that stock returns are higher during Democratic presidencies, and Belo, Gala, and Li (2013) report that the market is positively surprised by the spending policies of Democratic presidents. In contrast to their time-series methodologies, we examine the con- sumption and long-term return implications of government fiscal policy from a cross-sectional perspective. Thus, the motivation for our study is return variation attributable to differences in fiscal policy at the state level rather than errors in investor expectations.

To identify a state’s fiscal policy, we examine the response of its budget deficit (surplus) to different economic conditions. Annual data on state govern- ment revenue and expenditures as well as the gross state product (GSP) of each state from 1965 to 2008 are obtained from the Statistical Abstract of the United States maintained by the U.S. Census Bureau. Our sample ends in 2008 due to the availability of state-level data.

Annual budget deficits, and consequently surpluses, are defined as follows:

(1) DEFICITi ,t = EXPENDITUREi ,t −REVENUEi ,t

GSPi ,t .

When positive, this state-year observation represents a budget deficit for state i in year t . Conversely, when negative, this state-year observation represents a budget surplus. To clarify, withdrawals from a budget stabilization fund facilitate expenditures in excess of revenue that correspond to a budget deficit. For example, during the financial crisis, budget stabilization funds provided an average of 5.1% of state government expenditures. Deposits into a budget stabilization fund are included in expenditures and correspond to a smaller budget surplus in normal economic conditions.

The nature of each state’s fiscal policy is estimated using the following state- level time-series regression:

DEFICITi ,t = βi ,1GSP GROWTHi ,t +βi ,2GSP GROWTHi ,t−1(2) +βi ,3DEFICITi ,t−1 +�i ,t ,

based on GSP growth in state i between year t and t−1. Because economic condi- tions and budget deficits are autocorrelated, equation (2) includes a state’s lagged

114 Journal of Financial and Quantitative Analysis

GSP growth and lagged budget deficit. The inclusion of additional lags does not change our results. Svec and Kondo (2012) estimate a regression specification that is similar to equation (2), but their study does not investigate the impact of fiscal policy on consumption or stock returns.

The βi ,1 coefficient, which measures the sensitivity of a state’s government budget deficit (surplus) to contemporaneous economic growth, defines state i ’s fiscal policy beta. Hereafter, we abbreviate the fiscal policy beta of state i as FPBi . Because state governments can implement countercyclical fiscal policies by lowering taxes or increasing expenditures, the dependent variable DEFICIT focuses on their difference. Nonetheless, certain government expenditures may exert a greater impact on stock returns. For example, Belo and Yu (2013) find a positive relation between government expenditures on public-sector capital and stock returns.5

We also include the contemporaneous state-level unemployment rate and personal income growth as additional control variables in equation (2). We then define the alternative fiscal policy beta as the βi ,1 coefficient in this enhanced specification. Our later empirical results are consistent using both the original and alternative fiscal policy betas. The alternative fiscal policy beta denoted ALTERNATIVE FPB that accounts for a state’s unemployment rate and personal income has a 0.532 correlation with the original fiscal policy beta from equation (2). The similarity between these fiscal policy betas and their implications for consumption risk as well as returns is consistent with GSP growth providing a sufficient economic signal to guide government policy.

Panel A of Table 1 sorts the state-level FPBs from lowest to highest, with the average FPB being −0.014. A comparison of the 5 most countercyclical states and 5 most procyclical states indicates that their average fiscal policy betas equal −0.239 and 0.206, respectively. This 0.445 difference is highly significant.

For emphasis, only a relative ranking of state-level fiscal policy is required for our analysis.6 The actual magnitude, and hence significance, of an individ- ual state’s fiscal policy beta is irrelevant to our later empirical tests. Furthermore, fiscal policy is evaluated as a time-invariant state characteristic because our objec- tive is to investigate the cross-sectional relation between fiscal policy and equity returns over the long term. Nonetheless, we estimate the fiscal policy betas in 2 subperiods: from 1964 to 1985 and from 1986 to 2009. In unreported results available from the authors, the correlation in the fiscal policy betas across these 2 subperiods equals 0.580. This correlation is higher, 0.746, for the alternative fiscal policy beta. Therefore, the fiscal policy betas exhibit limited variation over time.

The bottom of Panel A in Table 1 reports on the difference in state character- istics between the most countercyclical and the most procyclical states as well as the correlation between these characteristics and the fiscal policy betas. Although the statistical significance of these differences and correlations is not reported in Table 1, these characteristics are utilized later in formal empirical specifications.

5 In unreported results, removing the government expenditure component from GSP produces nearly identical fiscal policy betas to those reported in Table 1.

6 Therefore, any bias in the state-level fiscal policy betas due to the correlation between GSP growth and budget deficits (GSP growth and personal income in the alternative fiscal policy betas) is unlikely to affect our conclusions.

Da, Warachka, and Yun 115

TABLE 1 State-Level Fiscal Policy

Panel A of Table 1 reports on the state-level fiscal policy betas (FPB) in equation (2) that measure the sensitivity of annual budget deficits to economic growth (GSP_GROWTH):

DEFICITi,t = βi,1GSP_GROWTHi,t +βi,2GSP_GROWTHi,t−1 +βi,3DEFICITi,t−1 +�i,t .

A state’s budget deficit is computed according to equation (1). The fiscal policy beta of state i equals βi,1. An alternative fiscal policy beta is estimated by including a state’s unemployment rate and its growth in personal income as independent variables in equation (2). Panel A also reports the fraction of the sample period that each state’s economic growth is negative (BUST) and its governor is a Republican. Political uncertainty is measured by the number of transitions in the political affiliation of a state’s governor. The stringency of the deposit rules governing each state’s budget stabilization fund on a scale of 1 to 4 is then reported, followed by the volatility of the state’s economic growth. Average state-level consumption growth is then reported, along with its volatility. Panel B reports summary statistics for the budget deficits of state governments, along with the correlation between their revenue and expenditures. Average revenue, expenditures, and outstanding debt normalized by gross state product (GSP) are also reported for each state. Panel C reports predicted fiscal policy betas that are estimated using equation (3):

FPBi = α0 +α1DRi +α2REP_GOVi +�i .

This regression conditions on the stringency of the deposit rules (DR) for each state’s budget stabilization fund and their frequency of having a Republican governor (REP_GOV). t-statistics are below each of the estimates in italics. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Panel A. Fiscal Policy Betas and State Characteristics

Alternative BUST Fraction Political Deposit GSP Consumption Consumption State FPB FPB Fraction Republican Uncertainty Rules Volatility Growth Volatility

NY −0.347 −0.402 0.022 0.286 3 4 0.030 0.049 0.040 OH −0.312 −0.213 0.022 0.643 5 2 0.035 0.052 0.043 PA −0.261 −0.290 0.022 0.357 5 2 0.026 0.051 0.041 IA −0.149 −0.077 0.043 0.571 3 1 0.044 0.051 0.037 WI −0.126 −0.285 0.022 0.429 6 3 0.031 0.054 0.044

NJ −0.124 −0.172 0.022 0.500 7 2 0.030 0.059 0.044 CT −0.117 −0.174 0.022 0.429 2 2 0.035 0.056 0.049 HI −0.117 −0.138 0.000 0.154 1 1 0.041 0.071 0.071 WV −0.115 −0.010 0.043 0.429 6 2 0.036 0.054 0.046 DE −0.110 −0.056 0.022 0.429 4 2 0.036 0.064 0.050 MI −0.109 −0.056 0.087 0.429 3 4 0.050 0.053 0.048 VA −0.107 −0.161 0.000 0.857 5 4 0.029 0.068 0.050 ME −0.091 −0.176 0.000 0.500 2 2 0.034 0.064 0.050 NV −0.072 −0.108 0.022 0.643 5 4 0.046 0.092 0.059 GA −0.060 −0.090 0.022 0.615 3 2 0.037 0.072 0.055 LA −0.040 −0.062 0.087 0.692 6 2 0.067 0.062 0.055 TN −0.038 −0.084 0.022 0.714 6 3 0.036 0.069 0.053 IL −0.031 −0.128 0.022 0.500 4 2 0.028 0.049 0.046 SC −0.030 −0.119 0.000 0.786 5 3 0.037 0.072 0.050 MN −0.030 −0.145 0.022 0.143 3 1 0.036 0.059 0.046 VT −0.027 −0.108 0.043 0.571 7 2 0.043 0.066 0.047 MA −0.020 −0.187 0.022 0.214 4 2 0.034 0.053 0.057 NM −0.020 −0.109 0.065 0.571 6 2 0.055 0.066 0.051 RI −0.017 −0.097 0.000 0.214 4 1 0.031 0.053 0.055 WA −0.012 −0.144 0.000 0.429 3 2 0.037 0.068 0.049 MT −0.009 −0.056 0.043 0.857 3 1 0.046 0.056 0.049 IN −0.007 −0.037 0.022 0.857 3 4 0.037 0.053 0.046 NE −0.006 0.012 0.000 0.857 7 2 0.039 0.052 0.040 AZ −0.004 −0.069 0.022 0.929 6 4 0.047 0.081 0.061 TX −0.003 −0.067 0.043 0.714 5 2 0.053 0.072 0.052 KY 0.002 −0.039 0.043 0.714 4 2 0.036 0.059 0.047 AR 0.005 −0.093 0.000 0.538 5 1 0.037 0.064 0.059 AK 0.009 0.047 0.152 0.923 6 1 0.137 0.076 0.096 MO 0.011 −0.066 0.022 0.571 6 1 0.031 0.055 0.045 NH 0.014 −0.042 0.043 0.643 6 2 0.044 0.075 0.057 FL 0.031 −0.140 0.022 0.714 5 2 0.042 0.078 0.058 KS 0.050 −0.081 0.022 0.929 7 3 0.035 0.055 0.042 OK 0.055 −0.053 0.087 0.929 5 2 0.056 0.062 0.053 MD 0.055 −0.133 0.000 0.286 3 3 0.028 0.061 0.051 AL 0.066 −0.104 0.022 0.769 4 4 0.036 0.064 0.047 CA 0.068 −0.236 0.022 0.571 5 2 0.036 0.062 0.050 ID 0.076 −0.074 0.000 0.929 2 1 0.044 0.066 0.053 NC 0.099 −0.051 0.000 0.643 4 2 0.031 0.070 0.053 ND 0.101 −0.007 0.109 0.929 2 2 0.087 0.054 0.042 SD 0.121 −0.009 0.022 1.000 2 2 0.053 0.055 0.045

(continued on next page)

116 Journal of Financial and Quantitative Analysis

TABLE 1 (continued) State-Level Fiscal Policy

Panel A. Fiscal Policy Betas and State Characteristics (continued)

Alternative BUST Fraction Political Deposit GSP Consumption Consumption State FPB FPB Fraction Republican Uncertainty Rules Volatility Growth Volatility

OR 0.131 −0.092 0.043 0.500 3 1 0.040 0.058 0.056 WY 0.148 −0.013 0.130 0.929 3 1 0.085 0.065 0.057 UT 0.227 −0.136 0.000 1.000 2 2 0.039 0.074 0.057 CO 0.248 −0.142 0.022 0.786 3 3 0.042 0.070 0.057 MS 0.278 −0.023 0.000 0.833 3 1 0.039 0.062 0.050

Average −0.014 −0.106 0.031 0.629 4.240 2.16 0.043 0.063 0.051 FPB correlation 0.532 0.071 0.517 −0.191 −0.216 0.243 0.295 0.261

Countercyclical −0.239 −0.253 0.026 0.457 4.400 2.400 0.033 0.051 0.041 Procyclical 0.206 −0.081 0.039 0.810 2.800 1.600 0.049 0.066 0.055

Panel B. State Government Deficit Characteristics

Budget Deficit

Correlation Alternative Std. Revenue/ Average Average

State FPB FPB Mean Dev. Skewness Expenditure Revenue Expenditure Debt

NY −0.347 −0.402 −0.004 0.017 1.610 0.411 0.301 0.289 0.095 OH −0.312 −0.213 −0.012 0.022 3.016 0.483 0.282 0.254 0.041 PA −0.261 −0.290 −0.004 0.016 2.535 0.497 0.259 0.247 0.050 IA −0.149 −0.077 −0.005 0.011 1.418 0.447 0.251 0.239 0.023 WI −0.126 −0.285 −0.011 0.027 2.238 0.228 0.264 0.240 0.054

NJ −0.124 −0.172 −0.006 0.011 1.864 0.598 0.216 0.207 0.078 CT −0.117 −0.174 −0.002 0.011 0.876 0.274 0.190 0.184 0.112 HI −0.117 −0.138 0.000 0.018 1.227 0.540 0.243 0.237 0.128 WV −0.115 −0.010 −0.008 0.011 −1.396 0.641 0.339 0.320 0.085 DE −0.110 −0.056 −0.006 0.014 0.343 0.258 0.149 0.140 0.124 MI −0.109 −0.056 −0.007 0.013 0.895 0.380 0.285 0.271 0.042 VA −0.107 −0.161 −0.006 0.012 1.047 0.331 0.170 0.158 0.035 ME −0.091 −0.176 −0.007 0.016 0.638 0.350 0.274 0.258 0.088 NV −0.072 −0.108 −0.006 0.012 1.061 0.363 0.143 0.136 0.035 GA −0.060 −0.090 −0.004 0.009 1.134 0.542 0.179 0.174 0.030 LA −0.040 −0.062 −0.006 0.012 1.362 0.401 0.249 0.235 0.071 TN −0.038 −0.084 −0.003 0.010 1.489 0.625 0.217 0.211 0.025 IL −0.031 −0.128 −0.005 0.012 2.721 0.417 0.217 0.209 0.051 SC −0.030 −0.119 −0.003 0.014 2.461 0.444 0.239 0.235 0.065 MN −0.030 −0.145 −0.009 0.015 1.642 0.308 0.236 0.222 0.034 VT −0.027 −0.108 −0.007 0.012 1.195 0.474 0.269 0.255 0.118 MA −0.020 −0.187 0.001 0.011 1.241 0.616 0.216 0.214 0.115 NM −0.020 −0.109 −0.017 0.023 3.038 0.465 0.263 0.241 0.053 RI −0.017 −0.097 −0.003 0.016 1.175 0.575 0.262 0.253 0.149 WA −0.012 −0.144 0.001 0.017 0.807 0.287 0.226 0.219 0.048 MT −0.009 −0.056 −0.016 0.014 1.186 0.497 0.311 0.281 0.078 IN −0.007 −0.037 −0.007 0.007 0.677 0.452 0.219 0.208 0.033 NE −0.006 0.012 −0.004 0.012 0.091 0.629 0.262 0.250 0.025 AZ −0.004 −0.069 −0.004 0.010 1.010 0.465 0.182 0.177 0.020 TX −0.003 −0.067 −0.006 0.008 0.462 0.519 0.167 0.157 0.019 KY 0.002 −0.039 −0.005 0.015 1.412 0.390 0.244 0.233 0.078 AR 0.005 −0.093 −0.010 0.013 0.857 0.510 0.231 0.211 0.031 AK 0.009 0.047 −0.040 0.077 −1.561 0.141 0.447 0.360 0.162 MO 0.011 −0.066 −0.009 0.013 1.415 0.314 0.217 0.199 0.039 NH 0.014 −0.042 −0.002 0.008 0.502 0.620 0.176 0.171 0.111 FL 0.031 −0.140 −0.005 0.012 0.758 0.304 0.198 0.187 0.036 KS 0.050 −0.081 −0.006 0.010 1.464 0.493 0.233 0.221 0.021 OK 0.055 −0.053 −0.008 0.012 0.215 0.591 0.245 0.227 0.059 MD 0.055 −0.133 −0.006 0.014 0.577 0.328 0.210 0.197 0.067 AL 0.066 −0.104 −0.004 0.013 1.771 0.380 0.253 0.247 0.047 CA 0.068 −0.236 −0.007 0.018 2.310 0.375 0.237 0.223 0.044 ID 0.076 −0.074 −0.013 0.015 1.493 0.188 0.223 0.201 0.040 NC 0.099 −0.051 −0.008 0.011 2.079 0.418 0.187 0.174 0.026 ND 0.101 −0.007 −0.016 0.015 −2.242 0.449 0.287 0.256 0.048 SD 0.121 −0.009 −0.010 0.015 0.819 0.154 0.212 0.194 0.079

OR 0.131 −0.092 −0.012 0.024 2.205 0.399 0.251 0.229 0.099 WY 0.148 −0.013 −0.028 0.032 −2.745 0.590 0.303 0.247 0.046 UT 0.227 −0.136 −0.006 0.016 1.649 0.614 0.210 0.197 0.044

(continued on next page)

Da, Warachka, and Yun 117

TABLE 1 (continued) State-Level Fiscal Policy

Panel B. State Government Deficit Characteristics (continued)

Budget Deficit

Correlation Alternative Std. Revenue/ Average Average

State FPB FPB Mean Dev. Skewness Expenditure Revenue Expenditure Debt

CO 0.248 −0.142 −0.008 0.015 2.734 0.775 0.183 0.170 0.026 MS 0.278 −0.023 −0.009 0.015 1.063 0.542 0.300 0.281 0.047

Average −0.014 −0.106 −0.008 0.016 1.117 0.442 0.238 0.223 0.061 FPB correlation 0.532 −0.223 0.040 −0.180 0.140 −0.089 −0.154 −0.178

Countercyclical −0.239 −0.253 −0.007 0.019 2.163 0.414 0.271 0.254 0.053 Procyclical 0.206 −0.081 −0.012 0.020 0.981 0.584 0.249 0.225 0.052

Panel C. Predicted Fiscal Policy Betas

Alternative Variable FPB FPB

DR −0.0386*** −0.0300*** −16.90 −16.13

REP_GOV 0.2939*** 0.2007*** 33.05 29.88

Intercept −0.1137*** −0.1668*** −17.79 −43.84

No. of obs. 2,208 2,208 Adj. R2 0.359 0.385

Panel A of Table 1 reports that a Republican is more likely to be governor in a procyclical state based on the 0.517 correlation between FPB and the fraction of the sample period during which a state’s governor is a Republican. Procyclical states are more likely to have periods of negative GSP growth according to BUST Fraction.

The fiscal policy betas have a 0.243 correlation with the volatility of eco- nomic growth and a 0.261 correlation with consumption volatility. These positive correlations suggest that countercyclical states have less variability in their eco- nomic output and consumption. We control for GSP volatility in our later empir- ical tests and examine consumption risk in the next section. Besides economic uncertainty, we construct a proxy for political uncertainty using the number of state-level transitions in the governor’s political affiliation. The relation between a governor’s political affiliation and fiscal policy is confirmed in a later empirical test.

States have adopted budget-stabilization funds to mitigate both the finan- cial constraints imposed by their respective balanced-budget amendments and the policy uncertainty that could arise from higher budget deficits. In unreported re- sults, state-level balanced-budget amendments were previously used to explain state-level fiscal policy. However, these budget amendments were often adopted before the U.S. Civil War. In contrast, budget-stabilization funds reflect more contemporary political decisions. Because the deposit rules governing a state’s budget-stabilization fund do not vary with economic conditions, these rules are suitable instruments for state-level fiscal policy. Wagner and Elder (2005) de- scribe the deposit rules governing these “rainy day” accounts and enumerate their stringency on a scale of 1 (weak) to 4 (strong). Along with Knight and Levinson

118 Journal of Financial and Quantitative Analysis

(1999), these authors conclude that the deposit rules of budget stabilization funds institutionalize government savings.

In practice, deposit rules usually require a minimum percentage of a state’s revenue to be deposited into its budget-stabilization fund. Following withdrawals from the budget-stabilization fund, the deposit rules also specify repayment pro- visions to ensure the fund is replenished to a prescribed maximum. Intuitively, stringent deposit rules facilitate more countercyclical fiscal policies through the accumulation of precautionary savings.

Panel B of Table 1 summarizes the mean, standard deviation, and skewness of state-level budget deficits. Because positive values of DEFICIT signify a budget deficit, countercyclical states have smaller average budget surpluses. This prop- erty can be explained by their larger budget-stabilization fund deposits, which are included in government expenditures.7 Moreover, the skewness results indicate that countercyclical states are willing to tolerate larger budget deficits. The cor- relation between state-level government revenue and expenditures in Panel B is also higher in countercyclical states. Intuitively, maintaining or increasing expen- ditures in poor economic conditions reduces the correlation between a counter- cyclical state’s expenditures and its revenue while inducing a positive skewness in its budget deficit.

The −0.089 correlation between FPB and revenue and the −0.154 correla- tion between FPB and expenditures in Panel B of Table 1 suggest that counter- cyclical states have higher expenditures and taxes as a fraction of their respective GSPs. Although the −0.178 correlation between a state’s fiscal policy beta and its outstanding debt-to-GSP ratio is consistent with countercyclical states hav- ing more debt, this debt averages 6.1% of GSP due to balanced-budget amend- ments. This average is far below the federal government’s debt-to-GDP ratio. Consequently, countercyclical fiscal policies at the state level are unlikely to in- duce high levels of tax uncertainty.

The limited cross-sectional variation in debt across states can be attributed to budget-stabilization funds financing countercyclical fiscal policies. The deposit rules (DR) of each state’s respective budget-stabilization fund and its tendency to elect a Republican governor (REP GOV) are examined in the context of fiscal policy using the following cross-sectional regression:

(3) FPBi = α0 +α1DRi +α2REP GOVi +�i.

We define the fitted values from equation (3) as predicted fiscal policy betas, which are denoted F̂PB. Although F̂PB depends on the frequency of electing a Republican governor during the entire sample period, there is no look-ahead bias in later empirical tests because our empirical analysis does not formulate a trad- ing strategy to exploit cross-sectional return differences based on fiscal policy. In- stead, the cross-sectional relation between fiscal policy and returns over the long term is the focus of our study.

7 Withdrawals from the budget-stabilization fund can increase a state government’s expenditures, and hence its budget deficit, without the corresponding issuance of debt. However, because with- drawals depend on the previous accumulation of funds, we focus our analysis on deposit rules.

Da, Warachka, and Yun 119

The negative coefficient of −0.0386 (t -statistic =−16.90) for DR in Panel C of Table 1 indicates that more stringent deposit rules are associated with counter- cyclical fiscal policies. Furthermore, the positive coefficient of 0.2939 (t -statistic = 33.05) for REP GOV indicates that Democratic governors are also associated with countercyclical fiscal policies. Ranking states according to FPB and F̂PB produces a similar ordering. Indeed, 10 of the top 15 countercyclical states are identical under both rankings, and their correlation exceeds 0.50. The alternative fiscal policy betas are also similar to their predicted counterparts. Thus, investors are not required to estimate their state’s fiscal policy beta provided they under- stand the two important determinants of its fiscal policy that explain nearly 40% of the variation in state-level fiscal policy according to the R2 metrics in Panel C.

III. Fiscal Policy and Consumption Risk According to Ricardian equivalence, government budget deficits cannot

stimulate consumption because households increase savings in anticipation of fu- ture tax increases. However, higher government spending can stabilize consump- tion during poor economic conditions due to household borrowing constraints (Zeldes (1989)) and difficulties in hedging unemployment (Hubbard, Skinner, and Zeldes (1994)). Moreover, at the state level, balanced-budget amendments and budget-stabilization funds reduce the association between debt issuance and countercyclical fiscal policy.

To determine whether countercyclical fiscal policies lower consumption risk, our state-level proxy for consumption is retail sales (Ostergaard, Sorensen, and Yosha (2002)). Following Korniotis (2008), retail sales data on non-durables are scaled upward to account for services. The annual scale factor equals aggregate per capita consumption in the United States divided by average per capita retail sales.

To examine the relation between state-level consumption risk and fiscal pol- icy, we estimate the following panel regression:

(4) CONSUMPTIONi ,t = β11BUSTi ,t +β2FPBi +β3(1BUSTi ,t ×FPBi )+�i ,t.

The indicator variable 1BUSTi ,t equals 1 in year t if state i has negative GSP growth relative to year t−1. The specification in equation (4) also includes year fixed effects and state fixed effects, with the inclusion of state fixed effects requiring the removal of the state-level fiscal policy beta. This specification is also esti- mated after replacing the FPB values with their predicted counterparts F̂PB from equation (3). Recall that the predicted fiscal policy betas capture salient aspects of each state’s fiscal policy that investors can condition on when evaluating their consumption risk.8

A negative β1 coefficient indicates that consumption growth decreases during bust periods. More important, a negative β3 coefficient indicates that consumption declines less during bust periods in states that implement countercyclical fiscal

8 The estimation is conducted using a generalized method of moments (GMM) procedure with the second stage accounting for the estimation error in the first stage. However, the results are similar using 2-stage least squares (2SLS) and limited information maximum likelihood (LIML).

120 Journal of Financial and Quantitative Analysis

policies. According to Panel A of Table 2, without state fixed effects, the β3 co- efficient for the interaction term in equation (4) is negative, equaling −0.1215 (t -statistic =−5.03). Thus, a more countercyclical fiscal policy is associated with a smaller decline in consumption during bust periods, and hence lower consump- tion risk. With the fiscal policy beta having a standard deviation of 0.121, the

TABLE 2 Fiscal Policy and Consumption Risk

Panel A of Table 2 reports on the sensitivity of state-level consumption to each state’s fiscal policy beta, as estimated by equation (4):

CONSUMPTIONi,t = β11BUSTi,t +β2FPBi +β3(1BUSTi,t ×FPBi )+�i,t .

The indicator variable BUST equals 1 if gross state product (GSP) growth in state i is negative in year t relative to the prior year, and 0 otherwise. The panel regression in equation (4) is also estimated using each state’s predicted fiscal policy beta from equation (3), F̂PB. These specifications include year fixed effects and state fixed effects, with state fixed effects requiring the removal of the state-level fiscal policy betas. Panel B reports the results of a similar panel regression that replaces a state’s fiscal policy beta with its political uncertainty. POLITICAL_UNCERTAINTY in a state equals the number of transitions between Democratic and Republican (Republican and Democratic) governors during our sample. Standard errors are clustered at the state-level. t-statistics are reported below each of the estimates in italics. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Panel A. Fiscal Policy and Consumption Risk

Variable 1 2 3 4 5 6 7 8

BUST −0.0280*** −0.0294*** −0.0265*** −0.0264*** −0.0380*** −0.0400*** −0.0464*** −0.0507*** −5.87 −5.20 −5.29 −5.02 −6.74 −7.29 −6.94 −8.56

FPB 0.0262*** 3.32

BUST × FPB −0.1215*** −0.1300*** −5.03 −4.97

F̂PB 0.0365** 1.98

BUST × F̂PB −0.1580*** −0.1908*** −3.93 −5.34

ALTERNATIVE_FPB 0.0304** 2.09

BUST × −0.1433*** −0.1733*** ALTERNATIVE_FPB −3.65 −4.93

ALTERNATIVE_F̂PB 0.0473* 1.80

BUST × −0.2074*** −0.2532*** ALTERNATIVE_F̂PB −3.72 −5.13

Year fixed effects Yes Yes Yes Yes Yes Yes Yes Yes State fixed effects No Yes No Yes No Yes No Yes

No. of obs. 2,300 2,300 2,300 2,300 2,300 2,300 2,300 2,300 Adj. R2 0.725 0.754 0.724 0.754 0.724 0.754 0.724 0.754

Panel B. Political Uncertainty and Consumption Risk

Variable 1 2 3 4

BUST −0.0208** −0.0118 −0.0225** −0.0220** −2.11 −0.96 −2.20 −2.05

POLITICAL_UNCERTAINTY 0.0007 0.0006 0.0007 0.84 0.76 0.82

BUST × POLITICAL_UNCERTAINTY −0.0020 −0.0044* −0.0019 −0.0020 −0.90 −1.78 −0.86 −0.86

FPB 0.0257* 1.88

F̂PB 0.0397 1.62

Year fixed effects Yes Yes Yes Yes State fixed effects No Yes No No

No. of obs. 2,300 2,300 2,300 2,300 Adj. R2 0.721 0.753 0.723 0.723

Da, Warachka, and Yun 121

−0.1215 coefficient implies a 1.47% smaller decline in consumption during bust periods in the more countercyclical state.

The predicted fiscal policy betas confirm the importance of fiscal policy to consumption risk. The interaction variable involving F̂PBi has a negative β3 coef- ficient equaling −0.1580 (t -statistic =−3.93) when F̂PB rather than state fixed effects are included in the specification. The alternative fiscal policy betas produce an even smaller decline in consumption during bust periods because the interac- tion variable has a negative β3 coefficient equaling −0.1433 (t -statistic =−3.65).

State governments may attempt to optimize their fiscal policy based on a trade-off between higher consumption growth versus lower consumption risk. This trade-off is consistent with the positive β2 coefficients in Panel A of Table 2, which indicate that consumption growth is higher in procyclical states. Intuitively, lower consumption growth may be the cost of a countercyclical fiscal policy that lowers consumption risk. Because the magnitude of the β3 coefficient is 4.64 times larger than that of the β2 coefficient, consumption growth is approximately 0.32% lower per year to obtain the 1.47% smaller decline in consumption during bust periods.

Finally, to capture the policy uncertainty associated with countercyclical fis- cal policies, we replace FPB in equation (4) with the political uncertainty vari- able that measures the number of transitions in each state governor’s political affiliation.9

The results in Panel B of Table 2 indicate that the interaction between po- litical uncertainty and the indicator function for bust periods has an insignifi- cant coefficient. Indeed, the β2 and β3 coefficients are consistently insignificant. Therefore, unlike fiscal policy, political uncertainty does not appear to influence consumption growth or consumption risk.

IV. Return Implications of Fiscal Policy Variation in consumption risk across states can have cross-sectional return

implications if investors have a local investment bias. Insufficient risk sharing within the United States due to poor diversification has been reported by Korniotis (2008) and Korniotis and Kumar (2013).

We first compute a firm-level local investor base measure. Individual firms are matched with specific states using the location of their headquarters in Compustat. The location of institutional investors is obtained from Nelson’s Directory of Investment Managers. Korniotis and Kumar (2013) document that individual investors have a similar local bias as institutional investors. This find- ing is consistent with the portfolio decisions of institutional investors reflecting the investment preferences of their clients.

An institution has a local investment bias in a nearby firm (headquartered in the same state) if its aggregate portfolio weight in this firm is at least 20% greater

9 To clarify, the frequency of electing a Republican governor differs from political uncertainty because both a low and a high frequency suggest low political uncertainty. Furthermore, a frequency near 50% does not indicate whether voters change the political affiliation of their governor once, in a permanent political shift from one political party to another, or whether the governor’s political affiliation alternates between elections and causes policy uncertainty.

122 Journal of Financial and Quantitative Analysis

than its market portfolio weight. A firm has a local investor base if at least 5% of its shares are held by institutions with a local investment bias.

To determine whether countercyclical fiscal policies lower the risk-adjusted returns of firms with a local investor base, the following panel regression is estimated:

(5) RETURNk,t = β0 +β1FPBi +γ SFt +�k,t.

The dependent variable refers to firm-level returns that have a local investor base in state i . Thus, all firms indexed by k are headquartered in state i . Returns are risk-adjusted using the methodology of Daniel, Grinblatt, Titman, and Wermers (1997), although unreported results are similar for industry-adjusted returns. The fiscal policy betas in equation (5) are replaced with their respective predicted values from equation (3). SFt contains state and firm characteristics. To capture cross-sectional variation in state-level industrial composition, the state character- istics are average economic growth and the volatility of economic growth (mea- sured as the standard deviation of annual GSP growth). The firm characteristics, which are computed each year, include institutional ownership (IO), capital asset pricing model beta (CAPM BETA), idiosyncratic return volatility (IVOL), book- to-market (BM) ratio, market capitalization (SIZE), and the firm’s return over the prior 12 months after omitting the most recent month (PRET).

A positive β1 coefficient indicates that stock returns are lower in counter- cyclical states for firms with a local investor base. Specifically, a positive β1 co- efficient is consistent with the discount-rate channel. To clarify, the discount-rate channel consists of a joint hypothesis. First, countercyclical fiscal policies are pre- dicted to lower the consumption risk of investors. Second, this lower consumption risk is predicted to lower the returns of firms that have a local investor base.

Panel A of Table 3 reports the results from equation (5) for firms whose insti- tutional investor base has at least a 5% local bias. On average, this local-investor- base restriction yields 666 firms in 43 states. The results in Panel B further restrict this subset to a 7% minimum local base that contains an average of 491 firms in 41 states. The discount-rate channel predicts a larger β1 coefficient in Panel B compared with Panel A under the more stringent local base threshold. Conversely, in Panel C, the estimation is performed on all firms regardless of whether their in- vestor base has a local bias. Thus, the discount-rate channel predicts a smaller β1 coefficient in Panel C compared with Panel A.

According to Panel A of Table 3, β1 equals 0.0583 (t -statistic = 2.54) in the full specification with all control variables. Thus, for firms with a local investor base, a higher fiscal policy beta, which corresponds to a less countercyclical fiscal policy (more procyclical fiscal policy), is associated with higher returns. The β1 coefficient increases to 0.1197 (t -statistic = 4.41) when FPB is replaced by its predicted value, F̂PB.

In terms of the β1=0.0583 coefficient’s economic significance in Panel A of Table 3, the standard deviation of FPB in Table 1 across all 50 states is 0.121. Thus, a 1-standard-deviation difference in the fiscal policy beta implies an annual return difference of 0.70% if at least 5% of a firm’s investor base has a local bias. This deviation is approximately the difference between the 10th and 90th percentiles of FPB. Furthermore, the 0.445 difference in the average fiscal policy

Da, Warachka, and Yun 123

beta between the 5 most countercyclical states and the 5 most procyclical states produces a return difference of 2.59% per year. Thus, the impact of fiscal policy on a firm’s cost of equity is economically significant.

For firms with a more salient local investor base, the β1 coefficient increases in magnitude to 0.0919 (t -statistic = 2.31) according to Panel B of Table 3. Therefore, countercyclical fiscal policies appear to lower stock returns via the discount-rate channel, with a more salient local investment bias increasing the sensitivity of stock returns to state-level fiscal policy. A similar increase in the β1 coefficient is found for the predicted fiscal policy betas. The alternative

TABLE 3 Fiscal Policy and Stock Returns

Table 3 reports the results from the Fama–MacBeth (1973) regression in equation (5): RETURNk,t = β0 +β1FPBi +γSFt +�k,t ,

which examines the return implications of fiscal policy for firms whose investors have a local investment bias. The state- level fiscal policy betas are measured according to equation (2) as the sensitivity of annual budget deficits to economic conditions. An alternative fiscal policy beta (ALTERNATIVE_FPB) is estimated by including a state’s unemployment rate and its growth in personal income as independent variables in equation (2). Predicted fiscal policy betas are defined in equation (3) using budget-stabilization fund deposit rules and the frequency of Republican governors. SFt contains state and firm characteristics. The state characteristics include each state’s average GSP growth (GSP_GROWTH) and the volatility of this growth (GSP_VOLATILITY). Annual firm characteristics include institutional ownership (IO), capital asset pricing model beta (CAPM_BETA), idiosyncratic return volatility (IVOL), book-to-market (BM) ratio, market capitalization (SIZE), and past returns over the prior year (PRET). Institutional investors have a local investment bias if the aggregate portfolio weight they assign to local firms (firms headquartered in the same state as their location) is at least 20% greater than the aggregate market portfolio weight of local firms. In Panel A, firms have a local investor base if at least 5% of their shares are held by institutions with a local investment bias. On average, this subset consists of 666 firms in 43 states. The results in Panel B pertain to a smaller subset averaging 491 firms in 41 states in which at least 7% of a firm’s shares are held by institutions with a local investment bias. Panel C reports the results for the entire sample of firms. t-statistics are below each of the estimates in italics. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Variable 1 2 3 4 5 6 7 8

Panel A. Fama–MacBeth Regression with 5% Local Bias

FPB 0.0606** 0.0583** 2.30 2.54

F̂PB 0.1197*** 0.1117*** 4.41 3.92

ALTERNATIVE_FPB 0.0846* 0.0742* 1.78 1.67

ALTERNATIVE_F̂PB 0.1740*** 0.1641*** 5.40 4.89

IO −0.0977*** −0.0997*** −0.1009*** −0.0997*** −0.0966*** −0.0989*** −0.1007*** −0.0990*** −3.59 −3.82 −3.89 −3.82 −2.83 −3.03 −3.11 −3.03

CAPM_BETA 0.0048 0.0061 0.0066 0.0061 0.0068 0.0080 0.0080 0.0079 0.28 0.35 0.39 0.35 0.45 0.51 0.53 0.51

IVOL −0.0573 −0.0561 −0.0558 −0.0562 −0.0527 −0.0512 −0.0507 −0.0513 −1.56 −1.50 −1.54 −1.51 −1.59 −1.51 −1.55 −1.51

GSP_GROWTH 0.1663 0.2128 0.1387 0.2033 0.1130 0.1917 0.1022 0.1796 0.65 0.80 0.48 0.76 0.44 0.74 0.35 0.68

GSP_VOLATILITY −0.0370 0.2144 0.1214 0.1833 −0.0538 0.2093 0.2014 0.1740 −0.06 0.31 0.10 0.27 −0.09 0.30 0.17 0.25

BM −0.0026 −0.0032 −0.0032 −0.0033 −0.37 −0.45 −0.45 −0.45

SIZE 0.0017 0.0015 0.0017 0.0015 0.85 0.80 0.93 0.80

PRET −0.0024 −0.0024 −0.0020 −0.0024 −0.11 −0.11 −0.09 −0.11

Intercept −0.1244 −0.1828 −0.0909 −0.1547 −0.0927 −0.1823 −0.0810 −0.1522 −0.47 −0.65 −0.30 −0.54 −0.36 −0.67 −0.27 −0.55

Adj. R2 0.028 0.027 0.028 0.027 0.039 0.038 0.039 0.038 (continued on next page)

124 Journal of Financial and Quantitative Analysis

TABLE 3 (continued) Fiscal Policy and Stock Returns

Variable 1 2 3 4 5 6 7 8

Panel B. Fama–MacBeth Regression with 7% Local Bias

FPB 0.0901** 0.0919** 2.12 2.31

F̂PB 0.1707*** 0.1674*** 6.34 6.10

ALTERNATIVE_FPB 0.1485*** 0.1451*** 2.58 2.78

ALTERNATIVE_ F̂PB 0.2433*** 0.2410*** 7.45 7.14

IO −0.1025*** −0.1041*** −0.1053*** −0.1044*** −0.1031*** −0.1055*** −0.1069*** −0.1059*** −5.14 −5.73 −5.85 −5.73 −3.94 −4.55 −4.69 −4.56

CAPM_BETA 0.0060 0.0077 0.0088 0.0076 0.0104 0.0120 0.0130 0.0119 0.43 0.55 0.67 0.55 0.74 0.84 0.94 0.83

IVOL −0.0644 −0.0636 −0.0623 −0.0638 −0.0618 −0.0608 −0.0593 −0.0609 −1.37 −1.33 −1.31 −1.33 −1.33 −1.28 −1.26 −1.29

GSP_GROWTH 0.1610 0.2008 0.1542 0.1923 0.1996 0.2682 0.2074 0.2574 0.54 0.61 0.52 0.59 0.67 0.86 0.73 0.83

GSP_VOLATILITY −0.3149 0.1909 −0.3429 0.1447 −0.3861 0.1801 −0.3436 0.1227 −0.65 0.20 −0.22 0.16 −0.81 0.19 −0.24 0.13

BM 0.0070 0.0057 0.0059 0.0057 0.77 0.60 0.64 0.60

SIZE 0.0027 0.0023 0.0025 0.0023 1.10 1.06 1.19 1.08

PRET −0.0044 −0.0046 −0.0040 −0.0046 −0.24 −0.25 −0.22 −0.25

Intercept −0.1086 −0.1684 −0.0837 −0.1340 −0.1901 −0.2759 −0.1798 −0.2393 −0.33 −0.48 −0.27 −0.39 −0.61 −0.85 −0.62 −0.75

Adj. R2 0.032 0.032 0.033 0.032 0.045 0.044 0.046 0.044

Panel C. Fama–MacBeth Regression with All Firms

FPB 0.0275 0.0288 1.26 1.32

F̂PB 0.0326 0.0276 1.15 0.94

ALTERNATIVE_FPB 0.0092 0.0015 0.24 0.04

ALTERNATIVE_F̂PB 0.0581 0.0518 1.50 1.32

IO −0.1131*** −0.1139*** −0.1143*** −0.1140 −0.1202*** −0.1207*** −0.1210*** −0.1208*** −5.01 −5.04 −5.12 −5.03 −4.07 −4.10 −4.16 −4.09

CAPM_BETA 0.0124 0.0125 0.0126 0.0125 0.0108 0.0109 0.0109 0.0110 0.75 0.76 0.78 0.76 0.67 0.67 0.69 0.67

IVOL −0.0557** −0.0556** −0.0556** −0.0556** −0.0562** −0.0560** −0.0561** −0.0560** −2.43 −2.43 −2.45 −2.43 −2.21 −2.20 −2.22 −2.20

GSP_GROWTH 0.7218*** 0.8012** 0.8082*** 0.7887** 0.7113*** 0.8020** 0.8016** 0.7891** 2.74 2.57 2.66 2.49 2.62 2.49 2.57 2.41

GSP_VOLATILITY −0.8232 −0.6895 −0.6635 −0.7277 −0.8418 −0.6757 −0.5971 −0.7161 −1.29 −1.14 −0.86 −1.22 −1.36 −1.14 −0.79 −1.22

BM −0.0034 −0.0035 −0.0035 −0.0035 −0.94 −0.93 −0.96 −0.94

SIZE 0.0036 0.0036 0.0036 0.0036 1.12 1.09 1.09 1.09

PRET 0.0009 0.0011 0.0011 0.0011 0.07 0.09 0.09 0.09

Intercept −0.6926** −0.7803** −0.7890** −0.7598** −0.7243** −0.8243** −0.8286** −0.8038** −2.52 −2.33 −2.32 −2.24 −2.35 −2.21 −2.21 −2.13

Adj. R2 0.018 0.017 0.017 0.017 0.023 0.022 0.022 0.022

Da, Warachka, and Yun 125

fiscal policy betas and their predicted counterparts also have larger coefficients in Panel B under the more stringent local-investor-base criteria.

The results in Panel C of Table 3 pertain to the entire cross section of firms, including firms whose investor base does not have a local bias. Thus, the discount- rate channel predicts a smaller β1 coefficient in this unrestricted cross section. Consistent with the discount-rate channel, the β1 coefficients are insignificant in Panel C for every specification. This lack of statistical significance, despite the larger sample size, highlights the importance of local investment bias to the rela- tion between fiscal policy and firm-level returns, thereby lending support to the discount-rate channel.

In unreported returns, we estimate a panel regression of firm-level stock re- turns on state-level consumption volatility (standard deviation of annual consump- tion growth). Higher consumption volatility does increase the equity premium for firms with a local investor base, as the coefficient for consumption volatility is positive. However, this analysis does not explicitly condition on fiscal policy. Indeed, besides fiscal policy, consumption volatility is influenced by other factors, such as economic volatility.

A. Political Uncertainty A large literature finds that government intervention in the economy can in-

crease policy uncertainty. For example, government budget deficits can increase uncertainty regarding future taxation. According to Panel A of Table 1, polit- ical uncertainty has a −0.191 correlation with FPB. This negative correlation indicates that countercyclical states have greater political uncertainty. However, there is no relation between political uncertainty and either consumption or out- put volatility. In particular, the correlation between political uncertainty and con- sumption volatility is −0.012, whereas the correlation between political uncer- tainty and GSP volatility is 0.021.

Although our study examines the return predictability attributable to the pre- dictable component of fiscal policy, we reestimate the Fama–MacBeth (1973) re- gression in equation (5) with the political uncertainty variable included as an ad- ditional control. Table 4 reports that controlling for political uncertainty does not eliminate the importance of fiscal policy to stock returns. At the 5% local bias threshold, the β1 coefficient for FPB remains positive, equaling 0.0688 (t -statistic = 2.33) after the inclusion of political uncertainty. The magnitude of this coeffi- cient does not decrease with the inclusion of political uncertainty. Furthermore, increasing the local bias threshold in the investor base to 7% confirms the impor- tance of fiscal policy. In particular, both FPB and its predicted counterpart F̂PB have significant β1 coefficients. These results also apply to the alternative fiscal policy betas that control for unemployment and income growth.

The impact of political uncertainty on returns is less consistent. In particu- lar, both political uncertainty and the predicted fiscal policy betas have a common dependence on the political affiliation of a state’s governor. Nonetheless, the re- sults at the 7% local bias threshold suggest that fiscal policy is more important to stock returns than political uncertainty. For example, the coefficient for F̂PB is positive, 0.1280 (t -statistic = 2.06), whereas the coefficient for political uncer- tainty is insignificant, 0.0038 (t -statistic = 1.16).

126 Journal of Financial and Quantitative Analysis

TABLE 4 Political Uncertainty, Fiscal Policy, and Stock Returns

Table 4 reports the results from the Fama–MacBeth (1973) regression in equation (5) that examines the return implica- tions of fiscal policy with an additional control variable for political uncertainty. The political uncertainty variable (POLITI- CAL_UNCERTAINTY) equals the number of state-level transitions from a Republican governor to a Democratic governor and, vice versa, from a Democratic governor to a Republican governor. State-level fiscal policy betas are measured according to equation (2) as the sensitivity of annual budget deficits to economic conditions. An alternative fiscal policy beta (ALTERNATIVE_FPB) is estimated by including a state’s unemployment rate and its growth in personal income as independent variables in equation (2). Predicted fiscal policy betas are defined in equation (3) using budget-stabilization fund deposit rules and the frequency of Republican governors. State characteristics in this analysis include each state’s average gross state product (GSP) growth (GSP_GROWTH) and the volatility of this growth (GSP_VOLATILITY). Annual firm characteristics include institutional ownership (IO), capital asset pricing model beta (CAPM_BETA), idiosyncratic re- turn volatility (IVOL), book-to-market (BM) ratio, market capitalization (SIZE), and past returns over the prior year (PRET). Institutional investors have a local investment bias if the aggregate portfolio weight they assign to local firms (firms head- quartered in the same state as their location) is at least 20% greater than the aggregate market portfolio weight of local firms. The 5% and 7% minimum thresholds apply to the percentage of shares held by institutions with a local investment bias. t-statistics are below each of the estimates in italics. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Panel A. 5% Threshold Panel B. 7% Threshold

Variable 1 2 3 4 5 6 7 8

FPB 0.0688** 0.0786** 2.33 2.01

F̂PB 0.0796 0.1280** 1.80 2.06

ALTERNATIVE_FPB 0.0996* 0.1400*** 1.87 3.00

ALTERNATIVE_F̂PB 0.1301* 0.2025** 1.85 2.60

POLITICAL_UNCERTAINTY 0.0073*** 0.0078** 0.0086** 0.0072 0.0045** 0.0038 0.0057** 0.0029 2.79 2.00 2.28 1.53 2.46 1.16 2.72 0.92

IO −0.110*** −0.1129*** −0.1141*** −0.1130*** −0.1054*** −0.1093*** −0.1109*** −0.1097*** −3.36 −4.01 −3.74 −3.69 −3.96 −4.79 −4.81 −4.80

CAPM_BETA 0.0104 0.0115 0.0118 0.0116 0.0110 0.0120 0.0129 0.0122 0.69 1.29 0.82 0.79 1.11 1.22 1.34 1.23

IVOL −0.0560 −0.0563 −0.0544 −0.0563 −0.0605 −0.0605 −0.0581 −0.0607 −1.41 −1.55 −1.36 −1.35 −1.38 −1.33 −1.33 −1.33

GSP_GROWTH 0.2740 0.2605 0.2079 0.2437 0.3390 0.3635 0.2923 0.3420 1.07 1.22 0.76 0.90 1.39 1.45 1.29 1.36

GSP_VOLATILITY −0.7166 −0.0452 −0.5750 −0.0981 −0.3154 0.2384 −0.5276 0.1576 −1.27 −0.08 −0.53 −0.13 −0.80 0.25 −0.42 0.17

BM 0.0035 0.0028 0.0027 0.0027 0.0073 0.0058 0.0059 0.0058 0.45 0.36 0.33 0.33 0.88 0.67 0.71 0.66

SIZE 0.0024 0.0024 0.0025 0.0024 0.0029 0.0028 0.0029 0.0028 1.17 1.59 1.30 1.22 1.34 1.44 1.54 1.45

PRET −0.0033 −0.0035 −0.0031 −0.0035 −0.0049 −0.0048 −0.0048 −0.0049 −0.16 −0.17 −0.15 −0.17 −0.27 −0.26 −0.27 −0.26

Intercept −0.2825 −0.2944 −0.2109 −0.2588 −0.3631 −0.4005 −0.2914 −0.3509 −1.09 −1.37 −0.76 −0.93 −1.40 −1.69 −1.39 −1.49

Adj. R2 0.045 0.044 0.044 0.044 0.047 0.047 0.048 0.047

B. Aggregate Fiscal Policy of Investor Base Our next analysis computes a firm-level investor-base beta denoted IB FPB

using institutional investor holdings and the FPB of each investor’s location. The investor-base beta reduces the dependence of our results on local investment bias by allowing a firm’s investor base to span several states. For a firm held by in- vestors located in multiple states, IB FPB weights each state’s fiscal policy beta (predicted fiscal policy beta) by the fraction of its investor base located in the

Da, Warachka, and Yun 127

state.10 Thus, IB FPB examines the aggregate fiscal policy affecting a firm’s in- vestor base instead of focusing on the fiscal policy affecting its local investors.

Because IB FPB is irrelevant if the combined portfolio holdings of institu- tional investors in a firm is too small, institutional ownership is required to be above a 30% threshold. This minimum threshold parallels the earlier firm-level local bias filter that required institutional investors to overweight a firm relative to its market portfolio weight. The 30% threshold creates a subset of 1,252 stocks in 48 states on average. This subset is considerably larger than the local-investor- base subset underlying Panel A of Table 3.11 A more stringent 40% minimum for institutional ownership reduces the subset to an average of 905 firms across 48 states.

We then reestimate equation (5) with the firm-level IB FPB replacing the state-level FPB:

(6) RETURNk,t = β0 +β1IB FPBk +γ SFt +�k,t.

This specification does not assume any local investor bias. Instead, any firm with sufficient institutional ownership is included in this analysis, with IB FPB repre- senting the aggregate fiscal policy of its investor base.

A positive β1 coefficient indicates that returns are lower for firms whose investor bases are located in countercyclical states. Panel A of Table 5 reports positive β1 coefficients for firms with at least 30% institutional ownership. In the full specification with all control variables, the β1 coefficient is 0.3170 (t -statistic = 4.40). This β1 coefficient increases to 1.0309 (t -stat. = 4.85) for the predicted fiscal policy betas. These β1 coefficients indicate that having an investor base concentrated in countercyclical states lowers a firm’s cost of equity.

Panel B of Table 5 reports larger β1 coefficients when institutional owner- ship is required to exceed a 40% minimum threshold. Specifically, the β1 coef- ficient increases to 0.4020 (t -statistic = 4.90). The stronger result is consistent with IB FPB capturing the fiscal policy exposure of a firm’s investor base more accurately because institutional portfolio holdings are more relevant in this subset of firms.

C. Investment Sensitivity to Fiscal Policy Although our study is intended to examine the investor-level implications

of fiscal policy, state-level fiscal policy may have implications for corporate investment. The lower cost of equity in countercyclical states versus the higher consumption growth in procyclical states creates a trade-off for firms.

However, firms can obtain equity financing from investors throughout the United States. Thus, the fiscal policy of the state in which a firm is headquar- tered does not necessarily determine its cost of equity. Indeed, a firm’s headquar- ters may appear in a procyclical state due to its higher consumption growth, ac- cess to natural resources, favorable regulation, or other considerations. With a local investment bias, firms headquartered in a procyclical state have a higher

10 Our analysis of long-term returns allows for heterogeneity in the prevailing economic conditions of states during each period. This heterogeneity is more important for the cash-flow channel.

11 In unreported results, intersecting the 5% minimum local-investment-bias filter with the 30% institutional ownership filter reduces the average number of stocks from 1,252 to 461.

128 Journal of Financial and Quantitative Analysis

TABLE 5 Aggregate Fiscal Policy of Investor Base and Stock Returns

Table 5 reports the results from the Fama–MacBeth (1973) regression in equation (6) that examines the return implications of a firm’s aggregate exposure to fiscal policy through its investor base. The location of every institutional investor that owns a firm’s stock determines its investor-base fiscal policy beta, denoted IB_PB. Specifically, this firm-specific aggre- gate fiscal policy beta is computed by weighting the fiscal policy beta (predicted fiscal policy beta) of each state by the percentage of a firm’s institutional investors in the respective state. The state-level fiscal policy betas are measured according to equation (2) as the sensitivity of annual budget deficits to economic conditions. Predicted fiscal policy betas are defined in equation (3) using budget-stabilization fund deposit rules and the frequency of Republican governors. State characteristics in this analysis include each state’s average gross state product (GSP) growth (GSP_GROWTH) and the volatility of this growth (GSP_VOLATILITY). Annual firm characteristics include institutional ownership (IO), capital asset pricing model beta (CAPM_BETA), idiosyncratic return volatility (IVOL), book-to-market (BM) ratio, market capitalization (SIZE), and past returns over the prior year (PRET). The results in Panel A impose a 30% minimum threshold on insti- tutional investment, whereas those in Panel B impose a 40% threshold. t-statistics are below each of the estimates in italics. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Variable 1 2 3 4

Panel A. Institutional Ownership Above 30%

IB_FPB 0.3093*** 0.3170*** 4.24 4.40

ÎB_FPB 1.0123*** 1.0309*** 4.41 4.85

IO −0.0256** −0.0233* −0.0294*** −0.0279** −2.23 −1.76 −2.84 −2.38

CAPM_BETA −0.0127 −0.0090 −0.0137 −0.0098 −0.76 −0.53 −0.85 −0.62

IVOL −0.1303*** −0.1297*** −0.1194*** −0.1199*** −5.19 −5.04 −4.34 −4.27

GSP_GROWTH 0.2932 0.3289 0.2805 0.3161 1.39 1.44 1.28 1.35

GSP_VOLATILITY −0.3361 −0.3769 −0.3501 −0.3984 −0.85 −1.00 −0.90 −1.05

BM −0.0025 −0.0045 −0.45 −0.81

SIZE 0.0062* 0.0056 1.75 1.62

PRET 0.0173 0.0178 1.33 1.32

Intercept −0.2553 −0.2426 −0.3282 −0.3057 −1.07 −1.00 −1.13 −1.05

Adj. R2 0.031 0.033 0.042 0.044

Panel B. Institutional Ownership Above 40%

IB_FPB 0.4020*** 0.4060*** 4.90 4.89

ÎB_FPB 1.4866*** 1.4818*** 5.82 5.76

IO 0.0262 0.0265 0.0333 0.0331 0.74 0.84 0.92 1.04

CAPM_BETA −0.0292 −0.0224 −0.0305 −0.0235 −1.42 −1.10 −1.57 −1.26

IVOL −0.1585*** −0.1588*** −0.1360*** −0.1337*** −4.28 −4.14 −3.46 −3.47

GSP_GROWTH 0.2044 0.2885 0.2077 0.2786 0.68 0.95 0.69 0.93

GSP_VOLATILITY −0.0431 −0.1787 −0.0251 −0.1473 −0.14 −0.63 −0.08 −0.50

BM −0.0017 −0.0041 −0.28 −0.70

SIZE 0.0112*** 0.0116*** 3.49 4.46

PRET 0.0120 0.0133 0.73 0.80

Intercept −0.1804 −0.1897 −0.3454 −0.3461 −0.57 −0.63 −0.98 −1.05

Adj. R2 0.047 0.051 0.064 0.069

Da, Warachka, and Yun 129

discount rate because (undiversified) local investors require a higher equity return premium. In response, these firms can target equity investors in countercycli- cal states to take advantage of their lower required equity return premium. This investor-base diversification is especially important for capital-intensive firms whose investments are sensitive to their respective discount rates. To clarify, the preference of local investors for local equity securities is not an advantage for firms headquartered in procyclical states because investors in these states require a higher equity return premium.

Because firms headquartered in a procyclical state have the ability to obtain financing from investors in countercyclical states to fund their capital expenditures (CAPEX), we estimate the following specification:

(7) CAPEXk,t = β0 +β1[FPBi − IB FPBk]+γ FCt +�k,t ,

whose dependent variable is capital expenditures normalized by total assets. Firms indexed by k are headquartered in state i . The firm characteristics contained in FC include controls for leverage and market-to-book ratios. Industry fixed effects based on 2-digit Standard Industrial Classification codes are also included, along with year fixed effects. In unreported results, the β1 coefficients are consistently positive. Thus, capital-intensive firms headquartered in a procyclical state (high FPB) attract equity investment from countercyclical states.

Note that firms headquartered in procyclical states with lower investment opportunities are less concerned about local investment bias, whereas capital- intensive firms have a stronger incentive to mitigate local bias in their investor base. Thus, not every firm will aggressively pursue equity financing from investors in countercyclical states.

D. Headquarters Relocations Assuming a local investment bias, relocations of firm headquarters provide

a quasi-natural experiment to analyze the impact of fiscal policy on average stock returns. During the 1986–2006 period, 1,265 firms change the state in which they are headquartered, according to Compact Disclosure. We observe an even split between the number of firms relocating to states with more countercyclical and less countercyclical fiscal policies. Relocations also occur in an array of differ- ent industries. Firms are required to have 3 years of returns before and after their headquarters relocation. The first year’s post-relocation return following a head- quarters relocation is omitted to mitigate the possible influence of takeover pre- miums associated with mergers and acquisitions.

We compute firm-level average returns before and after headquarters reloca- tions. For firms that relocated their headquarters, these average return differences are then regressed on changes in the fiscal policy beta associated with a relocation of firm k’s headquarters from state i to state j :

(8) RETURNk, j ,t+−RETURNk,i ,t− = β0 +β11FPBk,t +γ1Si , j +�k ,

where 1FPBk,t is defined as the difference in firm k’s fiscal policy beta after the relocation compared with before the relocation, FPBk, j ,t+−FPBk,i ,t−. The notation t+ refers to years after the relocation, and t− refers to years before the relocation

130 Journal of Financial and Quantitative Analysis

in year t . Firms can relocate between countercyclical states, with 1FPB being negative and positive, respectively, when the firm relocates to a more counter- cyclical state or a less countercyclical state. 1Si , j denotes state-level differences in GSP growth as well as GSP volatility before and after a headquarters relocation.

The positive β1 coefficients in Table 6 indicate that firms relocating to states with more countercyclical fiscal policies subsequently have lower average stock returns. The predicted fiscal policy betas also have positive β1 coefficients. Overall, the return implications of headquarters relocations confirm the impor- tance of fiscal policy to stock returns. However, the impact of a headquarters relo- cation on a firm’s cost of equity requires a local investor bias before and after the relocation.12

Furthermore, headquarters relocations may coincide with other corporate events. For example, the higher consumption growth in procyclical states may attract firms that benefit from having their headquarters located near customers. Conversely, the lower consumption risk in countercyclical states may attract em- ployees in volatile but high-value-added industries. Intuitively, although a head- quarters relocation decision has discount-rate implications, other important fac- tors, such as human capital (Romer (1990)), are relevant.

TABLE 6 Headquarter Relocations and Stock Returns

Table 6 examines firm-level average returns before and after headquarters relocations. Firm-level return differences are regressed on changes in the fiscal policy betas (FPB) associated with the relocations of firm k from state i to state j as in equation (8):

RETURNk,j ,t+−RETURNk,i,t− = β0 +β11FPBk,t +γ1Si,j +�k .

1FPBk,t is defined as the difference in firm k ’s fiscal policy beta after the relocation compared with before the relocation, FPBk,j ,t+−FPBk,i,t−. The notation t+ refers to years after the relocation, and t− refers to years before the relocation in year t . Firms are required to have at least 3 years of stock returns before and after their headquarters relocations. Returns in the first year after a relocation are eliminated from RETURNk,j ,t+ to ensure that takeover premiums arising from mergers and acquisitions are not influencing our results. 1Si,j denotes state-level differences in gross state product (GSP) growth (GSP_GROWTH), as well as its volatility (GSP_VOLATILITY), before and after a headquarters relocation. t-statistics are below each of the estimates in italics. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Variable 1 2 3 4 5 6 7 8

1FPB 0.2582*** 0.2478** 2.93 2.28

1F̂PB 0.5375** 0.4681* 2.53 1.92

1ALTERNATIVE_FPB 0.3253*** 0.2943* 2.61 1.85

1ALTERNATIVE_F̂PB 1.1477*** 1.0832*** 3.46 2.99

1GSP_GROWTH −0.8057 −0.8069 −0.7219 −0.8272 −0.84 −0.84 −0.75 −0.87

1GSP_VOLATILITY 1.0686 2.0341 1.2809 1.7019 0.50 1.02 0.58 0.89

Intercept 0.0090 0.0061 0.0103 0.0064 0.0072 0.0046 0.0098 0.0060 0.62 0.41 0.71 0.43 0.49 0.31 0.68 0.41

No. of obs. 816 816 816 816 816 816 816 816 Adj. R2 0.010 0.011 0.008 0.010 0.008 0.009 0.014 0.016

12 In unreported results, we find evidence that local investment bias does migrate with firms to their new headquarters locations.

Da, Warachka, and Yun 131

E. Market Participation The Panel Study of Income Dynamics (PSID) survey records state-level eq-

uity market participation rates for 1989, 1994, 1999, 2001, 2003, 2005, 2007, and 2009. Figure 1 illustrates the decline in market participation that corresponds to national recessions identified by the National Bureau of Economic Research (NBER) that occurred during part of the following time periods: 1990–1991, 2001, and 2008–2009.

FIGURE 1 Market Participation

Figure 1 provides a visual illustration of market participation in procyclical states and countercyclical states during national recessions. The Panel Study of Income Dynamics (PSID) survey records state-level market participation rates for 1989, 1994, 1999, 2001, 2003, 2005, 2007, and 2009. National recessions occurred during part of the following time periods: 1990–1991, 2001, and 2008–2009.

0.25

0.27

0.29

0.31

0.33

0.35

0.37

0.39

1989 1991 1994 1999 2001 2003 2005 2007 2008 2009

P ar

tic ip

at io

n

Year

Countercyclical

Procyclical

Recession

Consistent with countercyclical states having lower consumption risk, mar- ket participation is less volatile in countercyclical states as a result of smaller declines during recessions. In particular, the volatility of equity market participa- tion in the most countercyclical states equals 2.85% compared with 5.17% in the most procyclical states. This 2.32% reduction in market participation volatility is significant (t -statistic = 2.59).

Figure 1 provides empirical support for the discount-rate channel. Specifi- cally, the states with the most countercyclical fiscal policy exhibit lower consump- tion risk than the states with the most procyclical fiscal policy. Thus, countercycli- cal fiscal policies appear to finance investor consumption during poor economic conditions that otherwise would have been financed by investors liquidating their equity portfolios.

Along with the importance of investor location, the state-level equity par- ticipation provides evidence supporting the discount-rate channel. Nonetheless, countercyclical fiscal policies may lower stock returns by reducing cash-flow risk. Therefore, our next analyses examine the cash-flow risk channel.

F. Cash-Flow Risk To investigate whether countercyclical fiscal policies lower the sensitivity of

cash flow to economic conditions, we estimate the following panel regression:

(9) CASH FLOWk,t = β11BUSTi ,t +β2FPBi +β3(1BUSTi ,t ×FPBi )+�i ,t.

132 Journal of Financial and Quantitative Analysis

The dependent variable is the ratio of firm-level earnings normalized by total as- sets, for all firms headquartered in state i . This specification parallels equation (4), with state-level consumption replaced by firm-level cash flow. Year fixed effects are included with, standard errors clustered at the firm level. State fixed effects are also included when FPB is removed.

Unlike consumption, Table 7 indicates that cash flow does not decline less during bust periods in countercyclical states than in procyclical states. In particu- lar, the β3 coefficient for the interaction variable is either insignificant or positive. An insignificant β3 coefficient is consistent with the majority of firms in our sam- ple having diversified operations across the United States and with heterogeneity in economic conditions across U.S. states. Furthermore, a positive β3 coefficient indicates that firm-level cash flow declines more rather than less during bust peri- ods in countercyclical states.

To further analyze the cash-flow channel, we utilize the data of Garcia and Norli (2012) to examine the state-level operations of individual firms starting in 1994. Their data record instances where each state’s name occurs in a firm’s 10-K report. Delaware and Washington are removed from the sample given the large number of firms incorporated in Delaware and the possibility that Washington refers to the U.S. capital rather than the state. On average, firms operate in fewer than 10 states.

We compute firm-level cash-flow fiscal policy betas, denoted CF FPB, that weight the state-level fiscal policy betas in which a firm operates by the frequency of each state’s name in its 10-K report. Firms have more cash-flow exposure to a state’s fiscal policy if they have more operations in this state. The cash-flow

TABLE 7 Fiscal Policy and Cash-Flow Risk

Table 7 reports the results from the regression in equation (9): CASH_FLOWk,t = β11BUSTi,t +β2FPBi +β3(1BUSTi,t ×FPBi )+�i,t ,

which examines the firm-level cash-flow implications of fiscal policy during bust periods. Firms indexed by k are head- quartered in state i. Year fixed effects are included, as well as state fixed effects when the state-level fiscal policy betas are excluded. The state-level fiscal policy betas are measured according to equation (2) as the sensitivity of annual bud- get deficits to economic conditions. A bust period for a state is defined by negative gross state product (GSP) growth. An alternative fiscal policy beta is estimated by including a state’s unemployment rate and its growth in personal income as independent variables in equation (2). t-statistics are below each of the estimates in italics. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Variable 1 2 3 4

BUST −0.0052 0.0142* −0.0065 0.0156* −0.53 1.71 −0.70 1.76

FPB −0.0337 −1.00

BUST × FPB 0.0018 0.0979* 0.03 1.84

ALTERNATIVE_FPB −0.0397 −0.89

BUST × ALTERNATIVE_FPB −0.0598 0.0140 −1.03 0.22

Year fixed effects Yes Yes Yes Yes State fixed effects No Yes No Yes

No. of obs. 2,250 2,250 2,250 2,250 Adj. R2 0.080 0.304 0.080 0.304

Da, Warachka, and Yun 133

betas based on fiscal policy enable us to examine the cash-flow channel using the following Fama–MacBeth (1973) regression:

RETURNk,t = β0 +β1CF FPBk,t +β2BMk,t +β3SIZEk,t(10) +β4PRETk,t +γ FCk +�k,t.

Because this Fama–MacBeth regression uses firm-level CF FPB betas, the con- trol variables are book-to-market (BM) ratio, size (SIZE), and past return (PRET) characteristics. The firm-level returns in equation (10) are not risk-adjusted due to the inclusion of these firm characteristics. FCk contains the following firm characteristics: institutional ownership (IO), the capital asset pricing model beta (CAPM BETA), and idiosyncratic return volatility (IVOL).

A significant β1 coefficient is evidence that the cash-flow channel is respon- sible for the return implications of fiscal policy. However, according to Panel A of Table 8, the β1 coefficients are insignificant. The lack of empirical support for the cash-flow channel can be attributed to firms having diversified operations. Indeed, the cash flows of firms with diversified operations are unlikely to be affected by the fiscal policy and economic conditions of any individual state. Therefore, we restrict our next analysis to geographically concentrated firms whose operations are limited to a single state.

However, our results are biased against the cash-flow channel if single-state firms have operations in more than one state. Therefore, we refine the single-state classification of Garcia and Norli (2012) for our purposes by applying two addi- tional filters that eliminate firms using the following two criteria: i) firm describes itself as operating in multiple states but does not specifically mention their names, ii) firm refers to an international sales office. These filters eliminate 58 firms, leaving 419 firms that operate in a single state.

In contrast to our earlier analysis, the cash-flow channel does not require a local investment bias because the cash-flow channel requires local operations. For single-state firms, FPB equals CF FPB. The results in Panel B of Table 8 indicate that the firm-level cash flow betas are significant for firms operating in a single state. In particular, the β1 coefficient equals 0.0062 (t -statistic = 2.21) in the full specification with all control variables. Therefore, we find support for the cash-flow channel among geographically concentrated firms whose operations are limited to a single state. The positive β2 and β4 coefficients for BM and PRET, respectively, are consistent with the value premium and momentum, respectively.

To compare the discount-rate channel versus the cash-flow channel, recall that the β1 coefficient for FPB equals 0.0583 in Panel A of Table 3 with all control variables, whereas a similar specification in Panel B of Table 8 yields a larger β1 coefficient of 0.0744. These analyses focus on firms that have a local investor base and local operations, respectively. The larger coefficient associated with the cash- flow channel is consistent with single-state firms having a higher equity premium due to their undiversified operations. This difference in the equity premium is more pronounced for firms without a local investor base (Panel C of Table 3) but reverses for firms whose investor base is more local (Panel B of Table 3).

Finally, in unreported results, we estimate the panel regression specification in equation (5) for firms that have a local bias at the 5% threshold to examine

134 Journal of Financial and Quantitative Analysis

TABLE 8 Cash-Flow Beta and Stock Returns

Table 8 reports the results from equation (10): RETURNk = β0 +β1CF_FPBk +β2BMk +β3SIZEk +β4PRETk +γFCk +�k ,

which examines the return implications of firm-level cash-flow fiscal policy betas. The firm-level cash-flow fiscal policy betas, denoted CF_FPB, weight the state-level fiscal policy betas according to the operations of a firm. Thus, firms with more operations in a state have more cash-flow exposure to the state’s fiscal policy. Firm-level control variables include book-to-market (BM) ratio, size (SIZE), and past return (PRET) characteristics, where PRET equals the firm’s return over the prior 12 months after omitting the most recent month. FCk contains institutional ownership (IO), capital asset pricing model beta (CAPM_BETA), and idiosyncratic volatility return (IVOL). The results in Panel A pertain to all firms, whereas those in Panel B are for geographically concentrated firms that operate in a single state. t-statistics are below each of the estimates in italics. *, **, and *** indicate statistical significance at the 10%, 5%, and 1% levels, respectively.

Variable 1 2 3

Panel A. All Firms

CF_FPB 0.0432 0.0468 0.0552 0.47 1.12 1.64

BM 0.0252 0.0216** 1.57 2.19

SIZE 0.0000 −0.0036 0.01 −0.61

PRET 0.0444 0.0612** 1.41 2.57

IO −0.0120 −0.36

CAPM_BETA 0.0120 0.47

IVOL −0.0168 −1.63

Intercept 0.0912 0.0552 0.1044 1.41 0.51 1.32

Adj. R2 0.001 0.035 0.062

Panel B. Geographically Concentrated Firms

CF_FPB (FPB) 0.0876** 0.0972*** 0.0744** 2.29 3.06 2.21

BM 0.0252** 0.0372*** 2.48 3.97

SIZE 0.0024 −0.0060 0.37 −0.79

PRET 0.0456** 0.0540*** 2.39 3.58

IO 0.0276 0.47

CAPM_BETA 0.0396 1.61

IVOL −0.0192** −2.07

Intercept 0.1032*** 0.0360 0.1104 3.43 0.44 1.12

Adj. R2 0.005 0.049 0.095

the cash-flow channel among firms with a local investor base rather than local operations. However, we find that the β1 coefficients for CF FPB are insignificant. This evidence confirms that the cash-flow channel is limited to firms with geo- graphically concentrated operations. Therefore, our earlier results in Table 3 that support the discount-rate channel are not driven by firms having local operations.

Da, Warachka, and Yun 135

V. Conclusions Using state-level data within the United States, we find that consumption

risk is lower in states that implement countercyclical fiscal policies. Furthermore, firms whose investor bases are concentrated in countercyclical states have lower average stock returns. Equity market participation is also less volatile in counter- cyclical states as a result of smaller declines in equity market participation during recessions. The consumption and stock return implications of fiscal policy are confirmed using the political affiliation of state governors and the deposit rules of state-level budget-stabilization funds.

Our evidence is consistent with countercyclical fiscal policies being able to lower the consumption risk of investors and, consequently, the equity premium demanded by investors. This discount-rate channel requires geographic variation across firms in the location of their investor bases. An alternative cash-flow chan- nel, which requires geographic variation in firm-level operations, does not explain the relation between fiscal policy and stock returns. Intuitively, the implications of state-level fiscal policy are transmitted into asset prices through the location of a firm’s investor base rather than the location of its operations.

References Acemoglu, D.; S. Johnson; J. Robinson; and Y. Thaicharoen. “Institutional Causes, Macroeconomic

Symptoms: Volatility, Crises and Growth.” Journal of Monetary Economics, 50 (2003), 49–123. Belo, F.; V. Gala; and J. Li. “Government Spending, Political Cycles, and the Cross Section of Stock

Returns.” Journal of Financial Economics, 107 (2013), 305–324. Belo, F., and J. Yu. “Government Investment and the Stock Market.” Journal of Monetary Economics,

60 (2013), 325–339. Brav, A.; G. Constantinides; and C. Geczy. “Asset Pricing with Heterogeneous Consumers and

Limited Participation: Empirical Evidence.” Journal of Political Economy, 110 (2002), 793–824. Brogaard, J., and A. Detzel. “The Asset-Pricing Implications of Government Economic Policy

Uncertainty.” Management Science, 61 (2015), 3–18. Coval, J., and T. Moskowitz. “Home Bias at Home: Local Equity Preference in Domestic Portfolios.”

Journal of Finance, 54 (1999), 2045–2073. Croce, M.; H. Kung; T. Nguyen; and L. Schmid. “Fiscal Policies and Asset Prices.” Review of

Financial Studies, 25 (2012), 2635–2672. Croce, M.; T. Nguyen; and L. Schmid. “The Market Price of Fiscal Uncertainty.” Journal of Monetary

Economics, 59 (2012), 401–416. Daniel, K.; M. Grinblatt; S. Titman; and R. Wermers. “Measuring Mutual Fund Performance with

Characteristic-Based Benchmarks.” Journal of Finance, 52 (1997), 1035–1058. Durnev, A. “The Real Effects of Political Uncertainty: Elections and Investment Sensitivity to Stock

Prices.” Working Paper, University of Iowa (2010). Fama, E., and J. D. MacBeth. “Risk, Return, and Equilibrium: Empirical Tests.” Journal of Political

Economy, 81 (1973), 607–636. Garcia, D., and O. Norli. “Geographic Dispersion and Stock Returns.” Journal of Financial

Economics, 106 (2012), 547–565. Gordon, D., and E. Leeper. “Are Countercyclical Fiscal Policies Counterproductive?” Working Paper,

National Bureau of Economic Research (2005). Heaton, J., and D. Lucas. “The Effects of Incomplete Insurance and Trading Costs in a Consumption-

Based Asset Pricing Model.” Journal of Economic Dynamics and Control, 16 (1992), 601–620. Heaton, J., and D. Lucas. “Evaluating the Effects of Incomplete Insurance on Risk Sharing and Asset

Pricing.” Journal of Political Economy, 104 (1996), 443–487. Hong, H.; J. Kubik; and J. Stein. “Thy Neighbor’s Portfolio: Word-of-Mouth Effects in the Holdings

and Trades of Money Managers.” Journal of Finance, 60 (2005), 2801–2824. Hubbard, G.; J. Skinner; and S. Zeldes. “Expanding the Life-Cycle Model: Precautionary Saving and

Public Policy.” American Economic Review, 84 (1994), 174–179. Huberman, G. “Familiarity Breeds Investment.” Review of Financial Studies, 14 (2001), 659–680.

136 Journal of Financial and Quantitative Analysis

Ivković, Z., and S. Weisbenner. “Local Does as Local Is: Information Content of the Geography of Individual Investors’ Common Stock Investments.” Journal of Finance, 60 (2005), 267–306.

Ivković, Z., and S. Weisbenner. “Information Diffusion Effects in Individual Investors’ Common Stock Purchases: Covet Thy Neighbors’ Investment Choices.” Review of Financial Studies, 20 (2007), 1327–1357.

Johnson, D.; J. Parker; and N. Souleles. “Household Expenditure and the Income Tax Rebate of 2001.” American Economic Review, 96 (2006), 1589–1610.

Julio, B., and Y. Yook. “Political Uncertainty and Corporate Investment Cycles.” Journal of Finance, 67 (2012), 45–83.

Kelly, B.; L. Pastor; and P. Veronesi. “The Price of Political Uncertainty: Theory and Evidence from the Option Market.” Journal of Finance, 71 (2016), 2417–2480.

Knight, B., and A. Levinson. “Rainy Day Funds and State Government Savings.” National Tax Journal, 52 (1999), 459–472.

Korniotis, G. “Habit Formation, Incomplete Markets, and the Significance of Regional Risk for Expected Returns.” Review of Financial Studies, 21 (2008), 2139–2172.

Korniotis, G., and A. Kumar. “State-Level Business Cycles and Local Return Predictability.” Journal of Finance, 68 (2013), 1037–1096.

Kumar, A.; J. Page; and O. Spalt. “Investor Sentiment and Return Comovements: Evidence from Stock Splits and Headquarters Changes.” Review of Finance, 17 (2012), 921–953.

Lucas, D. “Asset Pricing with Undiversifiable Income Risk and Short Sales Constraints: Deepening the Equity Premium Puzzle.” Journal of Monetary Economics, 34 (1994), 325–341.

Ostergaard, C.; B. Sorensen; and O. Yosha. “Consumption and Aggregate Constraints: Evidence from U.S. States and Canadian Provinces.” Journal of Political Economy, 110 (2002), 634–645.

Parker, J.; N. Souleles; D. Johnson; and R. McClelland. “Consumer Spending and the Economic Stimulus Payments of 2008.” American Economic Review, 103 (2013), 2530–2553.

Pastor, L., and P. Veronesi. “Uncertainty about Government Policy and Stock Prices.” Journal of Finance, 64 (2012), 1219–1264.

Pastor, L., and P. Veronesi. “Political Uncertainty and Risk Premia.” Journal of Financial Economics, 110 (2013), 520–545.

Pirinsky, C., and Q. Wang. “Does Corporate Headquarters Location Matter for Stock Returns?” Journal of Finance, 61 (2006), 1991–2005.

Poterba, J. “State Responses to Fiscal Crises: The Effects of Budgetary Institutions.” Journal of Political Economy, 102 (1994), 799–821.

Romer, P. “Endogenous Technological Change.” Journal of Political Economy, 98 (1990), 71–102. Santa-Clara, P., and R. Valkanov. “The Presidential Puzzle: Political Cycles and the Stock Market.”

Journal of Finance, 58 (2003), 1841–1872. Storesletten, K.; C. Telmer; and A. Yaron. “Cyclical Dynamics in Idiosyncratic Labor Market Risk.”

Journal of Political Economy, 112 (2004), 695–717. Svec, J., and A. Kondo. “Fiscal Policy Cyclicality and Growth within the US States.” B.E. Journal of

Macroeconomics, 12 (2012), Article 2, 1–33. Tuzel, S., and M. Zhang. “Local Risk, Local Factors, and Asset Prices.” Journal of Finance, 72 (2017),

325–370. Wagner, G., and E. Elder. “The Role of Budget Stabilization Funds in Smoothing Government Expen-

ditures over the Business Cycle.” Public Finance Review, 33 (2005), 439–465. Zeldes, S. “Consumption and Liquidity Constraints: An Empirical Investigation.” Journal of Political

Economy, 97 (1989), 305–346.

Copyright of Journal of Financial & Quantitative Analysis is the property of Cambridge University Press and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.