Investment Decisions in Anticipation of Recessions and Outperformance of Pre-acting Firms ι Seungmin Chee †

Soo Young Kwon ∗

Ju Hyun Pyun ‡

Korea University Business School

Korea University Business School

Korea University Business School

March 2016

ABSTRACT We empirically examine whether firms make investment decisions in anticipation of recessions, and subsequently perform better. Using a large quarterly dataset of fixed asset investments for U.S. firms during 1984-2012, we show that not all firms efficiently adjust their investment decisions in anticipation of a recession. However, we find that pre-acting firms that properly adjust their investment decisions (i.e., underinvest) before a recession outperform re-acting firms that fail to make proper investment decisions (i.e., overinvest) before a recession in subsequent returns on assets, returns on investments, and marketadjusted return measures.

Keywords: investment decisions, recession, expansion, underinvestment, overinvestment

ι

We are grateful for insightful comments from three anonymous referees, seminar participants at Korea University and Research Forum of Korean Accounting Association. † Korea University Business School, 145 Anam-ro, Sungbuk-gu, Seoul 02841 Korea, e-mail: [email protected], Tel: +82-2-3290-2836 ∗ Corresponding author. Korea University Business School, 145 Anam-ro, Sungbuk-gu, Seoul 02841 Korea, email: [email protected], Tel: +82-2-3290-1937 ‡ Korea University Business School, 145 Anam-ro, Sungbuk-gu, Seoul 02841 Korea, e-mail: [email protected], Tel: +82-2-3290-2610

1. INTRODUCTION Firms’ investment decisions depend on their expectations of the benefits of an investment, which in turn depend on the expectations of future growth and product demand in the market. However, periods of high growth are followed by periods of slow or even negative growth. These fluctuations—upturns and downturns in economic activity over time—constitute the business cycle 1 and represent one of the most important attributes that determine the benefits of an investment. Thus, firms may adjust their investment decisions to manage the effects of business cycle fluctuations. Consistent with the theoretical argument of Bolton, Chen, and Wang (2013), we expect that firms reduce their investments when faced with a future recession. However, not all firms are able to anticipate macroeconomic negative shocks because the ability to predict the economic shocks varies across managers. We denote pre-acting firms as those that, in anticipation of the business cycle, underinvest before a recession starts, to the extent that they expect a decline in demand; and overinvest before an expansion starts to utilize a foreseeable growth in demand. In contrast, re-acting firms are those that, in response to the business cycle, underinvest during and/or after a recession and overinvest during and/or after an expansion. Pre-acting firms can avoid tying up resources, stockpiling inventories, and engaging in cost of capital savings by reducing their capital expenditures in an upcoming recession. On the other hand, they can fully benefit from the growth in demand in an upcoming boom through capital investments made in advance, leading to an increase in revenues and earnings. In addition, pre-acting firms are able to secure liquidity by saving cash and subsequent maintenance costs before a recession occurs. Such liquidity helps firms sustain themselves during these recessionary periods when external financing is severely constrained. However, re-acting firms are more likely to experience tied-up resources and stockpiling of inventories, and suffer from liquidity risks during a recession. These firms are less likely to fully benefit from the growth in demand during an expansion because they lack sufficient capital investments. Thus, we expect that pre-acting firms outperform re-acting firms in various accounting performance measures and stock returns.

1

The business cycle has two phases: recession and expansion. A recession (contraction) begins when the economy reaches a peak of activity and ends when the economy hits its trough. Between troughs and peaks, the economy is in an expansionary or contractionary period.

1

Using the National Bureau of Economic Research (NBER)’s definition of a recession and U.S. firm level data, this paper empirically examines whether firms adjust resource allocation in anticipation of a recession and whether their ability to predict macroeconomic negative shocks affects their performance. Both recession and expansion likely affect corporate investment behavior, thereby influencing corporate performance. However, we focus on recessions during which the impact of investment decisions on firm performance is exacerbated because of the following reasons: we expect that the effect of macroeconomic factors is larger during the recession since managers’ policy choices are extremely restricted due to the credit constraint during this period. The inability to borrow during the recession makes it difficult for firms to weather the period and leads to firms’ bypassing of attractive investment choices (Campello, Graham, and Harvey 2010). Moreover, the NBER defines the recessionary period depending on changes in real gross domestic products (GDP) and designates other periods as the boom. Therefore, the recession denotes the distinct nature of a consecutive decline in GDP whereas the boom, as such, is designated as other periods without such distinct characteristics. Under the null hypothesis that a firm’s resource allocation is not associated with recession, firms should exhibit investment levels consistent with their fundamentals. Our study follows the extensive body of literature that determines optimal investment decisions using a linear model that relates capital expenditures to investment opportunities (Alti 2003; McNichols and Stubben 2008). Specifically, we identify excess investments as those that differ from the amount predicted given a firm’s past level of investment, its investment opportunities (Tobin’s Q), and internal financial capabilities (cash from operations). Then, we test whether firms’ excess investments lead or lag recessions. Next, we examine whether firms’ investment decisions in anticipation of a recession affect its future performance by comparing the performance dynamics of pre-acting and re-acting firms. We find empirical evidence that not all firms efficiently adjust their investment decisions in anticipation of a recession. This observation is consistent with the finding that, after a surprise increase in risk and uncertainty, firms tend to halt or reduce investments that cannot be easily reversed―they “wait and see” (e.g., Bernanke 1983). More importantly, we find that pre-acting firms outperform re-acting ones in both accounting and market performance measures. This result indicates that pre-acting firms are more capable of reducing costs associated with recessions than re-acting firms are, implying that a firm’s

2

ability to predict recessions and adjust its investment decisions accordingly is a critical success factor in business. To date, relatively little research has addressed the relationship between recessions and firm-level investment decisions. Our study makes the following contributions. First, previous studies in finance, such as Bolton, Chen, and Wang (2013) and McLean and Zhao (2014), examined the effects of recession on a firm’s cost of external financing, investment sentiment, and consequently real investments, both theoretically and empirically. However, our work is distinct from theirs because our empirical analysis identifies firms’ ability to predict negative macroeconomic shocks in anticipation of a recession. In particular, our method of identifying pre-acting and re-acting firms is closely related to firm sensitivity to the future occurrence of recessions, as emphasized by Bolton et al. (2013). Indeed, their theory indicates that both heterogeneous financing constraints and the different probabilities that a firm attaches to future financing shocks generate different responses by a firm to investments. For instance, a small increase in the probability of a negative financing shock causes firms to underinvest significantly and hold on to excess cash, even in good times or before a crisis, rather than during a recession. Second, our study examines the extent to which firms’ heterogeneous investment decisions in anticipation of (or in response to) recessions affect their financial performance. To the best of our knowledge, our work is the first to empirically test whether firms that are aware of future macroeconomic risks have superior subsequent performance. Lastly, our study adds to the literature that investigates sources of investment inefficiencies. While most finance studies focus on the agency problem that causes investment inefficiencies (e.g., Bernardo, Cai, and Luo 2001; Grenadier and Wang 2005), we examine the effect of macroeconomic factors on a manager’s investment decisions. The results of this study suggest that the ability to predict macroeconomic factors, such as recessions, is important for reducing investment inefficiencies in addition to the firm-specific factors documented in prior research, such as information advantage. The remainder of this paper proceeds as follows. Section 2 outlines the theoretical considerations and the central research question. Section 3 describes the research design, and Section 4 reports the empirical results. Section 5 provides additional analyses to enhance the robustness of the results, and Section 6 concludes.

2. THEORETICAL BACKGROUND AND HYPOTHESES DEVELOPMENT 3

Is the business cycle (recession) predictable? A large body of literature exists on the sources of business cycle fluctuations. A line of research that approaches this question from the perspective of general equilibrium models tends to provide technology shocks a dominant role in business cycle fluctuations (see King and Rebelo 1999 for more details). In contrast, the empirical structural VAR literature usually introduces other macroeconomic disturbances as the main sources of business cycles (Cochrane 1994; Shapiro and Watson 1988). These macroeconomic studies consider the determinants of business cycle fluctuations and focus on the role of either technology or other types of macroeconomic shocks in explaining these fluctuations. In particular, the recent crisis draws attention to whether a recession is predictable. Although the business cycle follows a pattern, changes in activities are irregular and unpredictable. It is difficult to assess the duration of any part of the cycle or the extreme nature of booms or recessions. Despite this difficulty, firms attempt to predict fluctuations in the level of activity to avoid the costs associated with making investments in a recession, when demand in the economy is low and to fully benefit from the investments made in a boom through high demand. We expect that a differential ability exists in predicting the stages of the business cycle, particularly recessions, across firms.

Recession and firm’s investment decision Previous works such as Bloom (2009) and Bachmann and Bayer (2014) gauge business cycle fluctuations by considering firm-level investments. These economic studies primarily explain business cycle fluctuations by considering microeconomic firm-level data from a macroeconomic perspective. However, few prior studies examine the effect of macroeconomic shocks on firm-level heterogeneous decisions and their consequences on firm performance. While a firm’s investment decisions are affected by the idiosyncratic part of the risk, especially Jensen’s alpha (Pattitoni and Savioli 2011; Pattitoni, Petracci, Poti, and Spisni 2013), we focus on the undiversifiable systemic risk, such as macroeconomic factors, in influencing corporate investment decisions. 2 To the best of our knowledge, this study is the first to test whether firms adjust investment decisions in anticipation of recessions by Certainly, two types of risks―an individual firm’s idiosyncratic and macroeconomic risks―have exerted confounding influence on its investment decision. However, our main focus in this study is the relationship between firm performance and a firm’s ex-post investment decision before the recession, rather than an attribute that drives firmsʼ investment decisions as such. Thus, we identify two types of firms that show different investment behaviors immediately before the recession and investigate whether they make investment decisions in anticipation of the recession and their relationship to firm performance. 2

4

reducing investments to avoid extra capacity or stockpiled inventories during a recession. Furthermore, this study shows the importance of considering macroeconomic shocks to determine a firm’s investment decisions to improve financial performance. Our study is also related to the finance literature that examines investment inefficiency. Such literature primarily focuses on situations in which a manager’s short-term objective, combined with his private information, leads to investment inefficiency (Bernardo, Cai, and Luo 2001; Grenadier and Wang 2005). However, these studies do not consider the effect of macroeconomic factors on a manager’s investment decisions. This omission is important because a firm’s investment opportunities and returns are considerably affected by macroeconomic factors. Although each manager has private information about firm-specific factors that affect a firm’s investment opportunities, he does not necessarily have superior information about the macroeconomic factors. Prior research indicates that managerial styles explain a significant portion of the heterogeneity in investment decisions. For example, the literature examines how CEO overconfidence (Malmendier and Tate 2005), manager over-optimism (Heaton 2002), and early life experiences (Benmelech and Frydman 2015; Dittmar and Duchin 2016; Malmendier, Tate, and Yan 2011; Schoar and Zuo 2013) affect corporate investment decisions. In line with this research, we assume that the ability to predict the business cycle, especially recession, varies among different managers and that each manager adjusts the level of investment depending on his or her outlook of the cycle. By focusing on the recessionary period during which a macroeconomic factor’s effect on investment returns is exacerbated (Campello, Graham, and Harvey 2010), we assume that the level of underinvestment captures a manager’s ability to predict the economic cycle rather than investment distortions resulting from the agency problem. During a recession, firms need to consider various options for managing lower revenues and, therefore profits. In most businesses, there are inefficiencies and instances in which the costs of unprofitable customers or products are covered by profitable ones. Firms can focus on core profitable activities and identify areas in which costs can be reduced. They may take actions, such as eliminating redundancies, closing offices, shrinking capacity, or even selling off assets such as machinery and buildings. We investigate whether managers play a role in strategically making investment decisions in anticipation of a recession. Strategic decisions have long-term and companywide effects. The question is whether managers must respond to unexpected events, such as 5

recessions in the environment in which they operate, to remain competitive even during difficult times.

Hypotheses development We test whether firms strategically make investment decisions in anticipation of a recession. Investment decisions depend on the expectations of the benefits of an investment, which in turn depend on the expectations of future growth and product demand. Expectations of future growth vary with the business cycle. Growth declines and demand is weak in a recession, indicating that capital expenditures are costly for firms because they may have to manage inventory stockpiles and pay the cost of capital spent on investments. If firms are successful in foreseeing economic conditions before they commit to significant investments, we expect their investment decisions to lead to the business cycle. If firms are unsuccessful in predicting a recession, then we may observe that they concurrently adjust their investments after the recession. Thus, our first hypothesis is specified as follows.

H1: Firms are more likely to underinvest in anticipation of a recession.

However, one may argue that not all firms are capable of anticipating the business cycle. Pre-acting firms that reduce capital investments in anticipation of a recession can invest more efficiently than re-acting firms that reduce investments in response to a recession. In a recession, revenues decline because growth and demand decline, and maintenance costs decline because of excess capacity. Pre-acting firms reduce their capital expenditures to minimize inventory holding costs, thus avoiding the cost of capital needed for these expenditures. Further, these firms are able to secure internal capital by foregoing inefficient investment options and weathering recessionary periods during which liquidity is scarce. In contrast, re-acting firms already make commitments to investments and may have to bear the costs associated with inefficient resource allocation, the holding costs of inventory, and/or the cost of capital for investments, all of which dampen the earnings. 3 In addition, investors in the stock market might be able to identify the pre-acting and re-acting firms, thus affecting 3

If the under/overinvestment in the pre-recession period is related to the firms’ ability to predict the business cycle, it is natural to suppose that the pre-acting firms would overinvest before a boom initiates because they expect to secure much more market demand within a short period. However, owing to the nature of our dataset on the NBER recession date, the remaining periods except the recession are coded as expansion (boom) periods, which are not well identified in the dataset compared to recession periods. Thus, our analysis focuses on the recession periods to investigate a firm’s ability to predict the business cycle.

6

their stock prices. Therefore, we expect a significant difference in the performance in net income and market return between pre-acting and re-acting firms. The second hypothesis is specified as follows: H2: Pre-acting firms, which adjust investments in anticipation of a recession are more likely to perform better than re-acting firms, which make adjustments in response to recession.

3. RESEARCH DESIGN Our study requires measuring two key constructs: the recession and excess investments. We use several approaches to measure each construct to provide confidence that our findings are not driven by measurement errors in any one particular variable.

Measure of the recession Most prior research on the business cycle has utilized its economic definition. The reasoning behind using this definition is that the NBER provides an objective measure of the readily available economic business cycle, and using extensive analysis identifies ex-post peaks and troughs in economic activity, thus marking the onset and end of economic contractions and expansions. Consistent with prior research, we utilize the well-accepted measure based on the NBER business cycle recession dates. Essentially, the NBER data represent a timeline that defines economic cycles based on whether the economy is expanding or contracting at a given point, using several macroeconomic factors. Figure 1 illustrates the economic cycle from the first quarter of 1981 to the fourth quarter of 2012 using the NBER criteria for both economic performance and NBER announcement dates. Instead, to produce a monthly chronology of peaks and troughs, the NBER concentrates on monthly indicators of economic activity, such as total payroll employment, real income, industrial production, and real sales. To determine the date of a business cycle peak, the NBER examines the individual peaks in these and similar series. If an economic slowdown affects many different sectors, then the peaks in these individual series tend to cluster together. The monthly date of the central tendency of this cluster is designated as the overall business cycle peak. The currently available estimates of quarterly aggregate real domestic production do not clearly indicate dates of peaks in activity. The Business Cycle Dating 7

Committee of the NBER is responsible for maintaining a monthly chronology of the business cycle. 4 For this purpose, the committee mainly relies on monthly indicators. In contrast, monthly data on firms’ investment decisions are not available to the public. Furthermore, in practice, a popular rule of thumb is that two consecutive quarterly declines in real GDP signal a recession. 5 Indeed, in dating business cycles, the NBER also focuses on movements in quarterly real GDP, considers quarterly indicators, and maintains a quarterly chronology. Thus, this study uses quarterly investment variables and quarterly phases of a business cycle. In addition, we use annual investment variables and annual phases of a business cycle because firms may engage in investing in planning on an annual basis and quarterly investment allocations may be ad hoc. The NBER’s quarter/annual classifications are presented in Table 1. As Table 1 indicates, the recession periods are only from the third quarter of 1981 to the fourth quarter of 1982 (six quarters), the third quarter of 1990 to the first quarter of 1991 (three quarters), the first quarter of 2001 to the fourth quarter of 2001 (four quarters), and the fourth quarter of 2007 to the second quarter of 2009, also known as financial crises (seven quarters). Thus, for the 32year sample periods, only 20 out of 128 quarters (15.63 percent) are classified as recessions and the rest are viewed as expansions. Thus, to test our foregoing hypotheses, we categorize our sample period by quarter and year by coding “Recession” as 1 if a sample period is in a recession and 0 otherwise. Using this setting, we infer whether firms make investment decisions in anticipation of recessions.

Measure of excess investment We identify excess investments as those that differ from the amount predicted by a firm’s investment opportunities, using a model motivated by the finance and economics literature on optimal investments. A large body of literature in finance and economics analyzes investment decisions and attempts to understand the factors that influence investment behavior and how changes in monetary policy or other policies affect investments (Hubbard 1998). The empirical analyses in this literature generally employ the following investment model:

4

For further discussion on the definition and modeling of recessions, see Diebold and Rudebusch (1999). This rule is consistent with the dispersion and duration requirements for a recession and with the average recessionary path of real GDP. However, two very small quarterly declines might not produce the depth required for a recession.

5

8

INVit =+ α 0 α1Qi ,t −1 + α 2CFit + α 3GROWTH it −1 + ε it

(1)

where INVit is the investment level for firm i in time t, Qi,t−1 is the beginning of period t market value of assets divided by the book value of assets, and CFit is a measure of firm-level cash flows. The linear relation between investment and Qi,t−1 is motivated by investment models that incorporate adjustment costs and linear homogeneity in the production function. Modigliani and Miller (1958) show that in perfect capital markets, investments depend only on investment opportunities, and Tobin (1969) shows that investment opportunities are summarized in marginal q. Hayashi (1982) provides conditions under which marginal q is equivalent to average Q, which leads to the commonly used formulation as previously described. CFit is included to control for differences in internal financing capability. We also include asset growth at t-1 (GROWTHi,t−1), which equals the natural log of total assets at the end of t−1 divided by total assets at the end of t−2 because growth firms are more likely to invest independent of any misstatements, and to address potential measurement errors in Tobin’s Q. We also introduce a modified version of Equation (2), from which we will derive our baseline measure for excess investment. This version controls for past investment, which helps explain investment dynamics and also captures a firm-specific component to investment decisions not captured by the other variables in the model (McNichols and Stubben 2008).

INVit =+ α 0 α1 INVit −1 + α 2Qit −1 + α 3CFit + α 4GROWTH it −1 + eit

(2)

where INVit-1 is the lagged value of investment level for firm i. This augmented model shows our baseline specification to derive excess investment. Following McNichols and Stubben (2008), we estimate this investment model in Equation (2) separately for each industry and time (every industry-quarter pair), and calculate firm level excess investment using a residual from the predicted investment. Residual investment is measured incremental to the portion of an investment that is explained by investment opportunities and internal financing constraints. This approach implicitly assumes that the responsiveness of investment-to-investment opportunities is constant across firms in the same industry and year. 6

Data 6

Given that adjustment costs are not linear, and thus, the relationship between investments and Tobin’s Q is a function of Tobin’s Q (Abel and Eberly 2002), McNichols and Stubben (2008) also include quartile information on Tobin’s Q in the investment model.

9

The financial statement and market value data used in this study are obtained from the Compustat quarterly file. Our data cover the sample period from the fourth quarter of 1984 to the fourth quarter of 2012. 7 To calculate the ratios, we require net property, plant, and equipment (NPPE, item #8) to be greater than zero. Capital expenditures (INV) are taken from the statement of cash flows when available (item #128, otherwise item #30). Our proxy for Tobin’s Q is (MVE + TA − BVE)/TA, where MVE is the market value of equity (item #25 × item #199) and BVE is the book value of common equity (item #60), both measured at the beginning of the year. 8 Cash flows (CF) are obtained from the statement of cash flows when available (item #308); otherwise, the balance sheet approach is used. In this case, CF = OIAD – (∆CA − ∆Cash) – (∆CL − ∆STD − ∆TP) − DEPR, where OIAD is the operating income after depreciation (item #178), CA represents the current assets (item #4), Cash is cash and cash equivalents (item #1), CL is current liabilities (item #5), STD is the debt included in current liabilities (item #34), TP is the income taxes payable (item #71), and ∆ is the firstdifference operator. We obtain external financing measures using data from the statement of cash flows. We sum the cash proceeds from debt issuances (item #111) and from common and preferred stock sales (item #108). We also determine the change in accounts receivable from the statement of cash flows (item #302). We measure revenues from the income statement (item #12), and each variable used to calculate the discretionary revenues is deflated by the average total assets (item #6). Non-missing values for each variable previously described are required for each firm-year observation to be included in the sample. Table 2 presents the sample selection procedure. The initial firm-year sample is 13,658 firms (or 384,762 firm-year observations). From the sample, we exclude firms in the financial and insurance industry (1,904 firms or 41,935 firm-year observations) since investment decisions are made considering different factors due to the industryʼs highly regulated environment (Elloumi and Gueyie 2001). We also rule out firms in the public administration industry because they either typically oversee government activities or are engaged in the 7 The sample contains observations from years 1984-2012 with accounting data from the Compustat Fundamental quarterly database and the stock return data from CRSP. Our original US firm data is available from the year 1981. However, our sample excludes the first three years of 1981-1983, which are identified as the recession period in the NBER recession date because we did not collect information on firm’s investment in prerecession for that recession period. 8 One may argue that the book value of assets needs to be adjusted to reflect replacement costs for constructing a more precise Tobin’s Q. However, Perfect and Wiles (1994) suggest that this adjustment is not critical.

10

organization and financing of the production of public goods and services. The “publicprivate difference” has been studied from various aspects such as management, decisionmaking, and the criteria on investment appraisal (Boyne 2002; Brealey et al. 1997; Nutt 2006). Finally, we exclude firms in an industry with less than seven firms in a particular year because we utilize information on industry-level investments to measure excess investments. In this context, including data on a small number of firms in an industry in a specific year fails to provide accurate estimation results for excess investments. However, note that our results are not sensitive to this exclusion from the sample. The final sample includes 339,778 firm-year/quarter observations for 11,614 firms. Table 3 presents the distribution of sample firms across industries. In the sample, a large number of observations are in the Business Services industry, representing 12.2 percent of the total sample, followed by the Chemical & Allied Products industry, constituting 8.5 percent. In general, the observations are across every industry and not clustered in any particular one.

Empirical Procedures We investigate the behavior of U.S. firms’ excess investments through time relative to each stage of the business cycle, particularly recessions. We introduce the prediction errors from the estimated results for industry-time pair using Equation (2) as our main measure of excess investments for firm i in t. Then, we examine the mean level of excess investments over the six quarters (two years) preceding the contraction period, the quarters (years) during the contraction period, and the six quarters (two years) following the contraction period, as shown in Equation (3). Our estimation assumes that the proxies for the investment opportunity set capture the firm’s optimal investment at each point. Therefore, excess investments that are significantly different from zero reflect deviations from the optimal investment for the sample.

XINVit = β 0 + β1 Pre RC + β 2 RC + β3 PostRC + ∑ κ i + ξit

(3)

where RC is equal to 1 if the business cycle phase is a recession, and 0 otherwise. Similarly, PreRC (PostRC) is coded 1 if the business cycle phase is one to six quarters before (after) the recession, and 0 otherwise. If firms strategically reduce investments in anticipation of a recession (RC), we expect β1 to be negative because investment-reducing decisions lead to recessions. In contrast, if firms react to a recession by decreasing investments, we expect β3 to be negative. Firms that experience a decline in sales during a recession are more likely to 11

underinvest and those that experience a growth in sales during a boom are more likely to overinvest. Similarly, if firms overinvest, or rather, increase investments even in anticipation of a recession, we expect β1 to be positive because investment-increasing decisions lead to recessions. In contrast, if firms react to a recession by increasing investments, we expect β3 to be positive. Thus, Equation (3) provides an answer for Hypothesis 1 (H1). Furthermore, to test whether pre-acting firms that reduce capital investments in anticipation of a recession perform better than re-acting firms do in subsequent periods, we specify Equation (4).

RoAit += θ 0 + θ1, j ΧΙitj =( UPre,O− RC ) + Wit + s Θ + ∑ κ i + ∑ yeart + ε it for s=0, …, 4 s

(4)

The dependent variable RoAi,t+s is the return on assets (RoA) for firm i at s quarters from the reference period t, measured using net income divided by total assets—the accounting performance measure most commonly used. The test variable XIUit (XIOit) is the extent of firm i’s underinvestment (overinvestment) at t, before the recession, measured by the absolute value of excess investment (XINVit) at t. ΧΙUit = XINVit when XINVit < 0 and t is in the prerecession period, which indicates the amount that firm i underinvests for t that is one quarter ahead of a recession. ΧΙ Oit = XINVit when XINVit > 0 and t is in the pre-recession period, which indicates the amount that firm i overinvests for a t that is one quarter ahead of a recession. Wit + s is a vector of firm-level variables that affects RoAit + s , such as firm size, R&D, and leverage, among others. We also include firm fixed effects and year fixed effects to control for unobserved attributes in the firm and year dimensions. One may argue that return on investment (RoI) is a better measure to evaluate efficiency of investment than RoA is. However, in our specification, XINV or XI are constructed by investment. Thus, RoI itself has serious simultaneity and endogeneity with the XI variable. Therefore, we focus on a more general performance measure of firms such as RoA. For the robustness of the results, we replace RoAi,t+s with Reti,t+s, the market-adjusted return and RoIi,t+s, return on investment for the robustness of the results. We expect the coefficient of XIUit (θ1,U) to be positive for pre-acting firms because in a recession, these firms can minimize their cost to hold inventory and avoid capital expenditure costs. In contrast, we expect the coefficient of XIOit (θ1,O) to be negative for re-acting firms because in a recession, these firms must bear the costs associated with inefficient resource allocation, which dampens earnings. θ1,U is positive for XIUit and θ1,O is negative for XIOit to the extent that the market can distinguish between pre-acting and re-acting firms. 12

Note that in our two-step estimation procedure, XINV or XI in Equations (3) and (4) are measured as a residual from the first stage regression for the investment model. Thus, the hypothesis tests based on the estimated covariance matrix of the second-step estimator can be biased, even in large samples (Murphy and Topel 1985). Hardin (2002) shows that (heteroskedasticity corrected) the sandwich estimate of variance can be a good alternative to the Murphy-Topel estimate to improve statistical inference in this two-step estimation. Here, we report the sandwich standard errors clustered at the firm/industry level to strengthen our results.

4. TEST RESULTS Descriptive statistics and derivations of excess investments Table 4 presents the summary statistics for the entire sample. The descriptive statistics are presented in Panel A and correlations in Panel B. We observe large differences between the mean and median statistics and large standard deviations for most variables. These differences indicate the influence of outliers on the mean and standard deviation of many variables. For this reason, we focus our discussion on medians and quartiles. At the median, firms invest 172 percent of net property, plant, and equipment (NPPE). The interquartile ranges indicate significant cross-sectional variations in these amounts. Specifically, at the quartiles, investments range from 28 percent to 927 percent of NPPE. The mean excess investment (XINV) is close to zero by construction. However, the median is negative (–2.329), indicating a potentially right-skewed distribution (the mean is greater than the median) in the data. The median Tobin’s Qi,t–1 is 1.444, consistent with the unrecognized assets causing the market value of assets to exceed their book value. The median cash flow (CF) from operations is a positive 181 percent of NPPE; however, cash flows in the first quartile are negative (–62 percent of NPPE). The median asset growth (GROWTHt) is 0.01, with an interquartile range between –0.025 and 0.047. Correlations among the variables are tabulated in Panel B of Table 4. Investments are positively correlated with profitability as measured by cash flows, with asset growth, and with size. However, in our entire sample, investments show a slightly negative correlation with Q. To control for additional factors that influence investments, we base our inferences on the firm-adjusted and multivariate relations in Equations (1) and (2). Note that the estimated results of the investment model for each industry-quarter pair (total 59 industries for 1984 Q4~2012 Q4) can be obtained from the

13

authors on request. In the Appendix Table, we report the results of the investment model for the full sample to check the validity of the relationship between investment and other controls.

Results for Hypothesis 1 Table 5 shows the results of regressing excess investment on the pre-recession, recession, and post-recession dummy variables using quarterly data. In column (1), the dependent variable is XINV_B, the excess of actual investment over the expected investment estimated from the first investment model in Equation (1). The coefficients of Pre-recession and Recession are positive, but are not statistically significant at conventional levels. However, the coefficient of Post-recession is negative but statistically insignificant. These results suggest that firms neither anticipate a recession to adjust their investments nor react to a recession by reducing their investments. In column (2), the regression based on the excess investment XINV, measured by the residual from Equation (2), exhibits similar results. The results do not appear sensitive to how the excess investment is measured. The results based on the annual data in columns (3) and (4) are also consistent with the results in columns (1) and (2). In Table 5, we measure pre-recession and post-recession periods as six quarters before and after a recession, respectively. Alternatively, we use four-quarter and eight-quarter horizons, respectively, to measure pre-recession and post-recession periods, and then re-run the regressions in Table 5. The results exhibit similar patterns. Thus, we reject Hypothesis 1. However, this full sample including both pre-acting and reacting firms may attenuate either side of the investment decision in response to the recession.

Results for Hypothesis 2 The results for Hypothesis 1 show that not all firms anticipated a recession and were unsuccessful in adjusting their investments. However, a subset of firms may exist that better predicts a recession than other firms do. Thus, we first identify pre-acting firms that underinvest before a recession. We then test Hypothesis 2, which states that pre-acting firms that avoid the cost of tied-up resources perform better in subsequent periods than re-acting firms do.

14

The first measure that we employ is the accounting performance indicator RoAt+s, return on assets. Panel A of Table 6 presents the post-recession performance for pre-acting firms (XIU) and re-acting firms (XIO), respectively. As shown in columns (1) to (5), the coefficients of XIU are all positive and significant for RoAit–RoAit+4 except for RoAi,t+2, indicating that preacting firms perform well in subsequent periods. In contrast, the coefficient of XIO is negative and significant at t, implying that re-acting firms perform poorly in overinvesting immediately before a recession. The extent of excess investments and a firm’s financial performance may be influenced by the extent of cash flows, R&D expenditures, and/or leverage. Firms with large cash flows may afford to make large investments, whereas those with R&D expenditures and large longterm debt are less likely to make additional investments if firms already face financial constraints. Furthermore, these variables themselves affect the firm’s RoA. Panel B of Table 6 includes these variables in the regression model to control their effects on investments. The coefficients of XIU are positive and significant for RoAit, RoAit+1 and RoAit+4, while those of XIO are all negative but statistically significant at t. These results confirm that pre-acting firms perform better and re-acting firms perform relatively poorly in subsequent quarters. Panel C exhibits the results using the binary measure of excess investment. These results remain qualitatively unchanged. We also test for Hypothesis 2 by examining an alternative firm’s performance measure such as RoI and market performance (Ret), measured using market-adjusted returns. Table 7 presents the results of RoI. Consistent with the results in Table 6, both the coefficients of XIU and XIO are positive and negative, respectively. Note that to avoid any simultaneity between investment and R&D expenditure, we exclude the R&D intensity variable from the RoI equation. Interestingly, the estimated coefficients of XIO are significantly negative for RoIit―RoIit+3 and show that RoI captures more persistent negative effects from the reacting firm’s overinvestment before recession compared to RoA. Table 8 shows the results of the cross-sectional regressions of market-adjusted returns with respect to XIU (or XIO) and other control variables. Interestingly, note that both the coefficients of XIU and XIO are positive and negative, respectively, at t+1 and become insignificant thereafter. This short-lived significance contrasts with the accounting performance measure (RoA) that lasts for three or four quarters from the inception of a 15

recession. The market return measure appears to reflect the implications associated with the adjustment of investments to a recession at a faster rate than the accounting measure. This phenomenon may occur because accounting earnings reflect economic events with a lag given the revenue recognition criteria. Overall, the results confirm the second hypothesis that pre-acting firms that adjust investments in anticipation of a recession are more likely to perform better than re-acting firms that make adjustments in response to a recession. To confirm our findings, we divided the sample into pre-acting firms (XIU) and re-acting firms (XIO) in the recession period and traced their performance measures. Panel A of Figure 2 shows the accounting performance of publicly traded pre-acting firms in the non-financial industry in the periods subsequent to a recession. Pre-acting firms’ cumulative returns on assets are higher than that of re-acting firms in all periods. Panel B of Figure 2 exhibits similar results for only manufacturing firms. The manufacturing sector is one of the important sectors that can have a wide impact on employment, personal income, industrial production, business sales, and eventually GDP. Thus, the impact of recession on manufacturing is relatively enormous compared to that in other business sectors, eventually affecting the entire economy. Therefore, we illustrate the manufacturing sector although contracting in one sector does not imply that the rest of the economy will follow suit. Panel A of Figure 3 presents the cumulative market-adjusted returns of both pre-acting and re-acting firms. 9 Market-adjusted returns of publicly traded pre-acting firms in the nonfinancial industry are initially smaller than those of re-acting firms are before a recession and become slightly larger in the recessionary period. Moreover, in subsequent quarters, cumulative market-adjusted returns of pre-acting firms become substantially higher than those of re-acting firms. Although Panel B of Figure 3 presents the results only for manufacturing firms, similar patterns emerge for other firms as well. Taken together, both market and accounting performance measures indicate that pre-acting firms outperform reacting firms in periods subsequent to a recession.

9

The reason why we consider cumulative returns is to document the differences in long-term performance between pre-acting and re-acting firms after the recession. We also observe non-cumulative performance measures and significant differences in accounting and market measures.

16

5. ROBUSTNESS OF THE RESULTS We perform several additional analyses and sensitivity tests to check the robustness of our results.

Alternative measures of excess investments We examine the sensitivity of our findings by three alternative estimates of excess investments. The alternative measure of excess investments is the residual (XINV_B) estimated from the baseline investment equation in Equation (1). XI_BU (XI_BO) is the absolute value of XINV_B if XINV_B is negative (positive). XI_BU and XI_BO reflect preacting and re-acting firms in response to recessions, respectively. The results are shown in Table 9. The coefficients of XI_BU are all positive and significant in subsequent periods, consistent with the results in Table 7. The coefficients of XI_BO are all negative and significant at t, t+2 and t+3, also confirming the results in Table 7. Taken together, the finding that pre-acting firms outperform re-acting firms in subsequent periods is robust against alternative measures of excess investments. Use of annual data to measure the recession We use a quarterly classification of the recession to utilize firm-level information with as high frequency data as available. In this robustness test, we re-do the analyses by measuring the recession on a yearly basis. Table 10 presents the results of the primary analyses. The coefficients of pre-acting firms are positive and significant at t and t+1, thus exhibiting patterns similar to that in Table 6. The coefficients of re-acting firms are all negative but significant at t and t+3, consistent with the interpretation that the performance of these firms is poor in the subsequent periods. The results appear insensitive to the business cycle measurement interval. Alternative explanation for the results We document that pre-acting firms reduce capital investments earlier than re-acting firms do during the recession. We interpret this as the evidence that pre-acting firms invest more efficiently in anticipation of a recession than re-acting firms that reduce investments in response to a recession. However, it is possible that the former are more cyclical and the latter countercyclical, suggesting that each group of firms may be acting in an optimal way. However, if that is the case, we do not expect any difference in performance between pre17

acting and re-acting firms. Our findings show that pre-acting firms perform better and reacting firms perform relatively poorly in subsequent quarters, regardless of the performance measures employed. This result reduces the likelihood of supporting this alternative explanation. The result that firms do not underinvest before the inception of a recession may also be consistent with the alternative explanation that the firms’ investment can be countercyclical. This could be possible since the business cycle affects not only the market demand but also the investment costs. In particular, the investment costs might be cyclical if the impact of factor costs such as wages and rent dominates that of financing costs (Elliott et al. 2012; Jeon and Nishihara 2014; Solon et al. 1994). Thus, we cannot rule out the explanation regarding the fact that firms do not underinvest before a recession begins does not necessarily mean that they lack the ability to predict a recession.

6. CONCLUSIONS This study examines whether firms adjust resource allocation in anticipation of a recession. It also investigates whether firms (pre-acting firms) that properly underinvest before a recession to avoid the costs associated with tied-up resources, outperform firms (reacting firms) that overinvest before a recession. We examine fixed asset investments for a large sample of U.S. firms listed on the NYSE and ASE during the 1984-2012 period. First, we document that firms do not exhibit systematic changes in their investment in anticipation of the recession. However, we find that pre-acting firms outperform re-acting firms in subsequent periods by both accounting and market performance measures. Our study significantly contributes to the existing literature. First, our study extends prior studies in macro-finance by linking macroeconomic variables to firm-level microeconomic decisions. Second, our study documents whether firms can predict a recession at the firm level as opposed to the national aggregate level. Third, this study documents the unintentional investment decisions associated with a recession given firms’ inability to foresee the recession, whereas prior studies on investment efficiency view investment decisions as intentional. Several avenues exist for future research. First, our study can be extended to other firmlevel decisions in operating and financing activities in anticipation of the business cycle. Second, understanding the characteristics of pre-acting firms and the effects on their performance after a recession would be worthwhile. Third, consideration of the 18

macroeconomic shocks in the presence of agency costs in the literature on investment efficiency would enable us to better understand the consequences of investment decisions. Further, this study elucidates the linkage between macroeconomic variables and microeconomic decisions.

19

REFERENCES Abel, A., and J. Eberly. 2002. Investment and Q with Fixed Costs: An Empirical Analysis. Working Paper, Northwestern University. Alti, A. 2003. How Sensitive Is Investment to Cash Flow When Financing Is Frictionless? The Journal of Finance 58: 707–722. Bachmann, R. and C. Bayer. 2014. Investment Dispersion and the Business Cycle. American Economic Review 104(4): 1392–1416. Benmelech, E. and C. Frydman, 2015. Military CEOs. Journal of Financial Economics 117(1): 43-59. Bernardo, Antonio E., Cai, H. and Luo, J. 2001. Capital Budgeting and Compensation with Asymmetric Information and Moral Hazard. Journal of Financial Economics 61(3): 311–344. Bernanke, B. 1983. Irreversibility, Uncertainty and Cyclical Investment. Quarterly Journal of Economics 98: 85–106. Bloom, N. 2009. The Impact of Uncertainty Shocks. Econometrica 77(3): 623–685. Bolton, P., Chen, H. and Wang, N. 2013. Market Timing, Investment, and Risk Management. Journal of Financial Economics 109(1): 40–62. Boyne, G. A. 2002. Public and Private Management: What’s The Difference? Journal of Management Studies 39, No. 1: 97-122. Brealey, R. A., I. A. Cooper, and M. A. Habib. 1997. Investment Appraisal in the Public Sector. Oxford Review of Economic Policy 13, No. 4: 12-28. Campello, M., Graham, J. R., and Harvey, C. R. 2010. The Real Effects of Financial Constraints: Evidence from A Financial Crisis. Journal of Financial Economics, 97(3), 470-487. Cochrane, J. H. 1994. Shocks. Carnegie-Rochester Conference Series on Public Policy 41: 295–364. Dittmar, A., and Duchin, R. 2016. Looking In the Rearview Mirror: The Effect of Managers’ Professional Experience on Corporate Financial Policy. Review of Financial Studies 29 (3): 565-602. Elliott, D., S. Salloy, and A. O. Santos. 2009. Assessing the Cost of Financial Regulation. IMF Working Paper.

20

Elloumi, F. and J. P. Gueyie. 2001. Financial Distress and Corporate Governance: An Empirical Analysis. Corporate Governance: The International Journal of Business In Society 1, No. 1: 15-23. Grenadier, S.R. and Wang, N. 2005. Investment Timing, Agency, and Information. Journal of Financial Economics 75(3): 493–533. Hardin, J. 2002. The Robust Variance Estimator for Two–Stage Models. Stata Journal 2(3), 253-266. Hayashi, F. 1982. Tobin’s Marginal Q and Average Q: A Neoclassical Interpretation. Econometrica 50: 213–224. Heaton, J. B., 2002, Managerial Optimism and Corporate Finance, Financial Management 31(2), 33-46. Hubbard, R. 1998. Capital-Market Imperfections and Investment. Journal of Economic Literature 36: 193–225. Jeon, H. and M. Nishihara. 2014. Macroeconomic Conditions and a Firm’s Investment Decisions. Finance Research Letters 11: 398-409. King, R. G. and Rebelo, S. T. 1999. Resuscitating Real Business Cycles. Handbook of Macroeconomics, 1, 927-1007. Malmendier, U. and Tate, G. 2005. CEO Overconfidence and Corporate Investment, Journal of Finance 60, 2661-2700. Malmendier, U., Tate, G. and Yan, J. 2011. Overconfidence and Early-Life Experiences: The Effect of Managerial Traits on Corporate Financial Policies, Journal of Finance 66, 1687-1733. Mcnichols, M. and Stubben, S. 2008. Does Earnings Management Affect Firms’ Investing Decisions? The Accounting Review 83: 1571–1603. Mclean, R.D. and Zhao, M. 2014. The Business Cycle, Investor Sentiment, and Costly External Finance. The Journal of Finance, 69(3): 1377–1409. Modigliani, F., and M. Miller. 1958. The Cost of Capital, Corporation Finance and The Theory of Investment. American Economic Review 48: 261–297. Murphy, K. and R. Topel 1985. Estimation and Inference in Two-Step Econometric Models. Journal of Business & Economic Statistics 3(4), 370-379. Nutt, P. C. 2006, Comparing Public and Private Sector Decision-Making Practices. Journal of Public Administration Research and Theory 16, 2: 289-318. Pattitoni, P. and M. Savioli. 2011. Investment Choices: Indivisible Non-Marketable Assets and Suboptimal Solutions. Economic Modelling 28, 6: 2387-2394. 21

Pattitoni, P., B. Petracci, V. Potì, and M. Spisni, M. 2013, Cost of Entrepreneurial Capital and Under-Diversification: A Euro-Mediterranean Perspective. Research in International Business and Finance 27, 1:12-27. Perfect, S. and Wiles, K. 1994. Alternative Constructions of Tobin’s Q: An Empirical Comparison. Journal of Financial Economics 1: 313–341. Shapiro, M.D. and Watson, M. 1988. Sources of Business Cycle Fluctuations. NBER Macroeconomics Annual 3: 111–148. Schoar, A. and Zuo, L. 2011, Shaped By Booms and Busts: How the Economy Impacts CEO Careers and Managerial Styles. Unpublished Working Paper, NBER. Solon, G., R. Barsky, and J. A. Parker. 1994. Measuring the Cyclicality of Real Wages: How Important Is Composition Bias. Quarterly Journal of Economics 109: 1-25. Tobin, J. 1969. A General Equilibrium Approach to Monetary Theory. Journal of Money, Credit and Banking 1: 15–29.

22

Appendix TABLE Determinants of Investment for the full sample

α 0 α1Qi ,t −1 + α 2CFit + α 3GROWTH it −1 + ε it Basic model: INVit =+

(1)

α 0 α1 INVit −1 + α 2Qit −1 + α 3CFit + α 4GROWTH it −1 + eit Augmented model: INVit =+

Quarterly Variables Basic model

INVi,t-1 Qi,t-1 CFi,t

GROWTH i,t-1

Firm fixed effects Observations R

2

(2)

Annual

Augmented model

Basic model

Augmented model

0.6302***

--

0.6972***

[0.0140]

[0.0140]

0.5996***

0.4330***

2.8449***

4.4530***

[0.0899]

[0.0410]

[0.6349]

[0.3958]

0.1818***

0.0967***

0.3015***

0.1149***

[0.0081]

[0.0044]

[0.0127]

[0.0063]

2.4070***

-0.5169

27.0877***

12.8961***

[0.4161]

[0.3288]

[1.8522]

[1.3354]

Yes

Yes

Yes

Yes

331,274

331,274

91,570

91,570

0.527

0.844

0.835

0.909

Note: This table presents multivariate regression results for the determinants of investment using the full sample, as described in Equations (1) and (2). The second and the third columns present results based on the quarterly data and the third and the fourth columns present results based on the annual data. The basic model in the second column reveals that investment is also significantly positively related to Q, cash flows and asset growth. The results in augmented model in the third column are consistent with those in basic model. However, the third column shows that the estimated coefficient on asset growth turns out to be negative and statistically insignificant. The results with annual data are consistent with those with quarterly data. The high explanatory power of R-square, ranging from 53% to 91%, allows us to use these models to estimate the normal investment levels. Clustered robust standard errors (Huber-white sandwich) at the firm level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%. Constant is included but not reported.

23

TABLE 1 NBER Quarterly and Annual Business Cycle Classifications Business cycle classification Business cycle classification Year Quarter Quarterly Annual Quarterly Annual 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1981 Recession 1997 Expansion 3rd quarter Recession 3rd quarter Expansion 4th quarter Recession 4th quarter Expansion 1st quarter Recession 1st quarter Expansion nd nd 2 quarter Recession 2 quarter Expansion 1982 Recession 1998 Expansion 3rd quarter Recession 3rd quarter Expansion 4th quarter Recession 4th quarter Expansion 1st quarter Expansion 1st quarter Expansion nd nd 2 quarter Expansion 2 quarter Expansion 1983 Recession 1999 Expansion 3rd quarter Expansion 3rd quarter Expansion 4th quarter Expansion 4th quarter Expansion 1st quarter Expansion 1st quarter Expansion nd nd 2 quarter Expansion 2 quarter Expansion 1984 Expansion 2000 Expansion 3rd quarter Expansion 3rd quarter Expansion 4th quarter Expansion 4th quarter Expansion 1st quarter Expansion 1st quarter Recession nd nd 2 quarter Expansion 2 quarter Recession 1985 Expansion 2001 Recession 3rd quarter Expansion 3rd quarter Recession 4th quarter Expansion 4th quarter Recession 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1986 Expansion 2002 Expansion 3rd quarter Expansion 3rd quarter Expansion 4th quarter Expansion 4th quarter Expansion 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1987 Expansion 2003 Expansion 3rd quarter Expansion 3rd quarter Expansion th th 4 quarter Expansion 4 quarter Expansion 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1988 Expansion 2004 Expansion 3rd quarter Expansion 3rd quarter Expansion th th 4 quarter Expansion 4 quarter Expansion 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1989 Expansion 2005 Expansion 3rd quarter Expansion 3rd quarter Expansion th th 4 quarter Expansion 4 quarter Expansion 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1990 Expansion 2006 Expansion 3rd quarter Recession 3rd quarter Expansion th th 4 quarter Recession 4 quarter Expansion 1st quarter Recession 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1991 Recession 2007 Expansion 3rd quarter Expansion 3rd quarter Expansion th th 4 quarter Expansion 4 quarter Recession 1st quarter Expansion 1st quarter Recession 2nd quarter Expansion 2nd quarter Recession 1992 Expansion 2008 Recession 3rd quarter Expansion 3rd quarter Recession th th 4 quarter Expansion 4 quarter Recession 1st quarter Expansion 1st quarter Recession 2nd quarter Expansion 2nd quarter Recession 1993 Expansion 2009 Recession 3rd quarter Expansion 3rd quarter Expansion th th 4 quarter Expansion 4 quarter Expansion 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1994 Expansion 2010 Expansion 3rd quarter Expansion 3rd quarter Expansion th th 4 quarter Expansion 4 quarter Expansion 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1995 Expansion 2011 Expansion 3rd quarter Expansion 3rd quarter Expansion th th 4 quarter Expansion 4 quarter Expansion 1st quarter Expansion 1st quarter Expansion 2nd quarter Expansion 2nd quarter Expansion 1996 Expansion 2012 Expansion 3rd quarter Expansion 3rd quarter Expansion th th 4 quarter Expansion 4 quarter Expansion Note: Each quarter is classified as an expansion/contraction following the NBER Business Cycle Dates that provide monthly frequency business cycle classification. Each year is also classified according to the majority events of that year (i.e., greater than six months). Year

Quarter

24

TABLE 2 Sample Selection Procedure Number of firms

Sample selection criteria

Number of firm-year/quarters

--

538159

13658

384762

Less: Firms in the financial and insurance industry

1904

41935

Firms in public administration industry

125

2811

15

238

Number of firms available from COMPUSTAT Number of firms with relevant data

Firms with less than 7 firms in an industry

Final sample 11614 Note: the final sample consisted of 339,778 firm-year/quarters over the period 1984 Q4-2012 Q4.

25

339778

TABLE 3 Distribution of Sample firms by Industry Industry A. Agriculture, Forestry, Fishing B. Mining

C. Construction

D. Manufacturing

E. Transportation & Public Utilities

F. Wholesale Trade

G. Retail Trade

I. Services

2 digit SIC code 01 Agricultural Production – Crops 10 Metal, Mining 12 Coal Mining 13 Oil & Gas Extraction 14 Nonmetallic Minerals, Except Fuels 15 General Building Contractors 16 Heavy Construction, Except Building 17 Special Trade Contractors 20 Food & Kindred Products 21 Tobacco Products 22 Textile Mill Products 23 Apparel & Other Textile Products 24 Lumber & Wood Products 25 Furniture & Fixtures 26 Paper & Allied Products 27 Printing & Publishing 28 Chemical & Allied Products 29 Petroleum & Coal Products 30 Rubber & Miscellaneous Plastics Products 31 Leather & Leather Products 32 Stone, Clay, & Glass Products 33 Primary Metal Industries 34 Fabricated Metal Products 35 Industrial Machinery & Equipment 36 Electronic & Other Electric Equipment 37 Transportation Equipment 38 Instruments & Related Products 39 Miscellaneous Manufacturing Industries 40 Railroad Transportation 41 Local & Interurban Passenger Transit 42 Trucking & Warehousing 44 Water Transportation 45 Transportation by Air 46 Pipelines, Except Natural Gas 47 Transportation Services 48 Communications 49 Electric, Gas, & Sanitary Services 50 Wholesale Trade - Durable Goods 51 Wholesale Trade - Nondurable Goods 52 Building Materials & Gardening Supplies 53 General Merchandise Stores 54 Food Stores 55 Automotive Dealers & Service Stations 56 Apparel & Accessory Stores 57 Furniture & Home Furnishings Stores 58 Eating & Drinking Places 59 Miscellaneous Retail 70 Hotels & Other Lodging Places 72 Personal Services 73 Business Services 75 Auto Repair, Services, & Parking 76 Miscellaneous Repair Services 78 Motion Pictures 79 Amusement & Recreation Services 80 Health Services 82 Educational Services 83 Social Services 87 Engineering & Management Services

Total

26

Firm count 20 168 27 495 26 61 38 38 226 12 67 108 50 59 96 131 916 64 121 33 64 160 140 732 891 244 829 143 20 10 83 76 77 14 52 462 311 376 226 32 70 70 52 97 73 209 288 70 36 1775 35 11 150 154 347 64 26 389 11614

Observations 543 3421 550 12527 844 1537 1270 779 7650 196 2357 3591 1783 2428 3610 4465 25811 2187 4023 1528 2283 5461 5013 22653 29255 9380 27374 4004 882 161 3015 1968 2412 429 1211 9795 9064 11912 6935 817 2259 2591 1780 4132 2190 6848 7538 1766 1334 41583 1017 236 2874 4195 8880 1595 722 13114 339778

TABLE 4 Sample Summary Statistics Panel A: Descriptive Statistics

-0.589 16.758 -0.031 -0.213

Std. Dev. 5.634 49.856 14.891 28.753

-43.331 -0.149 -397.561 -549.560

-0.825 0.285 -1.617 -5.743

0.820 1.741 -0.244 -2.170

1.989 9.229 0.892 0.797

9.127 619.400 530.111 641.149

331274

2.538

8.596

0.000

0.000

0.244

1.617

397.561

ΧΙO

331274

2.507

11.624

0.000

0.000

0.000

0.892

530.111

Qt-1 GROWTHt GROWTHt-1 CF SIZE Leverage R&D intensity RoI Ret

331274 331274 331274 331274 331274 317271 143313 322494 329632

1.898 0.019 0.021 26.435 5.198 0.064 0.518 -2.325 0.272

1.362 0.134 0.133 87.755 1.966 0.193 2.327 17.477 26.879

0.569 -2.649 -2.649 -111.000 0.554 -0.494 0.000 -118.128 -58.737

1.097 -0.025 -0.024 -0.599 3.749 0.000 0.010 -0.633 -15.244

1.444 0.010 0.010 1.841 5.131 0.000 0.066 0.587 -2.127

2.159 0.047 0.048 15.896 6.592 0.051 0.165 2.035 11.858

12.385 3.409 3.409 1106.300 10.787 13.238 20.252 33.867 109.494

Variable

N

RoA INV XINV XINV_B

331274 331274 331274 331274

ΧΙU

Mean

Min

Q1

Median

Q3

Max

Variable Definitions: RoA= net income divided by total assets (×100) INV = capital expenditures scaled by beginning-of-year net property, plant, and equipment; XINV = excess investment, measured as the residual from an industry-year regression of INV onto Q CF and GROWTH. XINV_B= excess investment, measured as the residual from an industry-year regression of INV onto Q and CF. ΧΙUit = Under-investment in the period before recession starts

ΧΙOit = Over-investment in the period before recession starts Q = Tobin’s Q (market to book value of assets) at beginning of year; GROWTH = natural log of total assets at end of prior year divided by total assets two years prior; CF = cash flow from operations scaled by beginning-of-year property, plant, and equipment; SIZE= natural log of total assets Leverage= Long term debt divided by total assets R&D intensity= R&D expenditure divide by sales RoI= net income divided by total investment Ret = market adjusted return (×100)

27

Panel B: Spearman Correlation Matrix ROA

INV

XINV

RoA

1.000

INV

0.122

1.000

XINV

-0.004

0.262

1.000

XINV_B

XINV_B

ΧΙUit

ΧΙ O it

Qit-1

GROWTHit

GROWTHt-1

CF

SIZE

Leverage

R&D int.

Roi

-0.009

0.559

0.461

1.000

ΧΙUit

0.080

0.270

-0.652

-0.104

1.000

ΧΙ O it

0.058

0.558

0.801

0.524

-0.067

1.000

-0.110

-0.011

0.002

-0.002

-0.009

-0.004

1.000

GROWTHit

0.265

0.028

0.033

0.016

-0.001

0.043

0.158

1.000

GROWTHit-1

0.168

0.022

0.001

0.001

0.020

0.017

0.219

0.052

1.000

CF

0.163

0.714

-0.005

-0.023

0.361

0.279

0.024

0.020

0.021

1.000

SIZE

0.322

0.525

0.029

0.167

0.307

0.281

-0.088

0.085

0.075

0.522

1.000

Leverage

-0.003

0.020

0.008

0.018

0.017

0.023

-0.059

0.048

0.050

0.005

0.060

1.000

R&D int.

-0.369

-0.056

0.001

0.001

-0.039

-0.029

0.166

-0.040

-0.034

-0.067

-0.115

-0.027

1.000

RoI

0.608

0.078

-0.002

-0.001

0.057

0.042

-0.078

0.161

0.122

0.105

0.222

0.015

-0.318

1.000

Ret

0.118

-0.002

0.000

-0.019

-0.002

-0.001

-0.038

0.132

0.015

0.015

0.015

-0.009

-0.017

0.064

Qit-1

28

Ret

1.000

TABLE 5 Excess Investment and Recession: a full sample analysis

XINVit = β 0 + β1 Pre RC + β 2 RC + β3 PostRC + ∑ κ i + ξit Quarterly Dependent variable

Pre-recession (RCt-6 -RCt-1) Recession (RCt) Post-recession (RCt+1 -RCt+6) Firm fixed effects Observations R

2

Annual

XINV_B

XINV

XINV_B

XINV

(1)

(2)

(3)

(4)

0.2017 [0.1879] 0.0056 [0.2234] -0.2747 [0.1686]

0.0197 [0.0617] -0.0011 [0.0726] 0.0033 [0.0634]

0.6456 [1.0759] 0.9959 [1.4376] -0.1616 [0.8873]

-0.1414 [0.6604] 0.9113 [0.8656] 0.0691 [0.5969]

Yes

Yes

Yes

Yes

331,274

331,274

91,578

91,578

0.331

0.049

0.419

0.132

Note: Table 6 presents the results of regressing excess investment on pre-recession period dummy, recession dummy, and post-recession dummy variables based on the quarterly data. Columns (1) and (3) present the results when excess investment is estimated from equation (1), the basic model. Columns (2) and (4) present the results when excess investment is estimated from equation (2), the augmented model. All variables are defined as in Table 4. Clustered robust standard errors (HuberWhite sandwich) at the firm level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%.

29

TABLE 6 Pre-acting vs Re-acting Investment Decision before Recession and Firm Performance

RoAit += θ 0 + θ1, j ΧΙitj =( UPre,O− RC ) + Wit + s Θ + ∑ κ i + ∑ yeart + ε it for s=0,…,4 s Under-investment before the recession: ΧΙUit = XINVit , XINVit < 0 and t=Pre-Recession, Over-investment before the recession: ΧΙ Oit = XINVit , XINVit > 0 and t=Pre-Recession

Panel A. Baseline analysis Dependent variable RoAi,t+s (s=0,…,4)

RoAi,t

RoAi,t+1

RoAi,t+2

(1)

(2)

(3)

RoAi,t+3

RoAi,t+4

RoAi,t

RoAi,t+1

RoAi,t+2

(4)

(5)

(6)

(7)

(8)

Pre-acting firm: ΧΙ

Reacting firms: ΧΙ

U it

ΧΙ itj =,(UPre,O− RC ) Sizei,t+s Growthi,t+s Firm fixed effects Year fixed effects Observations R-squared

RoAi,t+3

RoAi,t+4

(9)

(10)

O it

0.0083*** [0.0027] 0.8117*** [0.2172] 7.7109*** [0.4978]

0.0057** [0.0023] 0.7966*** [0.2138] 7.7088*** [0.4938]

0.0018 [0.0026] 0.7724*** [0.2127] 7.6326*** [0.4703]

0.0046* [0.0026] 0.7673*** [0.2142] 7.6782*** [0.4827]

0.0083*** [0.0023] 0.7312*** [0.2064] 7.7050*** [0.4877]

-0.0058*** [0.0014] 0.8141*** [0.2171] 7.7123*** [0.4972]

-0.0003 [0.0015] 0.7974*** [0.2138] 7.7084*** [0.4937]

0.002 [0.0017] 0.7721*** [0.2129] 7.6329*** [0.4703]

0.0003 [0.0010] 0.7679*** [0.2142] 7.6781*** [0.4826]

0.0002 [0.0022] 0.7323*** [0.2065] 7.7058*** [0.4881]

Yes Yes 355,216 0.53

Yes Yes 331,274 0.531

Yes Yes 310,766 0.534

Yes Yes 292,620 0.531

Yes Yes 282,374 0.524

Yes Yes 355,216 0.53

Yes Yes 331,274 0.531

Yes Yes 310,766 0.534

Yes Yes 292,620 0.531

Yes Yes 282,374 0.524

Note: Table 7 presents the results of regressions of post-recession ROA on excess investment and other control variables. Columns (1) – (5) show the results for the sample of pre-acting firms, (XIU). Columns (6) – (10) show the results for the sample of re-acting firms (XIO). All variables are defined as in Table 4. Clustered robust standard errors (Huber-white sandwich) at the industry level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%. Constant is included but not reported.

30

Panel B. Full Specification Dependent variable RoAi,t+s (s=0,…,4)

RoAi,t

RoAi,t+1

RoAi,t+2

(1)

(2)

(3)

RoAi,t+3

RoAi,t+4

RoAi,t

RoAi,t+1

RoAi,t+2

(4)

(5)

(6)

(7)

(8)

Pre-acting firms: ΧΙ

Reacting firms: ΧΙ

U it

ΧΙ itj =,(UPre,O− RC ) Sizei,t+s Growthi,t+s CFi,t+s R&Di,t+s Leveragei,t+s Firm fixed effects Year fixed effects Observations R-squared

RoAi,t+3

RoAi,t+4

(9)

(10)

O it

0.0063* [0.0037] 1.2392*** [0.0777] 8.5983*** [0.2144] 0.0025*** [0.0003] -0.4719*** [0.0235] -1.7588*** [0.3116]

0.0063** [0.0032] 1.1995*** [0.0791] 8.6320*** [0.2254] 0.0028*** [0.0004] -0.4707*** [0.0252] -1.6003*** [0.3014]

-0.0076 [0.0063] 1.1625*** [0.0806] 8.4722*** [0.2345] 0.0030*** [0.0004] -0.4839*** [0.0274] -1.4911*** [0.2985]

0.0011 [0.0033] 1.1545*** [0.0832] 8.4989*** [0.2422] 0.0032*** [0.0004] -0.4869*** [0.0294] -1.4742*** [0.3060]

0.0072** [0.0029] 1.1221*** [0.0848] 8.4606*** [0.2473] 0.0030*** [0.0004] -0.4957*** [0.0307] -1.5304*** [0.3126]

-0.0081*** [0.0025] 1.2406*** [0.0777] 8.5999*** [0.2144] 0.0026*** [0.0003] -0.4718*** [0.0235] -1.7574*** [0.3115]

-0.0001 [0.0022] 1.2002*** [0.0791] 8.6312*** [0.2254] 0.0028*** [0.0004] -0.4707*** [0.0252] -1.6003*** [0.3014]

-0.002 [0.0039] 1.1621*** [0.0806] 8.4726*** [0.2345] 0.0030*** [0.0004] -0.4839*** [0.0274] -1.4908*** [0.2984]

-0.0033 [0.0036] 1.1550*** [0.0832] 8.4988*** [0.2422] 0.0032*** [0.0004] -0.4869*** [0.0294] -1.4740*** [0.3060]

-0.0037 [0.0031] 1.1232*** [0.0848] 8.4601*** [0.2473] 0.0030*** [0.0004] -0.4957*** [0.0307] -1.5297*** [0.3125]

Yes Yes 147,041 0.581

Yes Yes 138,000 0.58

Yes Yes 129,980 0.583

Yes Yes 123,212 0.581

Yes Yes 119,727 0.576

Yes Yes 147,041 0.581

Yes Yes 138,000 0.58

Yes Yes 129,980 0.583

Yes Yes 123,212 0.581

Yes Yes 119,727 0.576

Note: Clustered robust standard errors (Huber-white sandwich) at the firm level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%. Constant is included but not reported.

31

Panel C. Binary measure for Excess Investment

Under-investment before the recession: ΧΙ 2Uit = 1 , XINVit < 0 and t=Pre-Recession, Over-investment before the recession: ΧΙ 2Oit = 1 , XINVit > 0 and t=Pre-Recession Dependent variable RoAi,t+s (s=0,…,4)

RoAi,t

RoAi,t+1

RoAi,t+2

(1)

(2)

(3)

RoAi,t+3

RoAi,t+4

RoAi,t

RoAi,t+1

RoAi,t+2

(4)

(5)

(6)

(7)

(8)

Pre-acting firms: ΧΙ 2

Reacting firms: ΧΙ 2

U it

ΧΙ 2itj =,(UPre,O− RC ) Sizei,t+s Growthi,t+s CFi,t+s R&Di,t+s Leveragei,t+s

Firm fixed effects Year fixed effects Observations R-squared

RoAi,t+3

RoAi,t+4

(9)

(10)

O it

0.3733*** [0.0504] 1.2402*** [0.0777] 8.6009*** [0.2144] 0.0025*** [0.0003] -0.4716*** [0.0235] -1.7594*** [0.3116]

0.0054 [0.0501] 1.2002*** [0.0791] 8.6311*** [0.2254] 0.0028*** [0.0004] -0.4707*** [0.0252] -1.6004*** [0.3014]

0.0669 [0.0513] 1.1617*** [0.0806] 8.4724*** [0.2345] 0.0030*** [0.0004] -0.4838*** [0.0274] -1.4903*** [0.2984]

0.1054** [0.0533] 1.1542*** [0.0832] 8.5004*** [0.2422] 0.0032*** [0.0004] -0.4869*** [0.0294] -1.4731*** [0.3058]

0.2821*** [0.0578] 1.1225*** [0.0848] 8.4602*** [0.2473] 0.0030*** [0.0004] -0.4959*** [0.0307] -1.5306*** [0.3125]

-0.3155*** [0.0551] 1.2405*** [0.0777] 8.6055*** [0.2144] 0.0025*** [0.0003] -0.4717*** [0.0235] -1.7563*** [0.3113]

-0.0766 [0.0526] 1.2003*** [0.0791] 8.6322*** [0.2254] 0.0028*** [0.0004] -0.4707*** [0.0252] -1.5994*** [0.3013]

-0.1384** [0.0542] 1.1624*** [0.0806] 8.4736*** [0.2345] 0.0030*** [0.0004] -0.4839*** [0.0274] -1.4902*** [0.2984]

-0.0324 [0.0552] 1.1548*** [0.0832] 8.4988*** [0.2422] 0.0032*** [0.0004] -0.4869*** [0.0294] -1.4742*** [0.3060]

0.007 [0.0588] 1.1227*** [0.0848] 8.4603*** [0.2473] 0.0030*** [0.0004] -0.4957*** [0.0307] -1.5300*** [0.3126]

Yes Yes 147,041 0.581

Yes Yes 138,000 0.58

Yes Yes 129,980 0.583

Yes Yes 123,212 0.581

Yes Yes 119,727 0.576

Yes Yes 147,041 0.581

Yes Yes 138,000 0.58

Yes Yes 129,980 0.583

Yes Yes 123,212 0.581

Yes Yes 119,727 0.576

Note: Clustered robust standard errors (Huber-white sandwich) at the firm level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%. Constant is included but not reported.

32

TABLE 7 Pre-acting vs Re-acting Investment Decision before Recession and Firm Performance: Return on Investment (RoI) Dependent variable RoIit+s (s=0,…,4)

RoIit

RoIit+1

RoIit+2

(1)

(2)

(3)

RoIit+3

RoIit+4

RoIit

RoIit+1

RoIit+2

(4)

(5)

(6)

(7)

(8) Reacting firms: ΧΙ

Pre-acting firms: ΧΙ

U it

ΧΙ itj =,(UPre,O− RC ) Sizei,t+s Growthi,t+s CFi,t+s Leveragei,t+s

Firm fixed effects Year fixed effects Observations R-squared

RoIit+3

RoIit+4

(9)

(10)

O it

0.0084** [0.0036] 2.1948*** [0.1219] 13.0212*** [0.4199] 0.0037*** [0.0004] -2.0283*** [0.2786]

-0.0012 [0.0034] 2.1870*** [0.1272] 13.0225*** [0.4395] 0.0041*** [0.0005] -1.8782*** [0.2757]

-0.0008 [0.0034] 2.1484*** [0.1315] 13.0435*** [0.4643] 0.0042*** [0.0005] -1.8282*** [0.2786]

0.0016 [0.0032] 2.1496*** [0.1354] 13.1759*** [0.4802] 0.0043*** [0.0005] -1.8899*** [0.2876]

0.0049 [0.0040] 2.0979*** [0.1373] 13.6244*** [0.4962] 0.0039*** [0.0005] -1.9848*** [0.2982]

-0.0163*** [0.0021] 2.1992*** [0.1219] 13.0316*** [0.4202] 0.0038*** [0.0004] -2.0240*** [0.2784]

-0.0055** [0.0024] 2.1883*** [0.1272] 13.0222*** [0.4395] 0.0041*** [0.0005] -1.8773*** [0.2756]

-0.0054* [0.0028] 2.1497*** [0.1315] 13.0429*** [0.4643] 0.0043*** [0.0005] -1.8279*** [0.2786]

-0.0045** [0.0022] 2.1509*** [0.1354] 13.1759*** [0.4802] 0.0043*** [0.0005] -1.8898*** [0.2875]

-0.0021 [0.0038] 2.0989*** [0.1374] 13.6248*** [0.4962] 0.0040*** [0.0005] -1.9841*** [0.2982]

Yes Yes 330,760 0.349

Yes Yes 308,770 0.356

Yes Yes 289,862 0.364

Yes Yes 273,111 0.368

Yes Yes 263,570 0.365

Yes Yes 330,760 0.349

Yes Yes 308,770 0.356

Yes Yes 289,862 0.364

Yes Yes 273,111 0.368

Yes Yes 263,570 0.365

Note: Table 8 presents the results of regressions of Return on Investment (RoI) on excess investment and other control variables. Columns (1) – (5) show the results for the sample of pre-acting firms, (XIU). Columns (6) – (10) show the results for the sample of re-acting firms (XIO). All variables are defined as in Table 4. Clustered robust standard errors (Huber-White sandwich) at the firm level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%. Constant is included but not reported.

33

TABLE 8 Pre-acting vs Re-acting Investment Decision before Recession and Firm Performance: Market adjusted return Dependent variable Retit+s (s=0,…,4)

Retit

Retit+1

Retit+2

(1)

(2)

(3)

Retit+3

Retit+4

Retit

Retit+1

Retit+2

(4)

(5)

(6)

(7)

(8)

Pre-acting firms: ΧΙ

Reacting firms: ΧΙ

U it

ΧΙ itj =,(UPre,O− RC ) Sizei,t+s Growthi,t+s CFi,t+s R&Di,t+s Leveragei,t+s

Firm fixed effects Year fixed effects Observations R-squared

Retit+3

Retit+4

(9)

(10)

O it

-0.0036 [0.0203] -5.9434*** [0.1794] 27.3426*** [0.7972] 0.0083*** [0.0012] -0.1520** [0.0594] -1.4222** [0.5544]

0.0555** [0.0244] -5.6699*** [0.1847] 26.9336*** [0.8298] 0.0088*** [0.0012] -0.1226* [0.0633] -1.7177*** [0.5934]

0.0081 [0.0231] -5.6241*** [0.1921] 26.9073*** [0.8629] 0.0097*** [0.0013] -0.1628** [0.0639] -1.6968*** [0.6126]

0.0129 [0.0225] -5.7909*** [0.2062] 26.9957*** [0.8886] 0.0105*** [0.0013] -0.1519** [0.0686] -1.9168*** [0.6240]

0.0152 [0.0236] -5.8935*** [0.2086] 26.9796*** [0.9092] 0.0098*** [0.0013] -0.1815** [0.0709] -1.4308** [0.6060]

-0.0116 [0.0175] -5.9426*** [0.1793] 27.3476*** [0.7974] 0.0083*** [0.0012] -0.1520** [0.0594] -1.4215** [0.5544]

-0.0344** [0.0165] -5.6601*** [0.1847] 26.9255*** [0.8299] 0.0089*** [0.0012] -0.1225* [0.0633] -1.7141*** [0.5934]

0.0135 [0.0217] -5.6251*** [0.1920] 26.9074*** [0.8629] 0.0097*** [0.0013] -0.1628** [0.0639] -1.6977*** [0.6127]

-0.0233 [0.0216] -5.7867*** [0.2062] 26.9946*** [0.8886] 0.0105*** [0.0013] -0.1519** [0.0686] -1.9155*** [0.6239]

0.0204 [0.0248] -5.8945*** [0.2087] 26.9798*** [0.9092] 0.0098*** [0.0013] -0.1815** [0.0709] -1.4318** [0.6062]

Yes Yes 146,221 0.092

Yes Yes 137,256 0.093

Yes Yes 129,293 0.088

Yes Yes 122,569 0.091

Yes Yes 119,107 0.096

Yes Yes 146,221 0.092

Yes Yes 137,256 0.093

Yes Yes 129,293 0.088

Yes Yes 122,569 0.091

Yes Yes 119,107 0.096

Note: Table 8 presents the results of regressions of market-adjusted returns on excess investment and other control variables. Columns (1) – (5) show the results for the sample of pre-acting firms, (XIU). Columns (6) – (10) show the results for the sample of re-acting firms (XIO). All variables are defined as in Table 4. Clustered robust standard errors (Huber-White sandwich) at the firm level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%. Constant is included but not reported.

34

TABLE 9 Pre-acting vs Re-acting Investment Decision before Recession and Firm Performance: Alternative measure for Excess Investment (XINV_B) Dependent variable RoAi,t+s (s=0,…,4)

RoAi,t

RoAi,t+1

RoAi,t+2

(1)

(2)

(3)

RoAi,t+3

RoAi,t+4

RoAi,t

RoAi,t+1

RoAi,t+2

(4)

(5)

(6)

(7)

(8)

Pre-acting firms: ΧΙ _ B

Sizei,t+s Growthi,t+s CFi,t+s R&Di,t+s Leveragei,t+s

Firm fixed effects Year fixed effects Observations R-squared

RoAi,t+4

(9)

(10)

Reacting firms: ΧΙ _ B

U it

j= U,O ΧΙ _ Bit,( Pre − RC)

RoAi,t+3 O it

0.0078*** [0.0026] 1.2405*** [0.0777] 8.5968*** [0.2144] 0.0024*** [0.0003] -0.4718*** [0.0235] -1.7581*** [0.3115]

0.0081*** [0.0023] 1.1997*** [0.0791] 8.6326*** [0.2254] 0.0028*** [0.0004] -0.4707*** [0.0252] -1.6004*** [0.3014]

0.0042* [0.0023] 1.1614*** [0.0806] 8.4729*** [0.2345] 0.0030*** [0.0004] -0.4839*** [0.0274] -1.4899*** [0.2983]

0.0102*** [0.0024] 1.1536*** [0.0832] 8.5000*** [0.2421] 0.0031*** [0.0004] -0.4869*** [0.0294] -1.4722*** [0.3057]

0.0100*** [0.0023] 1.1225*** [0.0848] 8.4600*** [0.2473] 0.0029*** [0.0004] -0.4957*** [0.0307] -1.5300*** [0.3126]

-0.0018* [0.0010] 1.2407*** [0.0778] 8.5976*** [0.2144] 0.0025*** [0.0003] -0.4719*** [0.0235] -1.7577*** [0.3115]

-0.0012 [0.0010] 1.2006*** [0.0791] 8.6311*** [0.2254] 0.0028*** [0.0004] -0.4707*** [0.0252] -1.6000*** [0.3014]

-0.0031** [0.0014] 1.1628*** [0.0806] 8.4721*** [0.2345] 0.0031*** [0.0004] -0.4839*** [0.0274] -1.4905*** [0.2984]

-0.0033*** [0.0013] 1.1559*** [0.0832] 8.4988*** [0.2422] 0.0032*** [0.0004] -0.4869*** [0.0293] -1.4738*** [0.3059]

-0.0018 [0.0014] 1.1236*** [0.0848] 8.4601*** [0.2473] 0.0030*** [0.0004] -0.4957*** [0.0307] -1.5297*** [0.3126]

Yes Yes 147,041 0.581

Yes Yes 138,000 0.58

Yes Yes 129,980 0.583

Yes Yes 123,212 0.581

Yes Yes 119,727 0.576

Yes Yes 147,041 0.581

Yes Yes 138,000 0.58

Yes Yes 129,980 0.583

Yes Yes 123,212 0.581

Yes Yes 119,727 0.576

Note: Table 10 presents the results of regressions of the absolute value of excess investment estimated from basic model. XI_BU represents the absolute value of excess investment if XINV_B is negative and XI_Bo represents the absolute value of excess investment if XINV_B is positive. All other variables are defined as in Table 4. Clustered robust standard errors (Huber-White sandwich) at the firm level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%. Constant is included but not reported.

35

TABLE 10 Annual Data: Pre-acting vs Re-acting Investment Decision before Recession and Firm Performance Dependent variable RoAi,t+s (s=0,…,4)

RoAi,t

RoAi,t+1

RoAi,t+2

RoAi,t+3

RoAi,t

RoAi,t+1

RoAi,t+2

RoAi,t+3

(1)

(2)

(3)

(4)

(5)

(6)

(7)

(8)

Reacting firms: ΧΙ

Pre-acting firms: ΧΙ

U it

j= U,O ΧΙ it,( Pre − RC)

Sizei,t+s Growthi,t+s CFi,t+s R&Di,t+s Leveragei,t+s

Firm fixed effects Year fixed effects Observations R-squared

O it

0.0099*** [0.0026] 3.1812*** [0.2571] 25.3213*** [0.5145] 0.0017*** [0.0003] -3.1300*** [0.2132] -7.0139*** [1.2225]

0.0060* [0.0032] 2.8718*** [0.2752] 25.0000*** [0.5497] 0.0020*** [0.0003] -3.2674*** [0.2740] -6.5215*** [1.2478]

0.0011 [0.0028] 2.7538*** [0.2980] 24.6485*** [0.6094] 0.0022*** [0.0003] -3.1218*** [0.2916] -6.3268*** [1.2797]

-0.0031 [0.0027] 2.5221*** [0.3148] 24.5912*** [0.6538] 0.0024*** [0.0003] -3.1698*** [0.3218] -6.0104*** [1.2843]

-0.0077*** [0.0025] 3.1869*** [0.2570] 25.3292*** [0.5149] 0.0018*** [0.0003] -3.1299*** [0.2131] -7.0117*** [1.2223]

-0.0076 [0.0062] 2.8780*** [0.2751] 24.9932*** [0.5493] 0.0021*** [0.0003] -3.2674*** [0.2740] -6.5187*** [1.2471]

-0.0027 [0.0027] 2.7563*** [0.2980] 24.6479*** [0.6094] 0.0022*** [0.0003] -3.1218*** [0.2916] -6.3258*** [1.2795]

-0.0048** [0.0020] 2.5250*** [0.3148] 24.5897*** [0.6539] 0.0025*** [0.0003] -3.1696*** [0.3218] -6.0101*** [1.2834]

Yes Yes 56,955 0.709

Yes Yes 50,080 0.704

Yes Yes 44,843 0.697

Yes Yes 40,205 0.693

Yes Yes 56,953 0.709

Yes Yes 50,078 0.704

Yes Yes 44,841 0.697

Yes Yes 40,203 0.693

Note: Table 10 presents post-recession performance for pre-acting firms (XIU) and reacting firms (XIO) based on annual data. All variables are defined as in Table 4. Clustered robust standard errors (Huber-White sandwich) at the firm level are reported respectively in brackets. *, **, and *** are respectively significance level at 10%, 5% and 1%. Constant is included but not reported.

36

Figure 1 Quarterly and Annual Business Cycle Classified by NBER (1st quarter, 1981 ~ 4th quarter, 2012)

Note: Figure 1 shows the economic cycle from the 3rd quarter of 1983 to the 4th quarter of 2012 based on the NBER criteria in terms of both economic performance and NBER announcement dates.

37

-.004

-.002

0

.002

Figure 2 Cumulative ROA: firms with over-investment and under-investment Panel A: Publicly traded firms in the non-financial sector

-1

0

1 2 time horizon (crisis=0)

cumulative ROA_over-inv

3

4

cumulative ROA_under-inv

Note: The threshold of excess investment = +1,-1: over-investment (above 1), under-investment (below -1) at t-1

-.006

-.004

-.002

0

.002

Panel B: Manufacturing firms

-1

0

2 1 time horizon (crisis=0)

cumulative ROA_over-inv

3

4

cumulative ROA_under-inv

Note: The threshold of excess investment = +1,-1: over-investment (above 1), under-investment (below -1) at t-1

38

0

.1

.2

.3

Figure 3 Cumulative excess returns: firms with over-investment and under-investment Panel A: Publicly traded firms in the non-financial sector

-1

0

1 2 time horizon (crisis=0)

cumulative ret_over-inv

3

4

cumulative ret_under-inv

Note: The threshold of excess investment = +1,-1: over-investment (above 1), under-investment (below -1) at t-1

0

.1

.2

.3

Panel B: Manufacturing firms

-1

0

1 2 time horizon (crisis=0)

cumulative ret_over-inv

3

4

cumulative ret_under-inv

Note: The threshold of excess investment = +1,-1: over-investment (above 1), under-investment (below -1) at t-1

39