Expected economic growth and investment in corporate tax planning

Abstract

This study investigates whether expected economic growth is associated with investment in corporate tax planning. We predict that higher expected economic growth increases the net present value of tax planning opportunities and that thus firms’ investment in tax planning will be higher in periods when expectations about economic growth are optimistic. Consistent with expectations, we find that changes in fees for auditor-provided tax services are positively associated with GDP growth forecasts after controlling for known determinants of tax avoidance. Using path analysis, we find that expected macroeconomic growth influences firms’ investment in tax planning directly, rather than indirectly through other investments. Cross-sectional analyses show this association is more pronounced for firms that are financially constrained and those that are more likely to experience a change in tax status. Our study highlights that growth expectations at the macroeconomic level are an important determinant of time-series variation in firm’s investment in corporate tax planning.

1 Introduction

Understanding the factors that influence corporate investment decisions is fundamental to both accounting and finance research (e.g., Roychowdhury et al. 2019). While research finds realized macroeconomic conditions influence firms’ investing and financing policies (e.g., Barlevy 2007; Erel et al. 2011), there is limited evidence on whether expected macroeconomic conditions influence corporate policies. Because realized economic growth directly influences the availability of firm resources for investment, it is an empirical question as to whether expected economic growth will also influence corporate policies after holding realized economic growth constant. We fill this void in the literature by investigating whether expected economic growth influences firms’ investment in corporate tax planning.

Tax planning is one of a firm’s many risky investment opportunities (Armstrong et al. 2015). We focus on tax planning as opposed to other investments for three reasons. First, the net present value (NPV) of a tax planning opportunity often depends on a firm’s future profitability, because the expected tax savings from a given tax strategy are a function of a firm’s current and future pre-tax income. Therefore the expected benefits of a tax strategy likely scale proportionally with increases in future pre-tax income. For example, a tax strategy that shifts earnings from a high-tax to a low-tax jurisdiction will generate greater tax savings as future pre-tax earnings increase. As a result, tax planning provides an especially salient setting to examine the influence of expected economic growth on corporate investment.

Second, focusing on investments in tax planning allows for stronger identification than trying to examine other corporate decisions that may also be associated with expectations of economic growth. For example, acquisitions and seasoned equity offerings may well be influenced by expectations of economic growth, but, because they are discrete and relatively infrequent, they are more difficult to model. The extensive literature on tax avoidance provides a basis for identifying important determinants of tax avoidance to help us isolate the effect of growth expectations.

Third, although research generally investigates whether firm and manager characteristics influence corporate tax planning (Wilde and Wilson 2018), there is limited evidence on the factors that influence the time-series variation in tax planning. Understanding the time-series determinants of corporate tax planning matters from both a policy and academic perspective. The recent Tax Cuts and Jobs Act (TCJA) lowered the top U.S. statutory corporate tax rate by 14 percentage points, creating the first change in the U.S. corporate tax rate since 1993. For policymakers to anticipate the consequences of a tax reform bill that was partially designed to stimulate economic growth, understanding the factors that influence time-series variation in corporate tax planning is essential. Recent research has begun to investigate the connection between tax avoidance and both firm-level and market-wide financial constraints (Edwards et al. 2016; Law and Mills 2015). Our study extends this line of research by examining how macroeconomic expectations can influence current investments in tax planning.

Expectations about economic growth likely influence corporate tax planning because a firm only invests in tax planning when the NPV of the expected savings exceeds the marginal costs (Armstrong et al. 2015). The NPV of expected tax savings is positively associated with expected pretax income and cash flows. When expectations about economic growth are optimistic, expectations about pretax income and cash flows are higher, and the NPV of the expected savings from tax planning increases. In periods when the NPV of expected tax savings is higher, we expect that managers will be more likely to invest in tax planning strategies that were viewed as too costly in periods of lower economic growth. Consequently, we hypothesize that expected economic growth is positively associated with firms’ investment in tax planning.¹

We proxy for expectations about economic growth using consensus forecasts of real GDP growth from the Survey of Professional Forecasters (SPF), the oldest and most widely followed economic survey in the United States (Konchitchki and Patatoukas 2014).² In general, a firm’s total investment in tax planning consists of the use of internal resources (e.g., a firm’s tax department) as well as fees paid to outside consultants (e.g., lawyers or public accounting firms). Because a firm’s total investment in tax planning is unobservable, we proxy for tax planning investment using fees paid for auditor-provided tax services. Klassen et al. (2016) suggest that fees paid for auditor-provided tax services partially represent a firm’s investment in tax planning. Thus, although fees related to auditor-provided tax services do not fully capture a firm’s total tax planning, it directly proxies for a firm’s investment in tax planning in a given year.³

To test our prediction, we restrict our sample to firms that purchase tax services from their auditor and examine changes in tax fees.⁴ Based on a sample of 13,553 firm-year observations from 2003 to 2014, we find that changes in tax fees paid to auditors are positively associated with contemporaneous consensus forecasts of real GDP growth.⁵ The positive association between changes in the tax fees paid to auditors and GDP growth forecasts is robust to controlling for firm-level characteristics, market-wide discount rates, current macro-level financial constraints (Edwards et al. 2016), a temporal trend in corporate tax planning (Dyreng et al. 2017), and industry fixed effects. These findings suggest that firms increase their tax planning investment in periods when expectations about economic growth are more optimistic.

To provide additional insight into our findings, we examine how expected economic growth influences firms’ investment in tax planning. Our hypothesis argues that expected economic growth directly influences firms’ investment in tax avoidance. However, expected economic growth may also influence firms’ other investments, which creates additional tax planning opportunities. Using path analysis, we investigate the extent to which expected economic growth influences firms’ investment in tax planning directly and indirectly through other investments, such as research and development and capital expenditures. Our results suggest that expected economic growth is associated with greater investments in tax planning directly but not indirectly through increased investments in research and development and capital assets. Further, our results suggest that the total effect of expected economic growth on firms’ investment in tax planning is almost entirely attributable to the direct effect of expected economic growth. This result is consistent with firms directly investing more in tax planning when expected economic growth is greater.

To alleviate the concern of a coincidental relation between firms’ investment in tax planning and expected economic growth, we examine whether the relation varies predictably in the cross-section. First, we examine whether firm-level financial constraints influence the association between expected economic growth and tax planning. To the extent that financially constrained firms do not have access to cheaper external financing, we expect that their tax planning investments will be more sensitive to expected economic growth, because the firm needs to generate tax savings to pursue other investment opportunities. However, the tax planning investments of financially constrained firms may also be less sensitive to expected economic growth, because these firms either lack the resources to pursue additional tax planning or do not have any feasible incremental planning opportunities. We find the relation between the GDP growth forecast and changes in tax planning investment is more pronounced for financially constrained firms. Our results suggest financial constraints have an interactive effect on tax planning through expected economic growth, in addition to the direct effect documented by Edwards et al. (2016) and Law and Mills (2015).

Second, we examine whether uncertainty about a firm’s future tax status influences the relation between expected economic growth and tax planning. If a firm is tax exhausted and consistently faces a 0 % marginal tax rate, then it will derive less benefit from incremental tax planning, and its tax planning should be less sensitive to growth expectations. Likewise, if a firm consistently faces a marginal tax rate that equals the maximum statutory rate, its incentives to tax plan should depend less on the state of the economy. However, if a firm’s marginal tax rate is state dependent, then the firm’s tax planning investments should be more sensitive to growth expectations. We use the volatility of marginal tax rates to proxy for uncertainty about future tax status. Consistent with expectations, we find the association between GDP growth forecasts and changes in investment in tax planning is more pronounced for firms with more volatile marginal tax rates.

Our study contributes to nascent research on the association between corporate tax planning and macroeconomic conditions. In a cross-country setting, Shevlin et al. (2019) find that aggregate corporate tax avoidance is a leading indicator of realized future economic growth, particularly in countries with more corruption. In addition, Edwards et al. (2016) find that current macro-level lending constraints are associated with future tax avoidance. We extend both studies by examining whether managers’ expectations about economic growth help determine firms’ investment in tax planning. We also contribute to research that investigates the determinants of corporate tax planning more broadly. Research generally examines the influence of firm or manager characteristics on corporate tax planning (Wilde and Wilson 2018). While research acknowledges that firms’ investment in corporate tax planning can evolve, there is limited evidence on the factors that influence the time-series variation in tax planning. We contribute to this void in the literature by examining the effect of expectations about economic growth on firms’ investment in tax planning.

Our findings should also be of interest to policymakers and tax authorities. Tax revenues generally rise during economic expansions, but if investment in tax planning also rises in anticipation of economic expansion, that increase in corporate tax planning will partially offset the effect of increases in the tax base. Thus the relation between tax planning and expected economic growth should perhaps be considered in the dynamic scoring of proposed tax legislation.

2 Hypothesis development

2.1 Macroeconomic conditions and corporate decisions

As noted above, Shevlin et al. (2019) examine aggregate tax avoidance across countries and find that aggregate corporate tax avoidance is a leading indicator of realized future economic growth particularly in countries with more corruption. While aggregate performance metrics predictfuture economic growth, it is also likely that macroeconomic conditions influence firms’ decisions. Indeed, Cochrane (2011) argues that variation in macroeconomic factors has significant implications for both corporate finance and accounting decisions. Consistent with this notion, recent research finds that realized macroeconomic conditions influence a variety of corporate decisions. Specifically, studies have developed theoretical models demonstrating that macroeconomic conditions influence firms’ capital structure decisions (Chen 2010; Hackbarth et al. 2006). Korajczyk and Levy (2003) find that firms’ target capital structure is a function of macroeconomic conditions. Further, Erel et al. (2011) find that macroeconomic conditions influence firms’ ability to raise capital as well as the types and structures of securities that firms issue. Likewise, Haddad et al. (2017) find that aggregate risk premium primarily determines corporate buyout activity. Finally, a mature stream of research finds that realized economic growth influences corporate investment in the current period (e.g., Aghion et al. 2012; Barlevy 2007).

While research has focused on realized macroeconomic conditions, there is limited evidence on whether expected macroeconomic conditions influence firms’ decisions. Although realized macroeconomic conditions are associated with future conditions, the former conditions also reflect the current availability of resources. As a result, it is unclear whether, after holding current conditions constant, expected conditions will influence corporate decisions in the same manner as current ones. Accordingly, our study aims to begin a new line of research in the broader area of work focused on the relation between economic conditions and corporate decisions by providing evidence on whether expectations about macroeconomic conditions influence current investment.

2.2 Economic framework of investing in tax planning

One area of accounting and finance where macroeconomic conditions likely play an important role is corporate tax planning. Similar to other investments, we expect that firms will invest in additional tax planning opportunities if the NPV of the expected marginal benefits exceeds the marginal costs of tax planning (Armstrong et al. 2015). The initial investment in tax planning (i.e., marginal costs) can be significant. For example, the costs potentially include expanding the tax department, hiring outside tax or legal counsel, forming a new entity, and modifying information technology and financial reporting processes. Indeed, Dyreng and Markle (2016) note that implementing income-shifting strategies requires a significant initial investment, because the firm must develop and implement the appropriate legal and operational structures. Likewise, De Simone et al. (2017) argue that modifying existing income-shifting structures is costly.

Tax planning strategies that are less complex than income shifting also involve significant costs. For example, firms often change to a more favorable method of accounting for tax purposes to generate an immediate refund (i.e., a strategy that creates a temporary book-tax difference). To change a tax accounting method, firms are generally required to request formal approval from the Internal Revenue Service (IRS). As a result, tax accounting method changes involve substantial upfront costs in the form of fees paid to outside tax consultants and lawyers, modifications to reporting systems, and tax department resources to gather the appropriate documentation to quantify the tax benefit in the current period.

Because almost all tax planning strategies involve a significant initial investment, a key factor in managers’ investment decision is whether the NPV of the expected current and future tax savings exceeds the initial implementation costs and any expected outcome costs (e.g., IRS audit, penalties and interest, or reputational and political costs). We assume that the benefits of a given tax planning opportunity increase with pre-tax income and cash flows.⁶ In other words, the benefit of lowering a firm’s effective tax rate (ETR) or deferring the payment of cash taxes increases with pretax income.⁷ When expectations about economic growth are optimistic, expectations about future pretax income and cash flows are also higher, and the NPV of a tax planning strategy increases. In periods where the NPV of expected tax savings is higher, we expect that managers are more likely to invest in tax planning strategies that were viewed as too costly in periods of lower economic growth. Accordingly, we expect that firms will invest more in tax planning in periods where expected economic growth is greater, which leads to our first hypothesis.

• H1: Corporate tax planning is positively associated with expectations about economic growth.

2.3 Cross-sectional variation in the sensitivity of tax planning investment

Our first hypothesis focuses on the primary association between expectations about economic growth and tax planning. To provide additional insight, we also investigate whether the influence of expected economic growth on firms’ investment in tax planning varies predictably with firm characteristics.

While research (Edwards et al. 2016; Law and Mills 2015) finds that financial constraints influence tax planning, we expect that financial constraints also impact the extent to which tax planning investments are sensitive to growth expectations (i.e., an interactive effect between financial constraints and growth expectations). However, it is not clear whether the tax planning investment of financially constrained firms is more or less sensitive to growth expectations, relative to unconstrained ones. Edwards et al. (2016) argue that firms use tax planning as a source of internal financing, and constrained firms may rely more on tax planning to finance investment opportunities when growth expectations are favorable. To the extent financially constrained firms depend more on savings from tax planning, their investment in tax planning is likely more sensitive to growth expectations, relative to unconstrained firms. We also expect that managers increase their assessments about the need for internally generated cash to finance investment opportunities in periods where growth expectations are more favorable. Thus, to the extent financially constrained firms do not have a cheaper source of financing available, it is likely that their investment in tax planning is more sensitive to expectations about economic growth, relative to those of unconstrained firms. On the other hand, financially constrained firms may be avoiding as much tax as possible, regardless of the economic outlook, and thus may not have any feasible incremental planning opportunities available, suggesting a weakened sensitivity. Further, financially constrained firms may not have resources to invest in marginal tax planning. Because we have conflicting directional predictions, we state our second hypothesis in the null form as follows.

• H2: The relation between expectations about economic growth and tax planning is not different between financially constrained firms and unconstrained firms.

Our third hypothesis examines whether investments in tax planning are more sensitive to expectations about economic growth when a firm’s tax status is less stable. Economic growth can impact both a firm’s tax base and its marginal tax rate, and we expect that a firm’s tax planning investments will be more sensitive to expected growth if its marginal tax rate is also more likely to change, due to fluctuations in economic conditions. Firms that rarely experience changes in their tax status should have a more consistent level of tax planning that is less sensitive to expectations about growth. For example, firms that consistently reach tax exhaustion, a 0 % marginal tax rate, due to significant past losses and various tax shields, receive little benefit from tax planning, even in periods of greater expected economic growth. Similarly, firms whose marginal tax rate regularly equals the maximum statutory rate likely invest in tax planning opportunities consistently, because their benefits of tax planning opportunities are stable, regardless of the state of the economy. In contrast, investment in incremental tax planning is likely more state dependent for firms with volatile marginal tax rates, because the benefits largely depend on the firm’s expectations about its pre-tax income and tax status.⁸ Therefore, relative to firms with stable marginal tax rates, we expect that the tax planning of firms with volatile marginal tax rates is more sensitive to expectations about economic growth. Thus, our third hypothesis is as follows.

• H3: The relation between expectations about economic growth and tax planning is more pronounced for firms more likely to experience a change in their tax status.

3 Research design

3.1 Variable measurements

Our primary constructs of interest are expected economic growth and firms’ investment in tax planning. We use macroeconomists’ forecasts of growth in the gross domestic product (GDP FORECAST) to proxy for expected economic growth. We obtain consensus mean forecasts from the Survey of Professional Forecasters made available by the Federal Reserve Bank of Philadelphia. We calculate the averages of quarterly four-quarter-ahead (annualized) GDP growth forecasts in calendar year t.⁹ Note that we do not require that managers follow GDP forecasts from the Survey of Professional Forecasters. Rather, we argue that GDP forecasts from the Survey of Professional Forecasters are correlated with managers’ economic outlooks, which can be formed from alternative sources.

Unlike for expected economic growth, a comprehensive measure of firms’ investment in tax planning is not publicly available. Research often uses cash ETRs as a measure of tax planning (e.g., Dyreng et al. 2008, 2017), but using cash ETRs in our research setting potentially reduces the power of our tests for two reasons. First, cash ETR’s denominator, pre-tax book income, is likely correlated with our variable of interest, GDP growth forecasts. Cash ETRs could thus be associated with GDP growth forecasts, even absent any change in tax planning. Second, cash ETR is an outcome-based measure that reflects the outcomes of a firm’s past tax planning, which inhibits isolation of whether expected economic growth influences firms’ investment in tax planning in the current period. Further, it is not clear how quickly cash ETR reflects a firm’s tax planning investment in a given period.

In addition, cash ETRs potentially suffer from simultaneity and reverse causality problems in our research setting. Specifically, Shevlin et al. (2019) find that, at the country level, aggregate tax avoidance is associated with future macroeconomic growth. At the firm level, Shevlin et al. (2019) find that tax avoidance is associated with future capital expenditures. Combined, Shevlin et al.’s (2019) findings suggest firms use tax savings to increase capital investment, which leads to economic growth and potentially, increased expectations about future economic growth.¹⁰ This line of reasoning implies that increased capital investment that generates depreciation deductions reduce future cash ETRs. As a result of these concerns, we use the fees paid for auditor-provided tax consulting and compliance services (TAX_FEES) to measure a firm’s investment in tax planning, because it directly measures firms’ investment in tax planning in the current period and does not mechanically relate to expected macroeconomic growth. Consistent with prior research, higher levels of TAX_FEES represent a larger investment in tax planning (e.g., Kubick et al. 2019).¹¹

Despite its advantages, using TAX_FEES to measure firms’ investment in tax planning has three drawbacks. First, TAX_FEES only captures fees paid to the firm’s auditor and not fees paid to other tax services providers (Maydew and Shackelford 2007; McGuire et al. 2012). Second, TAX_FEES represent fees paid for both compliance and consulting services. However, Klassen et al. (2016) find that fees paid for auditor-provided tax services are associated with a firm’s level of tax avoidance, after controlling for whether the external auditor also prepared the firm’s tax return (i.e., performed tax compliance services). This result suggests that fees paid for auditor-provided tax services at least partially represent tax planning services provided by the external audit firm. Finally, while TAX_FEES can reflect an investment in a tax planning strategy that was previously considered to have a negative NPV, the fees also potentially represent tax planning that is necessitated by firms’ other investment opportunities. Specifically, if expected economic growth causes firm to increase overall investment, firms potentially purchase tax services to help implement their new investments in a tax efficient manner. While TAX_FEES represents the best available proxy for firms’ investment in tax planning, we also examine the robustness of our results to an outcome-based measure of tax planning in Section 5.

3.2 Test of H1: Expected economic growth and tax planning

To test our first hypothesis, we examine changes in tax fees paid to auditors. We implement a changes specification, because this design more closely captures the proposed dynamics in our hypotheses. Further, a changes specification improves identification by mitigating concerns related to correlated omitted variables. In addition, it helps reduce concerns that we are measuring tax compliance, as opposed to tax planning, because compliance-related fees are likely to be relatively constant across years.

To test H1, we estimate the following fixed effects regressions (firm and time subscripts are omitted).

$$ \Delta T\mathrm{A}X\_ FEES=\alpha +{\beta}_1 GDP\ FORECAST+\sum \beta \Delta Controls+\gamma +\varepsilon, $$

(1)

where ΔTAX_FEES represents our proxy for firms’ investment in tax planning and is measured as the change in TAX_FEES from year t-1 to year t. As discussed previously, GDP FORECAST proxies for expected economic growth. γ is an array of industry fixed effects. H1 predicts that expected economic growth influences firms’ investment in tax planning concurrently. Because ΔTAX_FEES measures firms’ investment in tax planning directly, we measure GDP FORECAST contemporaneously (i.e., year t forecasts of year t + 1 real GDP growth). As H1 predicts that higher expected economic growth is associated with greater investment in tax planning, we expect a significantly positive coefficient on GDP FORECAST.

In addition to our variable of interest, Eq. 1 controls for known determinants of cross-sectional variation in tax planning and other time-series macroeconomic variables. Specifically, we control for whether a firm has foreign operations, because multinational firms have different tax planning opportunities compared to domestic firms (MNE) (e.g., Chen et al. 2010). Consistent with prior research (e.g., Mills et al. 1998; Chen et al. 2010; Dyreng et al. 2017), we also control for changes in the natural log of total assets (ΔLOG_ASSETS), research and development expenses (ΔRD), intangible assets (ΔINTAN), advertising expense (ΔADV), the ratio of total debt to assets (ΔLEV), the presence of a tax loss carryforward (ΔNOL), the change in the magnitude tax loss carryforward during the current fiscal year (ΔCNOL), return-on-assets (ΔROA), the number of segments (ΔNUMSEG), the percentage of foreign sales (Δ%FOREIGN), net property, plant, and equipment (ΔPP&E), capital expenditures (ΔCAPEX), current special items (ΔSP), and lagged special items (ΔLAG_SP).

In addition, we include changes in corporate audit probability rates (ΔLAG_AUDITPROB), because resources devoted to tax enforcement may vary with expectations about macroeconomic conditions, and stricter enforcement deters corporate tax avoidance (Hanlon et al. 2014; Hoopes et al. 2012). We also include both the changes in audit fees (ΔAUDFEES) and non-audit, non-tax fees (ΔNONAUDTAXFEES) to control for any overall trends in fees paid to auditors.

We also control for several macroeconomic factors to ensure that we capture the influence of expected economic growth and not the influence of current macroeconomic conditions. Specifically, we include changes in bank lending standards (ΔTIGHTENING), because Edwards et al. (2016) find that firms avoid more cash taxes when bank lending standards tighten. We also control for changes in two discount rate components, the change in the market risk premium and the change in the risk-free rate, because the discount rate can affect the NPV of tax planning investments. To proxy for the change in the market risk premium, we follow Fama and French (1989) and use changes in the difference in yields between the Aaa-rated corporate bond portfolio and the one-month Treasury bill rate (ΔTERM SPREAD). To proxy for the risk free rate, we use the change in the three-year Treasury constant maturity rate (ΔTBILL 3YR) from the Federal Reserve Bank of St. Louis and compute the averages of daily rates in calendar t. We chose a Treasury bill with a three year maturity, because research suggests that firms can alter their tax positions fairly quickly.¹² A caveat of including these two variables is that the two variables are inevitably correlated with expected economic growth (Haddad et al. 2017), potentially underestimating the impact of expected economic growth on tax planning. All macro-level variables are measured in year t in a similar spirit to GDP FORECAST. We also include actual GDP growth during year t to isolate the effect of expected economic growth from realized economic growth (Actual GDP).¹³

Finally, we include a time trend variable scaled to increase from zero to one over our sample period (TREND) to control for the decline in auditor-provided tax services following the Sarbanes-Oxley Act and PCAOB restrictions on auditor-provided services (Lennox 2016). To the extent that this variable does not perfectly capture these effects, our changes model should further rule out concerns over the effects of Sarbanes-Oxley and PCAOB restrictions. Dyreng et al. (2017) also find a declining trend in cash ETRs over time, which could suggest a temporal trend in tax planning.¹⁴ Eq. 1 includes industry-fixed effects (γ) to remove any time-invariant cross-industry variation in tax planning investments and ensure that we capture time-varying within-firm associations between expected economic growth and changes in investment in tax planning.¹⁵ Unless otherwise noted, we measure each control variable in Eq. 1 as the change in the control variable from year t-1 to year t. Appendix Table 8 provides detailed definitions for each variable.

Finally, as discussed previously, we argue that the NPV of a tax planning strategy is a function of the future tax savings, discount rate, and expected growth. Accordingly, we also estimate Eq. 1 using a levels specification because the level of investment is a function of the NPV of future tax savings and therefore expected economic growth. Unless otherwise noted in Appendix Table 8, we measure TAX_FEES and all control variables in year t and include firm-fixed effects (φ) to remove any time-invariant cross-firm variation in tax planning investments. In both our changes and levels specifications, we cluster standard errors by firm and year (Gow et al. 2010; Petersen 2009).

3.3 Tests of H2 and H3: Cross-sectional analyses

H2 and H3 focus on cross-sectional variation in the association between expected economic growth and tax planning. H2 examines whether the association between expected economic growth and tax planning varies with firm-level financial constraints. We use two common measures of financial constraints, the Whited-Wu Index (Whited and Wu 2006) and the Size-Age index (Hadlock and Pierce 2010) to classify firms’ level of financial constraint. For both, higher values signal greater constraints. We supplement Eq. 1 by adding an indicator variable, CONSTRAINED, that equals one for firm-years with a given financial constraint measure above the sample median and zero otherwise. To test H2, we interact CONSTRAINED with GDP FORECAST and focus on the interaction coefficient.

H3 examines whether the association between expected economic growth and tax planning varies depending on the certainty around firms’ tax status. We proxy for uncertainty about firms’ tax status by measuring the variability of their marginal tax rates. Specifically, we calculate the standard deviation of Blouin, Blouin et al.’s (2010) marginal tax rates over the past five years (i.e., t-5 to t-1). We supplement Eq. 1 by adding an indicator variable, HIGH VOLATILITY, that equals one for firm-year observations with marginal tax rate volatility above the sample median and zero otherwise. To test H3, we interact HIGH VOLATILITY with GDP FORECAST. Consistent with H3, we expect that the coefficient on the interaction between HIGH VOLATILITY and GDP FORECAST will be significantly positive.

4 Results

4.1 Sample selection and descriptive statistics

Our initial sample includes U.S.-incorporated firms in the Compustat Annual Industrial file from 2003 to 2014. Our sample begins in 2003, because disclosures of fees paid for auditor provided tax services were limited before that date. The sample ends in 2014, because corporate audit probability rates (AUDITPROB) are not available after 2014. We exclude any observations with missing or zero tax fees paid to auditors in years t-1 and t, since these firms likely purchase tax services from other tax service providers.¹⁶ We also exclude financial institutions (SIC codes 6000–6999) and utilities (SIC codes 4900–4999), as firms in regulated industries face different tax planning incentives. In addition, we exclude firm-years with negative pretax income in year t, because tax planning is not likely a priority for firms with a current year loss.¹⁷ Our final sample consists of 13,553 firm-year observations. The sample size for cross-sectional tests varies based on the data requirements of each partitioning variable.

Table 1 provides the descriptive statistics for our sample. We present the descriptive statistics in a levels format to make economic interpretation more meaningful and provide a sense of the magnitude of tax fees. We find that the mean (median) TAX_FEES is 0.435 (0.187), which translates into $435 ($187) per $1 million of lagged total assets. Further, the mean GDP growth forecasts (GDP FORECAST) is 2.980% during our sample period with standard deviation of 0.501%. In addition, the descriptive statistics for the control variables are similar to prior research.

Table 1 Descriptive statistics

Table 2 presents the univariate correlations among the variables (the changes specification in Panel A and the levels specification in Panel B). Consistent with expectations, ΔTAX_FEES (TAX_FEES) is positively and significantly correlated with GDP FORECAST in both panels. This univariate result provides preliminary evidence that firms invest more in tax planning during periods when expectations about economic growth are more optimistic. Consistent with the need for multivariate analysis, we find that ΔTAX_FEES is significantly correlated with a number of the control variables.

Table 2 Correlations

4.2 GDP growth forecasts and tax planning

Table 3 presents the results of regressing changes in tax fees on GDP growth forecasts and control variables. Column 1 presents the results of the changes specification, while Column 2 presents the results of the levels specification. Because the findings in all our analyses are similar across both specifications, we only discuss the results using the changes specification for brevity.

Table 3 Expected Economic Growth and Tax Planning

Consistent with H1, we find that the coefficient on GDP FORECAST is significantly positive in both columns (p value <0.01). In terms of economic magnitude, the coefficient on GDP FORECAST (0.181) in Column 1 suggests that a 1% increase in forecasted GDP growth is associated with an additional $181 of tax fees per $1 million of lagged total assets. For a firm with average total assets, our results suggest that a 1% increase in forecasted GDP growth is associated with an increase of approximately $169,000 in tax fees paid to the external auditor, which represents a 41% increase in tax fees, relative to the sample mean.¹⁸ In the levels specification, we observe a similar 45% increase in tax fees, relative to the sample mean (approximately $185,000). To put the economic magnitude of our findings into context, the average firm in our sample pays approximately $3.1 million in total fees (audit fees, tax fees, and other consulting fees) to their external auditor, so the $169,000 increase in tax fees translates into approximately a 5.5% increase in total fees.

Turning to the macroeconomic control variables, we find that the coefficient on ΔTBILL 3YR is significantly negative, while the coefficient on TERM SPREAD is not statistically significant, which shows that firms invest less in tax planning when discount rates are higher. We find that the coefficient on TREND is not statistically significant, which suggests that our changes specification removes any time trend in the amount of fees paid for auditor provided tax services. In addition, we find that the coefficient on ΔTIGHTENING is significantly positive (p < 0.05), suggesting that firms pursue more tax planning when lending standards tighten. This result is consistent with the finding of Edwards et al. (2016) that cash ETRs are lower in periods of tightened lending standards. Further, Column 2 presents the results of our analysis using a levels specification, and we find that our inferences remain the same.¹⁹ Taken together, the results in Table 3 suggest that managers invest more in tax planning when expected economic growth is higher, because the benefits of tax planning are increasing in expected economic growth.

We conduct two additional analyses to further illuminate the association between expected economic growth and firms’ investment in tax planning. First, we examine how expected economic growth influences firms’ investment in tax planning. Because expected economic growth likely influences firms’ investments more broadly, it is not clear whether our results are attributable to tax planning that relates to firms’ new investments, firms investing in incremental tax planning opportunities previously viewed as too costly, or both.

Following Landsman et al. (2012), we use path analysis to decompose the association between GDP FORECAST and ΔTAX_FEES into direct and indirect, or mediated, paths. We focus on two specific forms of investment, research and development and capital expenditures, because they are commonly used in research (e.g., Shroff 2017) and have tax planning implications. However, we acknowledge that other mediators potentially exist that could reduce the direct effect of GDP FORECAST on ΔTAX_FEES.

Our path analysis relies on the following hypothesized chain of relations: more favorable growth expectations lead to greater investments in R&D and capital expenditures, which necessitate greater investments in tax planning to reap the associated tax benefits. We expect R&D will increase tax fees via the initial R&D tax credit study as well as the ongoing compliance and documentation costs associated with the credit. We also expect that capital expenditures will increase tax fees via tax planning related to accelerating depreciation, cost segregation studies, investment tax credits, and other aspects of the tax code, such as the domestic production activities deduction.

Table 4, Panel A, presents the results of our path analysis for our changes specification.²⁰ The results suggest that expected economic growth is associated with increased investments in tax planning directly. However, we do not find that expected economic growth indirectly influences firm’s investment in tax planning through changes in R&D or capital expenditures. For example, when we consider the indirect influence of expected economic growth through firms’ increased investment in R&D, our results suggest that approximately 99.7% [0.143 / (0.143 + 0.0005)] of the total correlation between GDP FORECAST and ΔTAX_FEES is attributable to the direct path between GDP FORECAST and ΔTAX_FEES, while the approximately 0.3% of the total correlation is attributable to the indirect influence of increased investment in R&D. Panel B presents the results using a levels specification, and we find similar inferences. Overall, these results are consistent with firms directly investing more in tax planning when expected economic growth is greater.

Table 4 Exploring the Channels through which Expected Economic Growth Affects Tax Planning

Second, we explore whether the association between expected economic growth and tax planning exhibits nonlinearity, because tax planning may be more sensitive to expected growth depending on whether the economic outlook is improving or deteriorating. Kim et al. (2019) find that firms increase tax avoidance more quickly than they scale it back, which suggests perhaps that tax planning may be more sensitive to expected growth when the economy improves. To determine whether the economic outlook is improving or deteriorating, we compare the GDP growth forecast to the current realization of GDP growth with more positive values indicating improving conditions. We then create an indicator variable equal to one for observations in the top tercile of this measure and zero otherwise (OPTIMISTIC) and interact this indicator variable with GDP FORECAST. Table 5 presents the results of this analysis. We find that the coefficients on the interaction terms are positive and statistically significant (p < 0.01 in the changes specification and p < 0.10 in the levels specification), which suggests that tax planning is more sensitive to growth expectations when the GDP growth forecast significantly outpaces current period growth.

Table 5 Is the Relation between Expected Economic Growth and Tax Planning Nonlinear?

4.3 Cross-sectional variation in the association between expectations about economic growth and investments in tax planning

Our findings thus far suggest that, on average, expectations about economic growth influence firms’ investment in tax planning. To alleviate the possibility that our results reflect similar but unrelated trends in tax fees paid to external auditors and expected economic growth, we controlled for several other macro-level indicators, estimated a changes specification, and included a time trend variable in the levels specifications. To further address correlated omitted variables and provide insight into the core relation between expected economic growth and firms’ investment in tax planning, we examine whether the sensitivity of tax planning to economic growth expectations predictably varies in the cross-section.

Table 6 presents the results of the tests of our cross-sectional hypotheses. Panel A presents the results for H2, which, stated in the null, predicts that the association between expected economic growth and tax planning is not different between financially constrained and unconstrained firms.²¹ We classify firms as financially constrained, CONSTRAINED, based on the Whited-Wu index (Whited and Wu 2006) in the odd-numbered columns and the Size-Age index (Hadlock and Pierce 2010) in the even-numbered columns. Columns 1 and 2 present the results of our changes specification, and coefficients on the interaction between GDP FORECAST and CONSTRAINED are significantly positive across both columns (p values <0.05 and 0.10, respectively), indicating that financially constrained firms’ investment in tax planning is more sensitive to expected economic growth. Our estimates suggest that the tax planning investments of financially constrained firms are 25 to 30% more sensitive to expectations about economic growth. Columns 3 and 4 present the results of our levels specification, and we find similar results. Overall, the results in Panel A suggest that the relation between expected economic growth and tax planning is more pronounced for financially constrained firms.

Table 6 Heterogeneity in the Association Between Expected Economic Growth and Tax Planning

Panel B reports the results of the test of H3, which predicts that the association between expected growth and tax planning is more pronounced for firms that are more likely to experience changes in their tax status. We proxy for the likelihood of future changes in a firm’s tax status using the standard deviation of Blouin et al.’s (2010) estimated marginal tax rate data from year t-5 to t-1. Under the changes specification in Column 1, the coefficient on the interaction between GDP FORECAST and HIGH VOLATILITY is significantly positive (p value <0.05). In terms of magnitude, the coefficient estimate on GDP FORECAST and HIGH VOLATILITY suggests that, relative to firms with less volatile marginal tax rates, the tax planning investments of the high volatility firms are 23% more sensitive to expectations about economic growth. Inferences are similar under the levels specification presented in Column 2. Collectively, these results support H3 and indicate that the influence of growth expectations on tax planning is greater among firms that are more likely to experience a change in tax status (i.e., those that are neither tax exhausted nor have a constant marginal tax rate equal to the statutory rate).

As an alternative way to examine H3, we test whether the association between expected economic growth and changes in firms’ investments in tax planning vary based on whether a firm is consistently tax exhausted. We consider a firm to be tax exhausted if it consistently reports net operating loss carryforwards throughout our sample period (Scholes et al. 1990; Beatty et al. 1995). To the extent a firm is consistently tax exhausted, H3 predicts that the influence of expected economic growth on changes in its tax planning investments will be less pronounced. Table OA4 of the online appendix presents the results of this analysis. Consistent with H3, we find that the coefficient on GDP FORECAST is significantly smaller among firms with NOL carryforwards in our changes specification (p < 0.10). This result complements our primary analysis of H3 by showing that firms’ investment in tax planning is less sensitive to expected economic growth when the firm is likely tax exhausted and thus has less of a need for immediate tax planning.

In combination, the results for H2 and H3 suggest that the influence of expectations about economic growth on tax planning varies predictably in the cross-section. These results are important for two reasons. From an economic perspective, our cross-sectional tests identify a set of variables that magnifies the influence of expectations about economic growth on firms’ investment in tax planning. From an econometric perspective, the cross-sectional results mitigate the possibility that the core relation between expected economic growth and tax planning is coincidental, reflecting a similar but economically unrelated trend. For this to be the case, the coincidental relation would also have to vary predictably by our partitioning variables.

5 Additional analyses

5.1 Controlling for firm-level growth expectations

Our results suggest that expected economic growth influences firms’ investment in tax planning. However, our results are potentially attributable to firm-specific earnings expectations as opposed to expectations about economic growth.²² Therefore we examine whether our primary results remain the same after controlling for firm-level growth expectations. We measure firm-specific growth expectations using analysts’ long-term earnings growth forecasts (EPS GROWTH FORECAST) and changes in annual earnings per share forecasts (ΔEPS FORECAST) from the Institutional Brokers Estimate System.

Table 7, Panel A, presents the results of controlling for firm-specific growth expectations using our changes specification. Column 1 (Column 2) presents the results of estimating a model that includes only EPS GROWTH FORECAST (ΔEPS FORECAST) and macroeconomic controls. We find that the coefficients on EPS GROWTH FORECAST and ΔEPS FORECAST are both insignificant. Column 3 presents the results of estimating a model that includes both EPS GROWTH FORECAST and ΔEPS FORECAST along with firm-level and macroeconomic controls. We find that the coefficients on both EPS GROWTH FORECAST and ΔEPS FORECAST are not statistically significant, which suggests that the firm-level determinants of tax planning subsume the influence of firm-specific growth expectations on a firm’s investment in tax planning.

Table 7 Firm-specific Growth Expectations and Tax Planning

Column 4 presents the results of estimating a full model that includes GDP FORECAST, EPS GROWTH FORECAST, ΔEPS FORECAST, and all control variables. We find that the coefficient on GDP FORECAST remains significantly positive (p < 0.01) after controlling for both EPS GROWTH FORECAST and ΔEPS FORECAST.²³ Panel B presents the results using our levels specification, and the results are similar to those presented in Panel A. Importantly, the coefficient on GDP FORECAST remains significantly positive (p < 0.01) after controlling for firm-specific growth expectations.²⁴ Overall, the results in Table 7 suggest that the influence of expected economic growth on a firm’s investment in tax planning is incremental to the influence of firm-specific growth expectations.

5.2 Outcome-based measure of tax planning

Although TAX_FEES directly measures a firm’s tax planning investment, it does not capture the investment of a firm’s internal resources or fees paid to non-auditor external consultants. Accordingly, we use an outcome-based measure, cash ETR, to proxy for a firm’s overall tax planning investment. As discussed previously, using cash ETR as a proxy for a firm’s investment in tax planning potentially reduces the power of our tests.

We replicate our tests of H1 through H3 using changes in cash ETR as our dependent variable. To allow cash ETR to reflect firms’ incremental investment in tax planning, we measure GDP FORECAST in year t-1.²⁵ We report the results in the Table OA6 of the online appendix. In general, we find that our inferences remain the same across all our analyses using changes in cash ETR as the dependent variable. The only exception is that the coefficient on GDP FORECAST is insignificant in our test of H1. Our inability to detect results for our test of H1 may reflect the fact that there is variability across firms in terms of when investments in tax planning translate into reduced cash ETRs, indicating measurement error in the dependent variable from an econometric standpoint. In addition, we find results consistent with all of our primary analyses when we use a levels specification. Collectively, these results suggest that our inferences are robust to using an outcome-based measure of firms’ investment in tax planning.

5.3 Validation of tax fees for auditor-provided tax services

As discussed previously, tax fees paid to the auditor represent both compliance and consulting services, creating the possibility that our results are due to compliance services as opposed to tax consulting.²⁶ Our changes specification mitigates this concern because it is unlikely that fees for compliance services change dramatically from year to year. Further, based on our discussions with tax practitioners, we assume firms that purchase a small dollar value of tax services from their auditor are more likely to be purchasing tax compliance services. Accordingly, we re-estimate the levels specification of our primary analysis using quantile regressions to allow the coefficient on GDP FORECAST to differ across the tax fee distribution. We present the results of this analysis in Fig. 1. We find that the coefficient estimate on GDP FORECAST increases monotonically from 0.023 in the 0.1 quantile to 0.354 in the 0.9 quantile of the tax fee distribution (all p < 0.01). In comparison, the coefficient estimate on GDP FORECAST in the levels specification of our main analysis is 0.198. This result suggests that the relationship between GDP FORECAST and TAX_FEES is strongest for the firms that we believe are most likely to be purchasing tax planning services.

Fig. 1

Estimation Across Tax Fee Quantiles. This figure shows coefficient estimates on GDP FORECAST at various quantiles of the distribution of TAX FEES

To provide additional comfort that tax fees for auditor-provided tax services capture firms’ investment in tax planning, we regress one-period-ahead cash ETR on TAX_FEES. We present the results in Table OA7 of the online appendix and find that the coefficient on TAX_FEES is negative and significant (p < 0.10). Combined with Klassen et al.’s (2016) finding that tax fees are associated with both tax haven intensity and unrecognized tax benefits, our results suggest that tax fees reasonably proxy for firms’ investment in tax planning.

Finally, to investigate whether our results are capturing an association between expected economic growth and general fees paid to a firm’s auditor, we replicate our primary analysis using changes in audit fees and non-audit non-tax fees as dependent variables. We present the results in Table OA8 of the online appendix. We find that the coefficient on GDP FORECAST is insignificant (significantly negative, p < 0.01) when changes in audit fees (changes in non-audit, non-tax fees) serve as the dependent variable.²⁷ Collectively, these results provide counterfactual evidence that our primary findings are not due to a general positive association between expected economic growth and changes in fees paid to the external audit firm.

6 Conclusion

Although research acknowledges that firms’ investment in corporate tax planning fluctuates, there is limited evidence on the factors that influence the time-series variation in tax planning. Expected economic growth likely influences corporate tax planning, because the net present value of additional tax planning is a function of expected cash flows. We fill this void in the literature by examining whether expectations about economic growth influence firms’ investment in tax planning.

Using changes in fees paid for auditor-provided tax services as a direct and contemporaneous measure of firms’ investment in tax planning, our results are consistent with firms investing more in tax planning in periods when expectations about economic growth are more optimistic. Our results are also robust to controlling for firm characteristics, current macroeconomic indicators, and industry fixed effects. Because expected economic growth likely influences firms’ other investments, we use path analysis to examine how expected economic growth influences firms’ investment in tax planning. Our results suggest that the total effect of expected economic growth on firms’ investment in tax planning is almost entirely attributable to the direct effect of expected economic growth, which is consistent with firms directly investing in more tax planning when expectations about macroeconomic growth are more optimistic. Finally, in cross-sectional analysis, we find that the influence of growth expectations on firms’ investment in tax planning is more pronounced for firms that are financially constrained and those that are more likely to experience changes in tax status. In combination, our cross-sectional analyses reinforce the core relation between expectations about economic growth and tax planning by mitigating concerns that our main results are due to correlated omitted variables.

Our study contributes to a new line of research examining the association between expectations of economic growth and current business decisions. Our findings also contribute to recent research that investigates the association between corporate tax planning and current macroeconomic conditions (e.g., Law and Mills 2015; Edwards et al. 2016; Shevlin et al. 2019). Edwards et al. (2016) find that current market-wide financial constraints are associated with higher levels of temporary tax planning. We extend this result by providing evidence that expectations of economic growth influence firms’ current investment in tax planning. Our findings should also be of interest to policymakers and tax authorities. Tax revenues generally rise during economic expansions, but our results suggest that increases in corporate tax planning can partially offset the effect of increases in the tax base that accompany economic growth.

Appendix

Table 8 Variable Definitions