Electricity Consumption, Future Earnings, and Stock Returns

Transcription

1 Electricity Consumption, Future Earnings, and Stock Returns Bok Baik* Business School, Seoul National University Gwanak_1 Gwanak-ro, Gwanak-gu, Seoul, Korea [08826] Jung Min Kim Business School, Seoul National University Gwanak_1 Gwanak-ro, Gwanak-gu, Seoul, Korea [08826] Woojin Kim Business School, Seoul National University Gwanak_1 Gwanak-ro, Gwanak-gu, Seoul, Korea [08826] * Please address all correspondence to Bok Baik, bbaik@snu.ac.kr Address: Business School, Seoul National University Gwanak_1 Gwanak-ro, Gwanak-gu, Seoul, Korea [08826] Phone: Fax: Current Draft: Jan, 2017

2 Electricity Consumption, Future Earnings, and Stock Returns Abstract In this paper, we examine whether the changes of electricity consumption are informative about future profitability and if so whether investors incorporate the implications of electricity consumption for future earnings in stock prices. To the extent that electricity consumption is necessary to produce goods and services and represents a firm s real production activity, we anticipate that electricity consumption is a leading indicator for future sales and earnings. Using hand-collected data on electricity consumption for Korean firms for the sample period of , we find a positive association between changes in electricity consumption and subsequent earnings even after controlling for other factors that may affect future profitability. More importantly, we find that changes in electricity consumption predict future stock returns. A positive relation between electricity changes and future returns still holds even after controlling for various risk factors. A hedged portfolio strategy that purchases the top quintile of the changes of electricity consumption and sells the bottom quintile of the changes of electricity consumption leads to an economically significant 3.36 to 6.48 percent annual abnormal return. Overall, our findings suggest that firms electricity consumption is useful in predicting future profitability and investors do not fully assimilate the implications of electricity consumption for future profitability. Keywords: Non-financial information; Fundamental analysis; Real production; Profitability; Stock returns JEL Classification: G11, G14, M41

3 1. Introduction The change in the value relevance of financial statements has increased interest in non-financial measures (Brown, Lo, and Lys 1999; Banker, Potter, and Srinivasan 2000; Liedtka 2002; Dichev and Tang 2008; Givoly, Hayn, and Katz 2016). Prior studies have investigated the usefulness of non-financial measures for compensation and firm valuation. For example, prior research indicates that nonfinancial metrics from customer, business process, and technology perspectives (Kaplan 1983; Kaplan and Norton 1992) can be utilized for the performance and reward systems. Researchers have also provided evidence that non-financial measures are an important value driver (Myers 1999; Trueman, Wong, and Zhang 2000; Demers and Lev 2001; Francis, Schipper, and Vincent 2003; Rajgopal, Shevlin, and Venkatachalam 2003; Ittner, Larcker, and Taylor 2009). Recently, there is a growing body of literature on distinctive non-financial measures of firms fundamental activities such as firm-level real-time corporate sales (RTCS) index or Google Trends search volume index (SVI) to predict earnings surprises (Choi and Varian 2012; Froot, Kang, Ozik, and Sadka 2016; Chiu, Huang, Teoh, and Zhang 2016). Interest in non-financial measures as predictors of future earnings and stock prices are not just restricted to academics. According to a recent Wall Street Journal report, professional investors including hedge funds actively utilize satellite images of parking lots of large retailers to get an early glimpse of future changes in revenues. 1 Related, Ljungqvist and Qian (2016) provide evidence that arbitragers visit firms production sites and utilize video recordings of production facilities to accurately assess their performance and correct mispricing. Despite the importance of non-financial measures, firms disclosures on non-financial indicators are scarce and limited to certain industries (Amir and Lev 1996; Trueman et al. 2000; Demers and Lev 2001; Liedtka 2002). It is challenging to investors to correctly forecast future earnings because they are mostly qualitative (for example, discussion on the firm s productivity and competitive advantages in MD &A). Moreover, these non-financial measures are based on firms voluntary disclosures, leading to self-selection 1 By Bradley Hope, August 14, 2016, Tiny Satellites: The Latest Innovation Hedge Funds Are Using to Get a Leg Up, The Wall Street Journal. 1

4 bias. As such, professional investors have spent enormous resources to extract alternative leading indicators (for example, by employing satellites and video surveillance or rummaging internet websites such as Amazon.com). In this study, we provide novel results that firm-level electricity consumption is useful in predicting future earnings and stock returns. Thus, we fill the gap in the literature by showing large sample evidence about real production activity and future performance. Firm-level electricity consumption has advantages over previously studied non-financial measures for several reasons. First, it is a quantitative measure applicable to a wide range of firms. Almost all firms across different industries rely on electricity as their major power source and this characteristic of electricity as a universal power source enables us to provide large sample evidence and perform a meaningful crosssectional analysis. Second, electricity consumption is an unbiased and independent measure less likely subject to managerial manipulation or self-selection than other non-financial measures. Strict regulations and extensive disclosure requirements are imposed on the utilities industry and as such, electricity consumption data is compiled by an independent external party, namely Korea Electric Power Corporation (KEPCO), which is the sole monopolistic distributor of electricity in the nation. This centralized control over electricity consumption makes it a reliable measure of firm-level real production. We expect that electricity consumption is a leading indicator of future firm performance because it captures the management s response to expected change in market conditions, like the expected change in demand. These managerial judgements on future market conditions are likely to be reflected in a firm s decision regarding investment level and production level. In other words, electricity consumption is a manifestation of management s view on a firm s future prospects. Alternatively, electricity usage may capture a firm s input investment which is realized as earnings in following months. For example, firms with a lot of profitable opportunities and on-going projects will introduce higher levels of labor input investments into the work. Subsequently, electricity usage is likely to increase as a result of longer working hours of employees. 2 Hence, electricity is a refined measure of resource input to investments, which will be realized as the firm s 2 A recent paper by Fitzgerald et al. (2015) analyzes the effects of average annual working hours on total energy consumption and discovers a positive relationship between the two. They further find out that the effect of working hours on energy consumption has increased through time. 2

5 profit in later periods. 3 Turning to the market reaction to electricity consumption, we expect no relation between current changes in electricity consumption and stock prices, to the extent that the systematic relation between current changes in electricity consumption and future profitability is fully incorporated into stock prices. However, if investors have limited attention and fail to process the implications of electricity consumption for future earnings (Hirshleifer et al. 2004), we expect to observe a significant relation between changes in electricity consumption and future stock returns. In this study, we utilize a unique informational environment for Korean firms. 4 In Korea, KEPCO (Korea Electric Power Corporation) exclusively transmits electricity to entities and records all electricity consumption within the country. We manually collect information on branch-level, or physically separate business unit-level, electricity consumption provided by KEPCO, aggregate it at the firm-level and then relate firm-level electricity consumption to future profitability and stock prices. Using a large sample of Korean firms over the period of , we examine whether electricity consumption is associated with future profitability. In a univariate analysis, we find strong evidence that changes of electricity consumption predict future earnings changes. We find that those firms with an increase in electricity consumption tend to report higher earnings even after controlling for the fundamental signals from prior research, suggesting that firms with increased electricity consumption are more likely to report higher future earnings. We also examine whether the informative content of the changes of electricity consumption on future profitability is fully incorporated in the market. We find a significant positive relation between changes in electricity consumption and future returns, suggesting that investors do not fully impound the implication of electricity consumption changes for future earnings. This positive relationship between changes in electricity consumption and future 3 Da, Yang and Yun (2014) use residential electricity usage as a direct proxy for the service flow from household production capital. Their line of reasoning is similar with ours in that they rely on the role of electricity usage as a measure of capital investment. 4 Korea offers a particularly unique setting that allows us to examine the effect of electricity consumption on accounting profitability and stock prices at the firm level. Unlike the deregulated U.S. electricity industry, electricity customers in Korea are served by a regulated monopoly utility, Korea Electric Power Corporation (KEPCO). Currently, 6 power generation companies, independent power producers, and community energy systems produce electric power in Korea. But unlike the production of electricity, which is carried out by several producers, the distribution of electricity is exclusively undertaken by KEPCO. KEPCO solely transports the electric power it purchases from the Korea Power Exchange through the transmission and distribution networks, and sells to end-use customers including firms. In other words, KEPCO, as a highly regulated monopolistic electricity transmitter, has a tight control over the transmission networks and electricity service. As such, electricity usage data provided by KEPCO is a comprehensive representation of the entire electricity consumption in the nation. 3

6 returns continues to hold even after controlling for risk factors. To gauge the economic significance, we form a hedged portfolio that takes a long position in the highest quintile of electricity changes and a short position in the lowest quintile of electricity changes and find that this hedge strategy generates a 3.36 to 6.48 percent excess annual return. We further condition the relation between changes in electricity consumption and future returns on information asymmetry and discover that the positive relation is most salient for firms with high information asymmetry. To summarize, these results support the hypothesis that electricity consumption is an important value driver but financial markets do not quickly incorporate the implications of electricity consumption for future profitability into stock prices. We contribute to prior research in several ways. First, we contribute to a stream of research documenting the role of non-financial indicators (Deng, Lev, and Narin 1999; Amir and Lev 1996; Ittner and Larcker 1998; Demers and Lev 2001; Rajgopal et al. 2003; Ittner et al. 2009; Chiu et al. 2016; Froot et al. 2016). Our research is also related to extant research on fundamental signals (Ou and Penman 1989; Lev and Thiagarajan 1993; Abarbanell and Bushee 1997; Piotroski 2000; Beneish, Lee and Tarpley 2001; Mohanram 2005; Piotroski and So 2012) because we highlight the informativeness of electricity consumption for future performance. In addition, our study adds to prior research on analysts forecasting ability by documenting that electricity consumption is useful for future profitability analysis (Fairfield, Sweeney, and Yohn 1996; Banker and Chen 2006). Finally, our study adds to the literature on stock price anomalies (for example, Sloan 1996; Hirshleifer, Hsu, and Li 2013; Rapach and Zhou 2013) by showing that firm-level electricity consumption, a leading indicator of a firm s future performance, is mispriced in the market. This finding should be of interest to investors where real production activity measures are available since we provide compelling evidence on the usefulness of real production measures to predict future earnings and stock prices. Specifically, our study is related to recent work by Da, Huang, and Yun (2016) reporting that the industrial electricity usage growth rate is associated with aggregate future stock returns. We complement Da et al. (2016) by showing that firmlevel electricity consumption is informative about individual stock s future returns. Our empirical also has policy implications. For the financial regulators, our study suggests that additional disclosure requirements on 4

7 a firm s real production activity such as electricity consumption may help to facilitate price discovery. We also acknowledge that our study has certain limitations. First, the fact that our results pertain to Korean firms means that they may not generalize to firms in other countries. Second, in contrast to Korean firms, we cannot directly observe electricity usage on US or Canadian firms, implying that investors may not be able to directly exploit electricity information to forecast future earnings. However, analysts and investors can solicit such information from conference calls. More importantly, the regulator may consider recommending that firms report audited utility usage such as electricity, water, and gas in their annual reports. The rest of this paper is organized as follows. In section 2, we develop our hypotheses with a review of the literature. We describe research design and data in section 3. We report results of our hypothesis testing in section 4. Section 5 provides the summary and conclusions. 2. Related Literature and Hypothesis Development This paper extends prior research on non-financial leading indicators and stock price anomalies by introducing an electricity consumption measure, a non-financial leading indicator, to predict future performance. A stream of research examines the relation between fundamental signals and future performance (see a detailed review by Richardson et al. 2010). Ou and Penman (1989) suggest the Pr-measure using financial ratios from financial statements and predict the direction of future earnings changes. They also devise a trading strategy based on the predictions. Holthausen and Larcker (1992) extend Ou and Penman (1989) by employing a logit model to forecast future stock returns based on accounting ratios. In another study, Lev and Thiagarajan (1993) adopt a regression framework to examine the value-relevance of twelve fundamental signals from analysts research reports. They find that financial signals such as inventory, capital expenditure, and gross margin, are closely related to contemporaneous returns. Similarly, Abarbanell and Bushee (1997) find a significant relationship between changes in fundamental signals and subsequent earnings changes. Extending the findings from Lev and Thiagarajan (1993) and Abarbanell and Bushee (1997), Abarbanell and Bushee (1998) focus on nine variables out of Lev and Thiagarajan (1993) and show that these signals 5

8 anticipate both future earnings changes and analyst revisions. They further find that the fundamental signals generate abnormal returns. Beneish et al (2001) also document the usefulness of contextual fundamental analysis in predicting extreme stock returns, and Mohanram (2005) combines traditional fundamentals with measures tailored for growth firms to earn significant excess returns. Additionally, Piotroski (2000) and Piotroski and So (2012) underscore the importance of historical financial statements by showing that even within the portfolio of high book-to-market firms, financial statement information separates winners from losers. Overall, these findings suggest that financial measures contain value-relevant information. While prior research suggests that the information contained in financial statements is value-relevant, there also exists another stream of research showing that the explanatory power of accounting numbers has decreased over the last few decades (Francis and Schipper 1999; Collins et al. 1997; Brown et al. 1999). Relatedly, several studies document that non-financial indicators such as satisfaction measures (Ittner and Larcker 1998; Ittner et al. 2009), patent (Deng et al. 1999; Hirshleifer et al. 2013), market penetration (Amir and Lev 1996), order backlogs (Rajgopal et al. 2003), eyeball measures and web traffic performance measures in the internet industry (Trueman et al. 2000; Demers and Lev 2001) are value-relevant and informative about future performance. These studies imply that non-financial measures can be leading indicators of financial performance. Recently, several papers propose distinctive non-financial measures of firms fundamental activities. For example, Froot et al. (2016) construct firm-level real-time corporate sales (RTCS) index from multiple sources including 50 million mobile devices. They find that their measure of RTCS index effectively captures firm-specific real-time economic activities which track consumer activities for US big-box retail firms. Several papers also analyze Google Trends search volume index (SVI) as a proxy for potential customer demand (Choi and Varian 2012; Chiu et al. 2016). They provide evidence that SVI can forecast future sales. These studies suggest a growing interest in unique non-financial data that are indicative of firms fundamental operating activities. In this study, we extend prior research on fundamental analysis and non-financial leading indicators by introducing a unique measure of a firm s real activity to predict future performance. We collect 6

9 information on a firm s electricity consumption to examine whether this measure is informative about future profitability and stock performance. While there is abundant research on fundamental signals, we are unaware of any research that relates firm-level electricity consumption to its future performance. Recent study by Da et al. (2016) employ U.S. industrial electricity consumption to predict aggregate stock market returns and find that high industrial electricity consumption predicts low stock market returns in the future. However, they stop short of examining the impact of firm-level electricity consumption on individual stock returns due to data constraints. By utilizing a unique setting for Korean firms, we investigate whether electricity consumption information can be used to predict future performance in a firm-level analysis. We rely on electricity consumption as our key measure because it is likely a manifestation of managerial judgements regarding the firm s future earnings prospect and because it effectively captures firmlevel input investment that will subsequently be realized as gains. We expect that firms that increase electricity consumption are likely to be the ones increasing their production levels in accordance with projected demand. Consequently, they will exhibit better operating performance than firms that do not. Alternatively but not mutually exclusively, it is also possible that electricity consumption is a reflection of firm-level resource inputs into projects that will generate earnings in later periods. To the extent that electricity consumption reflects production levels to match with varying demand and to the extent it captures investment levels preceding future profits, we expect to observe a positive relation between changes in electricity consumption and future earnings changes. Our hypothesis follows, as stated in an alternative form: HYPOTHESIS 1. Changes in electricity consumption are positively associated with future earnings changes. Furthermore, we examine whether investors see through the implication of changes in electricity consumption for future profitability, if any. In an efficient market, any systematic relation will be impounded efficiently into stock prices so we do not observe any relation between the changes in electricity consumption and future stock prices. However, it is also possible that investors fail to understand the implication of 7

10 electricity consumption for subsequent performance because electricity consumption data is compiled at branch-level in the first place and thus requires investors to convert them at the firm-level. Moreover, due to its high cyclicality, utilizing electricity consumption information requires understanding its seasonal nature and devising accurate measurement model based on seasonally adjusted changes in electricity consumption. Provided that investors fail to fully see through the implication of electricity consumption and slowly respond to the information, there should be a significant positive relation between a firm s electricity consumption growth and subsequent stock returns. Thus, whether or not investors unravel this relation is an empirical question. We generate our second hypothesis, formulated in an alternative form as follows: HYPOTHESIS 2. Changes in electricity consumption are positively associated with future stock returns. 3. Data and Research Design 3.1. Data We report our sample reconciliation procedure in Table 1.. Our primary electricity usage data are assembled from a proprietary raw dataset obtained from KEPCO which provides monthly electricity consumption for all types of entities including factories, universities, hospitals, government organizations and residential buildings. In order to ensure economic significance of our analysis, we limit the sample to the entities having power contract of at least 5,000 kwh on an annual basis. This requirement yields our initial sample consisting of 231,246 branch-month observations from year 2005 to [Insert Table 1 about here] We delete branch-month observations if the number of monthly observations for a given year is less than 12, losing 8,154 branch-months. To mitigate the effect of the irregularities of electricity consumption when KEPCO first begins to provide electricity to a certain client firm, we also eliminate first six months of 5 Since electricity usage figures that sporadically appear or disappear in the middle of our sample period are likely to yield extreme outliers, we calculate the standard deviation of electricity usage (ELECSTD) and delete branch-months if ELECSTD exceeds 1. ELECSTD is computed as the standard deviation of firm-level electricity usage divided by the average firm-level electricity usage, calculated at branch-level. 8

11 branch-month observations from the sample unless the electricity consumption figure starts from January 2005, the very first month of our sample period. This leads our sample to amount to 218,004 branch-month observations. Next, we transform branch-level consumption into firm-level consumption by aggregating the branch-month observations into 54,312 firm-months. 6 Additionally, we require that our sample firms have stock prices and accounting information on KISVALUE and FNDATAGUIDE (a local provider comparable to CRSP and Compustat, respectively), yielding 39,828 firm-month observations. To mitigate backfilling biases, we also require that a firm must be listed in the sample for 6 months before it is included in the data set (Fama and French 1993). Our final sample comprises 38,876 firm-month observations for the sample period of Research Design To test the association between changes in electricity consumption and subsequent earnings, we follow prior research that has examined the indicators of future profitability growth (Abarbanell and Bushee 1997; Soliman 2008) and estimate the following OLS regression model: (1) ΔROA q+1 is subsequent quarter s year over year changes in ROA (i.e., ROA q+1 ROA q 3 ). Throughout the analysis, subscript q refers to the quarter to which the month t belongs. ΔELEC t is our main variable of interest. We calculate the variable as the year over year growth in a firm s monthly electricity consumption to account for seasonality in electricity consumption (i.e., (Electricity Consumption t Electricity Consumption t-12 ) / Electricity Consumption t-12 ). SIZE t is the natural logarithm of (1 + a firm s market 6 We manually collect the names and stock codes of the KOSPI (the main bourse) and KOSDAQ (tech bourse similar to NASDAQ) firms to ensure that the names of the firms in the electricity consumption data are correctly matched with KISVALUE (comparable to CRSP) and FNDATAGUIDE (comparable to COMPUSTAT). In case there are more than two firms with the same name, we use the firm s location information and the name of the firm s CEO to match each firm with an appropriate stock code. 9

12 capitalization divided by 100,000). BTM t is the natural logarithm of (1 + a firm s book value of equity to market value of equity). PASTRET t is the monthly compounded return of a firm for period [t 12, t 1], and LEVERAGE t is the natural logarithm of a firm s total asset deflated by book value of equity. ΔROA q is the year over year change in ROA (i.e., ROA q ROA q 4 ), where ROA q is calculated as the net income of quarter q, divided by the beginning of the quarter total assets, multiplied by 100. D_LOSS t is an indicator variable equal to one for firms reporting a loss in the quarter to which the month t belongs, and zero otherwise. 7 As we hypothesize a positive relation between electricity changes and future profitability changes, we expect a positive β 1 estimate. To explore whether the market recognizes the implications of changes in electricity consumption for future profitability, we estimate cross-sectional regressions forecasting one-month-ahead stock returns with changes in electricity consumption, ΔELEC t, and a host of characteristics that we use as controls as follows: (2) We include control variables indicated by the prior literature (Rajgopal et al. 2003; Jung, Wong, and Zhang 2014). Specifically, we include size (SIZE t ), book-to-market (BTM t ), past returns (PASTRET t ), leverage (LEVERAGE t ) and earnings to price ratio (E/P t ). E/P t is a firm s net income divided by the ending price of a firm s stock, divided by 1,000. All other variables are defined in Appendix. If the market does not fully incorporate the implication of the changes of electricity consumption for future profitability, we anticipate a significant and positive β Empirical Results [Insert Table 2 about here] 7 Because information from financial statements is available only at the quarterly level, the most recent quarter s financial statement information is used in calculating monthly control variables when necessary. 10

13 We report the descriptive statistics of the variables used for our analysis in Table 2. In order to maintain consistency with our regression analysis, we follow the Fama-Macbeth approach in calculating the descriptive statistics. Our key variable of interest is the changes of electricity consumption denoted as ΔELEC t. As can be seen from the table, the changes of electricity consumption have a mean of 0.078, implying that the average firm s year over year monthly electricity growth is 7.8 percent. We find that ΔELEC t has a standard deviation of 0.068, suggesting substantial variation in monthly electricity consumption. Both mean and median values of one-month-ahead stock returns, RET t+1, are 1.1 percent. We also report that one-quarterahead changes in ROA, denoted as ΔROA q+1 have a mean of We additionally consider a variety of control variables from prior literature that are potentially correlated with future profitability and stock returns. The mean value for BETA t, defined as the market beta calculated over previous 60 months, is SIZE t shows a mean (median) of (1.433). The mean (median) values for BTM t and PASTRET t are (0.823) and (0.139), respectively. These variables are included in our analysis to control for the size effect, the value effect and the momentum effect. LEVERAGE t has mean and median values of and with a standard deviation of 0.299, and E/P t, a firm s net income divided by stock price, has mean and median values of and ΔROA q is negative on average, with a mean of D_LOSS t exhibits a mean value of 0.248, indicating that about 25 percent of our sample firms incur losses. [Insert Table 3 about here] Table 3 presents correlations among the key variables. As expected, the changes of electricity consumption are positively related to the changes in one-quarter-ahead profitability at the 1 percent level. This relationship implies that increased electricity consumption is informative about an improvement in future profitability. We also find that the changes of electricity consumption are positively related to one-monthahead stock returns. These results provide initial evidence consistent with our hypotheses. The correlation coefficients among the control variables indicate that except for a few pairs, correlations between the variables are largely statistically significant. However, the magnitudes of the correlations are not that large. For example, the largest correlation (in absolute terms) is between changes in 11

14 ROA and contemporaneous ROA, which is Most remaining correlations are at modest levels, implying that multicollinearity would not be a serious concern. Nevertheless, we appropriately control for these control variables in a multivariate framework to establish a causal link between electricity consumption and future performance. [Insert Table 4 about here] To test the impact of changes in electricity consumption on future profitability, we estimate model (1) and present the results in Table 4. Specifically, we regress changes in ROA one quarter ahead against current changes in electricity consumption cross-sectionally for every calendar month during our sample period, and take the time-series averages of the coefficients to obtain t-statistics. In the first three columns, cross-sectional regressions are based on raw continuous variables, while the last column is based on decile ranks of each explanatory variable. We also control for a variety of firm characteristics in both panels that could potentially affect changes in future profitability. The results reported in panel A of Table 4 indicate that changes in current electricity consumption are positively associated with subsequent earnings changes even after controlling for size, book to market, past returns, leverage, current ROA and changes in contemporaneous ROA. The economic significance is also non-trivial. Specifically, a one standard deviation increase in changes in electricity consumption leads to 2.3 to 2.5 percent increase in subsequent changes in ROA. Because our dependent variable, ROA q+1, is measured at a quarterly interval, we also regress ROA q+1 on year over year changes in quarterly electricity consumption as a robustness check. Panel B of Table 4 reports time-series average of the coefficients estimated from the 36 cross-sectional regressions from 2006 Q1 to 2014 Q4. Note that our main variable of interest, ΔELEC qtr is measured over a quarter in this analysis. The results reported in panel B of Table 4 mirror largely those based on monthly changes in electricity consumption reported in panel A of Table 4. These results support our hypothesis that electricity consumption predicts future profitability. 12

15 Thus far, we find evidence that the changes of electricity consumption predict a firm s future profitability. To further test the predictive ability of electricity consumption in the stock market, we analyze whether changes in electricity consumption are predictors of future returns. [Insert Figure 1 about here] In Figure 1, we report annualized hedged portfolio returns for each calendar year during our sample period. The hedged portfolio is constructed by ranking firms into deciles based on ΔELEC t each month, and then taking a long position in the highest decile firms and a short position in the lowest decile firms. The annualized return results indicate that except for 2006 and 2011 where the returns are slightly negative, the returns for all other years are positive and economically significant. The results reported in Figure 1 provide preliminary evidence for hypothesis 2 and suggest a positive association between changes in electricity consumption and future returns. [Insert Table 5 about here] To parsimoniously test whether changes in electricity consumption predict future stock returns, we estimate model (2). Specifically, for every calendar month during our sample period, we regress monthly returns on previous month s changes in electricity consumption and additional control variables crosssectionally. We then take the time-series averages and standard errors to obtain t-statistics. In Table 5, we report the regression results with future returns as the dependent variable. In Panel A, we employ year-overyear changes in monthly electricity consumption to see whether they have predictive ability on one-monthahead returns. The first three columns are based on raw continuous variables, while the last column is based on decile ranks to mitigate the impact of outliers. Column (1) presents the regression results of the return predictability of changes in electricity consumption, controlling for market beta, firm size, book-to-market and past returns. We find that the coefficient on ΔELEC t is positive and significant at the 1 percent level, suggesting that a one unit increase in electricity consumption leads to a 0.7 percent increase in subsequent returns. In columns (2) and (3), we include additional control variables, LEVERAGE t and E/P t, in the regression analysis. The coefficient estimate of ΔELEC t reported in column (2) continues to be positive and statistically significant at the 1 percent level, indicating that the return predictability of changes in electricity 13

16 consumption holds even after additionally controlling for a firm s leverage. In column (3), we continue to find that changes in electricity consumption are positively associated with future returns. In the last column, we use decile-ranked variables to check the robustness of our findings to outliers, and find that the coefficient on ΔELEC t is positive and statistically significant with a t-statistic of All the other control variables that are statistically significant exhibit signs in the expected directions. In Panel B, we additionally test whether yearover-year changes in quarterly electricity consumption are predictive of subsequent three months returns. The results are in line with the ones we reported in Table 5, Panel A as we continue to observe a significant and positive relationship between changes in quarterly electricity consumption and future stock returns. To summarize, the results clearly indicate that increases in electricity consumption lead to higher stock returns in the subsequent month, consistent with hypothesis 2. To gauge the economic significance of the relation between changes in electricity consumption and future returns, we calculate returns for quintile portfolios sorted by changes in electricity consumption. Specifically, for each month during our sample period, we sort all stocks in our sample based on electricity consumption of the previous month and compute size-adjusted abnormal returns and DGTW benchmarkadjusted abnormal returns suggested by Daniel, Grinblatt, Titman, and Wermers (1997). [Insert Table 6 about here] The results reported in Table 6 indicate that abnormal returns are increasing with changes of electricity consumption. Specifically, the size-adjusted abnormal return for the top quintile is 0.25 percent per month, while the corresponding figure is 0.29 percent per month for the bottom quintile. We observe a similar increasing pattern for DGTW benchmark-adjusted abnormal return as we move from the lowest quintile to the highest quintile. Looking at the differences between the abnormal returns of the lowest and the highest quintiles, we can see that the hedged portfolio return obtained by buying top quintile portfolio stocks and shorting the bottom quintile portfolio stocks yields an average size adjusted monthly return of 0.54 percent and DGTW benchmark adjusted return of 0.52 percent. Based on either abnormal return measure, the hedged return is statistically significant. Moreover, the results in Figure 1 show that the hedged portfolio return is positive and significant in 7 out of the 9 years, implying stability of the excess returns across years. 14

17 We also calculate abnormal hedged portfolio returns using the Fama-French factor model. The excess return estimates, or alphas, are calculated using monthly returns for quintiles sorted by ΔELEC t based on the three-factor model and the four-factor model, respectively. In an untabulated analysis, we find that that the excess returns increase as we move from the lowest quintile to the highest quintile and that the differences between the highest quintile and the lowest quintile excess returns are statistically significant. Specifically, a hedged portfolio return generates an economically significant 0.40 percent (0.28 percent) monthly abnormal return based on the three-factor (four-factor) model, corroborating our inferences about the association between electricity consumption and future stock returns. [Insert Figure 2 about here] We acknowledge that one potential concern with respect to electricity consumption is that it exhibits a highly seasonal pattern. Figure 2 reports the average normalized electricity consumption and average normalized energy degree days (EDD) for each month. 8 A firm s electricity consumption for a given month t is scaled by its own annual electricity consumption of the year to which month t belongs. The normalized electricity consumption in each month is then averaged across different firms in our entire sample to yield a single representative value for each month. Figure 2 clearly indicates that electricity consumption hits its peak during winter when there is a lot of heating demand. In order to illustrate the temperature movement along with the electricity consumption, we calculate the EDDs for each month and plot the normalized EDDs in Figure 2, coupled with the normalized electricity consumption. In calculating EDDs, we first calculate cooling degree days (CDD) defined as max[0, (T max +T min )/2 18 C], and heating degree days (HDD) defined as min[0, 18 C (T max +T min )/2], where T max (T min ) is the maximum(minimum) temperature during that month. These measures are designed to capture deviations from 18 degrees Celsius, the temperature at which energy is least consumed. We then add CDD and HDD for a given month to obtain EDDs. As expected, Figure 2 illustrates that EDDs are high during winters and summers, while they are low during springs and falls. 8 In order to control for weather fluctuations in the analysis, we extract information on temperature from National Climate Data Service System (NCDSS). 15

18 To control for the seasonal effect described above, we conduct a two-stage regression analysis. In the first stage regression, we estimate the following rolling regression for the 24 months prior to month t to create our firm-specific measure of weather sensitivity. (3) Using the estimates from the rolling regression, we estimate the predicted weather effect by calculating the fitted ELEC t for each firm month. This measure is specific to each firm month and it represents the incremental effect of year over year temperature deviation on firm-level changes in electricity consumption. We use past 24 months of ELEC t and weather data, and require at least past 10 months of data for each firm-month regression. Temp_Deviation is defined as the temperature difference between current month and the same month of prior year (Temperature t Temperature t 12 ). SpringFall, Summer and Winter are indicator variables for each season devised to capture different implications of temperature deviation on electricity consumption for each season. 9 Using the estimates from the first-stage regression, we obtain the weather-adjusted ELEC t, or the fitted ELEC t for each firm month observation. The difference between actual ELEC t and the fitted ELEC t is defined as ELEC t (residual). We then use these residuals as regressors in the second stage regression to predict the next month s stock returns. We report the second-stage regression results in Table 7. [Insert Table 7 about here] The first three columns report the coefficient estimates on raw variables and the last column on decile ranked variables. We find that the coefficient estimates on ELEC t (residual) is positive and statistically significant, suggesting that changes in electricity consumption, after netting out the effect of temperature changes, still predict future stock returns. In summary, the results reported in Table 7 suggest that return predictability of electricity consumption continues to hold even after controlling for any seasonal effect. 9 Firm months ending in Jan-Mar are considered to be winter and firm months ending in July-Sep are considered to be summer. The rest of the firm months (i.e., firm months ending in Apr-June and Oct-Dec) are considered to be spring and fall. 16

19 [Insert Table 8 about here] In Table 8, we implement two sets of additional robustness tests. In Panel A, as an alternative measure for changes in electricity consumption, we scale the year-over-year difference of electricity consumption by average total asset. This scaling effectively yields a growth rate in electricity consumption controlling for firm size. As shown in Table 8, panel A, we confirm that the positive correlation between electricity consumption and future stock returns still holds under this alternative definition of changes in electricity consumption. We also control for a number of other explanatory variables and report the results in panel B of Table 8. In the first column, we control for changes in industry-level electricity consumption based on 2-digit SIC industry classification. The results again suggest that industry-level changes in electricity consumption do not drive our results. In the second column, we include quarterly accruals as an additional control variable to make sure that the accruals anomaly is not driving our results. The results suggest that the changes in electricity consumption still maintain the predictive power even after controlling for the accruals. In the third to fifth columns, we include control variables that capture year-over-year changes in the number of employees, employment costs, and total wage, respectively. The coefficient estimates on changes in electricity consumption continue to be positive and statistically significant, indicating that the measure has predictive power beyond other employee-related non-financial measures. It is likely that investors find it hard to collect and process information for firms with greater information asymmetry. If investors do not fully incorporate and underreact to the extent to which changes in electricity consumption signal future earnings for opaque firms, we expect that the mispricing of electricity consumption will be more pronounced for firms with greater information asymmetry. To gain further insight into the relation between electricity consumption and future returns, we split firms into two groups based on the degree of information asymmetry and examine whether the relation differs between the two groups. We consider proxies for information asymmetry such as firm size, R&D (research and development expense scaled by sales), and return volatility (variance of monthly returns calculated each year, relative to the industry average) following prior research (Collins et al. 1987; Aboody and Lev 2000). To the extent that stock return predictability is attributed to market participants mispricing of electricity consumption, we 17

20 anticipate that the relation between the changes of electricity consumption and future returns will be more pronounced in firms with greater information asymmetry (i.e., small firms, high R&D firms, and firms with high return volatility). We report the results in Table 9. Consistent with our expectation, we find that the positive association between changes in electricity and future stock returns is largely driven by high information asymmetry firms such as small firms, high R&D firms, and firms with higher return volatility. [Insert Table 9 about here] In the first two columns, we report regression results for subsamples of firms divided based on the firm size. As expected, ΔELEC t has a strong return-predictive power for a subsample of small firms with a t- statistic of In comparison, for a subsample of large firms, we find an ΔELEC t estimate with a t-statistic of only The third and fourth columns report regression results for the subsamples partitioned by R&D expenditures (relative to sales). We find that the return predictability of ΔELEC t is observed only in high R&D firms. Specifically, we observe that the coefficient estimate on ΔELEC t is statistically significant for high R&D firms (estimate = 0.008, t-stat = 2.31) but we do not find any significance in low R&D firms (estimate = 0.004, t-stat = 0.94). In the last two columns, we also run separate regressions based on return volatility. We document that the coefficient on ΔELEC t is (t-statistics = 2.18) for firms with higher return volatility than the industry average and (t-statistics = 1.94) for firms with lower return volatility than the industry average. This finding further corroborates the view that the predictability of ΔELEC t for future returns is stronger for firms with greater information asymmetry. Overall, these findings suggest that investors do not recognize the implication of electricity consumption growth for future profitability and slowly impound the information into stock prices at the firm level. [Insert Table 10 about here] To reconcile the firm-level results with the seemingly contradictory market level results reported by Da et al. (2016), we aggregate electricity consumption and repeat the analysis. In Table 10, we show the return predictability of changes in electricity consumption at the aggregate level. Using cumulative aggregate market returns up to 1-month, 3-month, 6-month, 9-month and 12-month ahead, we find that the aggregate 18

21 changes in electricity consumption are negatively correlated with future aggregate returns at a marginally significant level. This is consistent with the findings from Da et al. (2016) where high aggregate industrial electricity usage today is found to be predictive of low stock returns in the future. This evidence suggests that the negative relation between changes in electricity consumption and market returns at an aggregate level coexist with the positive relationship we find at a firm level. 5. Summary and conclusions Despite great interest in evaluating firms non-financial information for future financial performance, there is limited large sample evidence regarding the impact of a firm s real production activity on firm performance. In this paper, we complement extant literature by testing whether changes in firm-level electricity consumption would be useful in predicting a firm s future performance. In other words, we study whether the changes of electricity consumption are informative about future profitability and whether market participants fully recognize the implication of the changes for subsequent accounting profitability. As expected, we find that changes in a firm s electricity consumption have incremental ability to predict subsequent earnings changes, implying that firms which increased (decreased) electricity consumption exhibit higher (lower) subsequent earnings. We find that this explanatory power is not subsumed by other firm characteristics and fundamental signals that may affect future performance. More important, we also find that unexpected changes in electricity consumption are associated with substantial future abnormal returns, even after controlling for various risk factors. We also report that the relation between changes in electricity consumption and abnormal returns is pronounced in firms with greater information asymmetries. This evidence is consistent with the market s mispricing of electricity consumption. Overall, we document that electricity consumption is an important driver of firm value. Our evidence highlights the role of electricity consumption, a parsimonious proxy for a firm s real production activity, in a capital market research context and underscores the importance of non-financial leading indicators for firm valuation. 19

22 References Abarbanell, J. S., and Bushee, B. J Fundamental analysis, future earnings, and stock prices. Journal of Accounting Research 35(1): Abarbanell, J. S., and Bushee, B. J Abnormal returns to a fundamental analysis strategy. The Accounting Review 73(1): Aboody, D., and Lev, B Information asymmetry, R&D, and insider gains. The Journal of Finance 55(6): Amir, E., and Lev, B Value-relevance of nonfinancial information: The wireless communications industry. Journal of Accounting and Economics 22(1): Atiase, R. K Predisclosure information, firm capitalization, and security price behavior around earnings announcements. Journal of Accounting Research Banker, R. D.,and Chen, L Predicting earnings using a model based on cost variability and cost stickiness. The Accounting Review 81(2): Banker, R. D., Potter, G., and Srinivasan, D An empirical investigation of an incentive plan that includes nonfinancial performance measures. The Accounting Review 75(1): Beneish, M. D., Lee, C. M., & Tarpley, R. L Contextual fundamental analysis through the prediction of extreme returns. Review of Accounting Studies 6(2-3): Brown, S., Lo, K., and Lys, T Use of R 2 in accounting research: measuring changes in value relevance over the last four decades. Journal of Accounting and Economics 28(2): Chiu, P., Huang, X., Teoh, S. H., Zhang, Y Using Google searches of firm products to Nowcast sales revenues and detect revenue management. Working paper. Choi, H., and Varian, H Predicting the present with Google Trends. Economic Record 88(s1): 2-9. Collins, D. W., Kothari, S. P., and Rayburn, J. D Firm size and the information content of prices with respect to earnings. Journal of Accounting and Economics 9(2): Collins, D. W., Maydew, E. L., and Weiss, I. S Changes in the value-relevance of earnings and book values over the past forty years. Journal of Accounting and Economics 24(1): Da, Z., Huang, D., and Yun, H Industrial electricity consumption and stock returns. Journal of Financial and Quantitative Analysis forthcoming. Da, Z., Yang, W., and Yun, H Household production and asset prices. Management Science 62(2): Daniel, K., Grinblatt, M., Titman, S., and Wermers, R Measuring mutual fund performance with characteristic based benchmarks. The Journal of Finance 52(3): Demers, E. and B. Lev A Rude Awakening: Internet Shakeout in Review of Accounting Studies 6: