Reexamining the (Mis)pricing of Customer
Satisfaction and the Impact of Economic
Situation
Jorick G.M. Beernink
Nassaulaan 61, 9717CH, Groningen, me@jorickbeernink.nl
Abstract
This paper examines the possible mispricing of customer satisfaction and extends previous work by testing the influence of economic situation on the relationship between customer satisfaction and stock returns. Stock portfolios are formed based on customer satisfaction scores. The sample period of this study is from April, 1996 to September, 2013. The abnormal returns of the portfolios are examined using Carhart’s four-factor model. Results indicate an insignificant alpha during up and down markets. We conclude that (lagged) change in customer satisfaction has no strong predictive power for stock returns and that customer satisfaction causes no excess returns but lowers systematic returns. These findings hold over different time intervals, as well as over periods of up and down market. There is no clear indication for the mispricing of customer satisfaction. However, increased customer satisfaction lowers systematic risk.
Keywords
Customer Satisfaction, Firm Value, Portfolio Analysis, (Mis)pricing
University of Groningen Faculty of Economics and Business
MSc Finance & MSc Marketing
Groningen,
Introduction
Measurement is the company’s nervous system. The measurement of financials is communicated back and forth through the company, yielding the general picture of a company’s well being, the equivalent of heart rate, pulse, and calories taken in and spent (Ambler, 2003). The incompetence of marketing to engage with senior management through common financial metrics has limited the influence of the discipline (Verhoef and Leeflang, 2010). In order to recapture the respect of the top management and to enhance the influence of marketing in the firm, the accountability of marketing activities and strategies becomes important (Doyle, 2000). To increase this accountability, marketing researchers started to develop a link between marketing and firm value, shifting away from the focus on the link between marketing and accounting-based relationships (Rust, Ambler and Carpenter, 2004). For a superior long-term performance, marketing is a necessity to both consumption and financial markets. However, shareholders can draw a firm's marketing focus toward the financial markets (Lovett and MacDonald, 2005).
Marketing researchers have suggested that the models for predicting firm performance lack intangible customer-based metrics (Gupta, Lehmann and Stuart, 2004). The financial value of a firm is determined by its intangible assets, such as customer satisfaction and customer loyalty (Srivastava, Shervani and Fahey, 1998). However, financial analysts cover these critical determinants only to a small extend, focusing on tangible data reported in a firm’s financial statement instead (Gupta et al., 2004).
An intangible asset found to influence firm value is customer satisfaction (Anderson, 2004; Fornell, Mitha, Morgenson III and Krishand, 2006; Gruca and Rego, 2005). Aksoy, Cooil, Groening, Keiningham, and Yalcin (2008) studied the link between customer satisfaction and firm value. They find that a portfolio of firms with high customer satisfaction generates an excess return of 0.78% a month. However, others studies reexamined the possible mispricing of customer satisfaction and found no excess returns related to customer satisfaction (Bell, Ledoit and Wolf, 2012; Jacobsen and Mizik, 2009; O’Sullivan, Hutchinson, and O’Connell, 2009). Since customer satisfaction information is publicly available, our study expects to find no excess returns. However, since customer satisfaction reduces the volatility of cash flows, our expectation is that customer satisfaction will lower systematic risk.
This paper continues the previous line of research on the relation between customer satisfaction and stock returns. For instance, Fornell et al. (2006) and O’Sullivan et al. (2009) studied the relationship by making use of the American Customer Satisfaction Index (ACSI) data. However, some questions remain, including, “Are firm’s excess returns related to its level of customer satisfaction?” and “Do these results hold over different time horizons?” In this study, the previous research in the field is extended by exploring whether economic downturn exerts any influence on the relationship between customer satisfaction and market value. Recessions can affect the performance of firms. However, not all firms perform poorly or fail during economic downturns (Srinivasan, 2005). Customer satisfaction influences customer retention and loyalty, which are important during economic downturns (Anderson and Sulivan, 1993). By using the approach of Henrikkson and Merton (1981), our study tests whether the alphas and the betas differ during up and down markets.
University of Groningen 3
Literature Review
High customer satisfaction ratings are widely assumed to be the best indicator of a company’s future profit. Firms increasingly use customer satisfaction as a criterion for diagnosing product or service performance and tie customer satisfaction ratings to both executive and employee compensation (Kottler, 1991). According Kottler (1991), satisfaction can be characterized as a post-purchase evaluation of product quality, given pre-purchase expectations. Oliver (2010) defines customer satisfaction as the customer’s fulfillment response. It is a judgment that a feature of a product or a service, or this product or service itself provides a pleasurable level of consumption-related fulfillment, including levels of under- or over-fulfillment. The drivers of customer satisfaction, perceived quality, customer expectations, and perceived value are incorporated in the American Customer Satisfaction Index (ACSI).
It may seem highly intuitive that firms should perform better if their customers are happier. A positive relationship between customer satisfaction and stock returns requires two channels, namely: (1) customer satisfaction is beneficial to firm value, and (2) the benefits of customer satisfaction are not fully valued by the market. The second channel is motivated by the non-incorporation of other intangibles (e.g., employee satisfaction).
Marketing researchers have recently put more emphasis on how various marketing activities and metrics affect the bottom-line financial performance of a firm (Williams and Naumann, 2011). These performances may include positive outcomes, such as increases in revenue or market share growth (Anderson and Mittal, 2000; Fornell et al., 2006). Terpstra, Kuijlen and Sijtsma (2012) empirically studied the influence of customer satisfaction on revenues. They tested the relationship between customer satisfaction at time t = 0 and revenues at time t > 0 and found that customer satisfaction had a positive impact on revenues with 1-year and 2-year time lags. The relation between customer satisfaction and market share growth is less clear. As Rego, Morgan and Fornell (2013) indicate, a firm’s customer satisfaction can predict future market share of the firm when it is benchmarked against that of its nearest rival, and when customer-switching costs are low. In general, it is hard to predict whether customer satisfaction is beneficial for firm value, since, besides affecting long-term revenues, it increases costs.
For a positive relationship between customer satisfaction and stock returns, the benefits should not be fully valued by the market. The efficient-market hypothesis asserts that financial markets are informationally efficient. Since customer satisfaction information is publicly available, it is expected that the market already incorporates the benefits of customer satisfaction.
However, one of the first studies exploring the link between customer satisfaction and stock returns expected to see a positive relationship (Ittner and Lacker, 1998). This study investigated the stock market reactions to the release of the ACSI data. The authors calculated cumulative abnormal returns for firms over five or ten days following the ACSI disclosure. Their findings indicated that the level of customer satisfaction of firms was economically relevant to the stock market, which was not completely reflected in the accounting book values, however. More recently, Anderson et al. (2004) studied the link between satisfaction and market value. This study developed a theoretical framework that specified how customer satisfaction affected future customer behavior. For this framework, the authors used the price-to-book value, equity prices, and Tobin’s Q using ACSI data over a time horizon from 1994 to 1997. Their results suggested a positive relation between customer satisfaction and shareholder value. Furthermore, Gruca and Rego (2005) linked customer satisfaction to shareholder value and found that satisfaction creates shareholder value by increasing future cash flow growth and reducing its variability. Fornell et al. (2006) found that customer satisfaction is significantly related to the market value of equity. However, news about the ACSI results do not move share prices. Their event study consisted of 161 events for 89 companies for which stock data were available, and where a one-day event period was specified. Their findings were mixed and inconclusive, but fairly consistent with the results of Ittner and Lacker (1998). Ittner and Larcker (1998) found no significant effects of ACSI news for the event window of five days. These results support our expectation that there is no relationship between customer satisfaction and stock returns. Similarly to Gruca and Rego (2005), Fornell et al. (2006) also found that firms that do better than their competitors in terms of customer satisfaction perform at lower systematic risk.
University of Groningen 5
2009). To reexamine the mispricing of customer satisfaction, Jacobsen and Mizik (2009) match the ACSI data with monthly and daily stock returns from the Center for Research in Security Prices (CRSP) database. The authors suggest that the mispricing of customer satisfaction is not widespread, but is limited to firms in the computer and internet sector. O’Sullivan et al. (2009) take the study by Fornell et al. (2006) as a guideline for their study to examine the mispricing of customer satisfaction. In line with Fornell et al. (2006), O’Sullivan et al. (2009) start by examining the performance of a portfolio based on the top 20% of ACSI. Therefore, in both studies the same models (CAPM, three-factor and four-factor) are used for analyses. Their results indicate a higher return for the top 20% portfolio compared to the lowest 80%. In addition, a positive alpha is found for the top 20% portfolio; however, this alpha is not significant for any of the estimated models. These findings are in line with the findings of Bell et al. (2012) and Jacobsen and Mizik (2009). To examine the mispricing, Bell et al. (2012) used a new portfolio formation approach. In this approach, they formed portfolios that do not load on any risk factor in the four-factor model (i.e. their aim was to be market-, size-, book-to-market-, and momentum-neutral). Under this approach, mispricing can be observed by looking at the expected return of the resulting portfolio (Bell et al., 2012). The studies of Bell et al. (2012), Jacobsen and Mizik (2009), and O’Sullivan et al. (2009) provide evidence that there is no mispricing of customer satisfaction.
As stated above, marketing researchers have put more emphasis on how various marketing and metrics affected the financial performance of a firm. However, their studies focused on the impact of customer satisfaction on financial performance in general circumstances. Recessions can affect the performance of firms; for some firms, recessions can be seen as opportunities to strengthen their business over their weaker competitors (Srinivasan et al., 2005). Recessions can affect the performance of individual firms, industries, and entire economic sectors (Zarnowitz, 1985). However, not all firms perform poorly or fail during economic downturns, some firms will even grow (Srinivasan, 2005). Despite the effects of recessions on the performance of firms, there is little knowledge about the appropriate response during challenging times. ACSI states that it is important to focus all on customers, even in times of recession. Barwise (1999) notes that firms that invest during economic downturns experience significant benefits afterwards. However, most firms dramatically decrease their marketing expenditures during recessions (Barwise and Styler, 2002). A key necessity for the maintenance of a brand’s life in the long term is to win customer loyalty (Aydin et al., 2005). Customer satisfaction can lead to positive behavior and directly enhance customer loyalty (Deng et al., 2010). Therefore, customer satisfaction influences customer retention and loyalty that are crucial during economic downturns (Anderson and Sulivan, 1993). This suggests that investments in customer satisfaction can reasonably be expected to enhance business performance during economic downturns.
In this study, we hypothesize that there is no relationship between customer satisfaction and stock returns, as the customer satisfaction information is fully valued by the market. Firms that do better than their competitors in terms of customer satisfaction have, due to their increased customer loyalty and retention, lower volatility of cash flows. Therefore, we hypothesize that firms that score better on customer satisfaction generate their returns with a lower systematic risk.
Research Methodology
Customer satisfaction was measured in this study using the American Customer Satisfaction Index (ACSI), developed by the National Quality Research Center (NQRC). This index is used to measure the quality of goods and services purchased in the United States that are produced by both domestic and foreign firms with substantial U.S. market shares. As ACSI collects the customer satisfaction data from over 50,000 customers through telephone interviews, it is a national barometer of customer satisfaction (Fornell et al., 1996). The drivers of customer satisfaction, such as perceived quality, customer expectations, and perceived value, are incorporated in the index. The overall scores are scaled from 0 to 100 for more than 375 company brands and have been released in the public domain since 1994. These company brands can be classified in 10 economic sectors and 43 industries. The ACSI scores of individual firms are updated once a year; however, the publication date depends on the economic sector. The National ACSI score has proven to be a strong predictor of Gross Domestic Product (GDP), Personal Consumption Expenditure (PCE), and the stock market (Fornell et al., 2010). Therefore, the index is becoming an important indicator of economic performance for the macro economy.
To test the relationship between customer satisfaction and stock returns, portfolios have been formed based on the ACSI scores. All NYSE- and NASDAQ-listed firms that are measured by the ACSI have been taken into account. This study covers the time period from 1996 to 2013, which is a period with market ups and downs. In total, 2,095 firm-yearly observations have been analyzed. The number of unique firms in each year averaged 116 firms with a range from 86 to 143. For firms where customer satisfaction was measured for multiple departments, a firm average has been calculated.
Portfolios have been formed by ranking the companies from high to low based on their ACSI score. To have a diversified portfolio of reasonable size, Fornell et al. (2006) selected firms in the top 20% of ACSI, relative to their competition. To have portfolios of reasonable size, this study also creates quintile portfolios. Due to the annual updates of ACSI, the inclusion of a stock within a portfolio has been re-examined on a yearly basis. Within this study, not only quintile portfolios, but also high minus low portfolios have been taken into account. For example, the high minus low 10% portfolio is the portfolio of the 10% firms with the highest ACSI scores minus the 10% firms with the lowest ACSI scores).
University of Groningen 7
depending on the re-examination of the updated ACSI score in 2013. The reason for using lagged data was to make sure that any investor forming a portfolio at a given time had full access to the publicly available ACSI data.
After constructing the portfolios, their abnormal returns have been analyzed. Due to the risk-return tradeoff (Ghysels et al., 2005), it was expected that riskier portfolios were accompanied by higher returns. Therefore, the returns have to be controlled for several validated risk factors. This was done using the four-factor model.
The CAPM model of Sharpe (1964) and Lintner (1965) posits a market reward for variation of returns. Fama and French (1992, 1993) extended the CAPM model with two systematic risk factors. The factors SMB (Small Minus Big) and HML (High Minus Low) are common factors in stock returns related to size and book-to-market equity. The Small Minus Big factor measures the additional return that the investors have received historically, by investing stocks of companies with a relatively small market capitalization compared to large market capitalizations. The High Minus Low factor measures the premium provided to investors for investing in companies with high book-to-market values. Carhart (1997) extends the Fama and French (1992, 1993) model with the momentum factor. Momentum in a stock described as the tendency of the stock price to continue previous changes. In other words, if the stock has a positive (negative) return in the previous period, it is expected to show a positive (negative) return in the next period as well. The Carhart model is specified in equation (1).
(1) RCShigh,t– RCSlow,t = αhl+ βhl(Rm,t – Rf,t) + βhl2(SMBt) + βhl3(HMLt) + βhl4(MOMt) + εhl,t
where RCShigh,t– RCSlow,t is the difference in rate of return on the portfolios with the highest and the lowest scores on customer satisfaction on day t; Rm,t is the market rate of return using the Standard & Poor’s (S&P) 500 composite index on day t; Rf,t is the risk free return using the 3-monthly U.S. treasury bills; αhl is the intercept that measures the impact of customer satisfaction. The subscript hl stands for high—low. The betas measure the systematic risk of the portfolio. Systematic risk is measured by four systematic risk factors: the market factor, the size factor, the book-to-market factor, and the momentum factor, respectively. The sign of alpha is used to determine whether the portfolios with high customer satisfaction scores minus those with low corresponding scores outperform in the market without taking additional risk.
For the sake of completeness, the three-factor model has been considered as well (see Appendix 2). The three-factor model is specified in equation (2).
(2) RCShigh,t – RCSlow,t = αhl+ βhl(Rm,t – Rf,t) + βhl2(SMBt) + βhl3(HMLt) + εhl,t
To test whether the results of the Carhart four-factor model hold, the sample period (April 1996 – September 2013) has been divided into four sub-periods. The first split divided the sample period in two periods of equal size, a period from April 1996 till the end of 2004 (2204 observations) and from the beginning of 2005 until September 2013. The second split divided the sample period in a pre- (i.e. before August 2007) and post- (i.e. after August 2007) start of the financial crisis period.
heteroskedasticity. Time series of financial assets (e.g. stock returns) often exhibit volatility clustering. In the presence of heteroskedasticity, the OLS is consistent but inefficient. In this study, the residuals have been tested for heteroskedasticity by an ARCH test, which indicated that variance of the error term was not constant. To relax the assumption of constant variance of the error term and to allow the conditional variance to depend on previous lags, a GARCH(1,1) model was estimated. The conditional variance of the GARCH(1,1) model is specified in equation (3).
(3) σ!! = ω + α!ε!!!! + β!σ!!!!
As indicated by Fornell et al. (2010), the National ACSI score has proven to be a strong predictor of Gross Domestic Product (GDP) and Personal Consumption Expenditure (PCE). As stated by Fornell et al. (2010), the ACSI index is becoming an important indicator of economic performance for macro economy. However, since there is no confirmation from macro economists on this claim, the predictive power of customer satisfaction has to be tested. To test whether customer satisfaction is a strong predictor of stock returns, our study has extended Carhart’s (1997) four-factor model with a factor that measures the change in the National ACSI scores. The relative change in the ACSI score has been calculated with the following equation (4):
(4) !"#$%&"' !"#$ !"#$%!! !"#$%&"' !"#$ !"#$%!!!
!"#$%&"' !"#$ !"#$%!!!
The extended model is specified in equation (5), where ∆CSt is the change in customer satisfaction at time t. Since the stock returns are measured with a lag, to ensure that any investor forming a portfolio at a given time has full access to the publicly available ACSI data, a lagged change factor has also been created. This factor is included in equation (6), where ∆CSlaggedt is the lagged change in customer satisfaction at time t.
(5) RCShigh,t– RCSlow,t = αhl+ βhl(Rm,t – Rf,t) + βhl2(SMBt) + βhl3(HMLt) + βhl4(MOMt) + βhl5(∆CSt)+ εhl,t
(6) RCShigh,t– RCSlow,t = αhl+ βhl(Rm,t – Rf,t) + βhl2(SMBt) + βhl3(HMLt) + βhl4(MOMt) + βhl5(∆CSlaggedt)+ εhl,t
University of Groningen 9
(7) RCShigh,t– RCSlow,t = d+α!!!+ d+β!!!(Rm,t – Rf,t) + d-α!!!+ d-β!!!(Rm,t – Rf,t)+ εhl,t
where d+and d- are the dummies for up- and down-market periods, respectively. The up-market dummy is a 0, 1 dummy, in which d = 1 for an up-market and 0 otherwise. An up market is defined as a period in which the market returns exceeds the risk-free returns (Rm,t>Rf,t). The inclusion of the dummy variables allows for testing the differences in portfolio betas across different market conditions. The down-market dummy is specified as (1 – d+). There is also a up- and down-market beta specification based on the Carhart model (see equation (8)).
(8) RCShigh,t – RCSlow,t = d+α!!!+ d+β!!! (Rm,t –Rf,t) + d+β!!!! (SMBt) + d+β!!!! (HMLt) + d+β!!!! (MOMt) d-α !! !+ d-β !! ! (R m,t –Rf,t) + d-β!!!! (SMBt) + d-β!!!! (HMLt) + d-β!!!! (MOMt) + εhl,t
A Wald test for model coefficients has been executed to test if the differences in up- and down market coefficients are significant. Thereby, we have tested whether the risk factors differ significantly during up- and down market.
Data and Descriptive Statistics
The American Customer Satisfaction Index website provides customer satisfaction data. The scores are reported on a 0-100 scale for about 375 brands. The first measurement was published in 1994, and new scores have been published every quarter from 1996 and on. This paper uses the ACSI data from 1995 until 2013. The association of customer satisfaction to stock returns relates to the brand-company connection, since satisfaction scores are measured on brand level, while stock returns are measured on company level. Ittner, Larcker and Taylor (2009) provide guidelines for the connections between brand and company levels. These guidelines are somewhat arbitrary, but they attempt however to minimize the problems that arise when the ACSI scores are linked to company identifiers. The guidelines are shown in Appendix 1. By connecting the brand-level ACSI scores to the company-level stock returns, the 375 company brands for which ACSI data were available for at least one year have been connected to 219 individual firms.
Quintile portfolios have been created by ranking the data from high to low based on their ACSI scores. The average number of the firms included in a portfolio is 116, with a minimum of 86 and a maximum of 143. The portfolios of the top and the lowest 20% consisted both of 23 firms on average. The minimum and the maximum number of the firms in these portfolios was 17 and 28, respectively. From 1995 to 2012, the customer satisfaction score was 75.6 on average. The lowest ACSI score over this time horizon was 49, whereas the highest score was 91.
by the website of Kennet French1. The descriptive statistics of these variables are shown in Table 1. The ACSI universe portfolio contains all firms (100%) that could be included in one of the quintile portfolios.
Table 1: Descriptive Statistics
Market Return Risk Free SMB HML MOM ACSI Universe (100%) Average daily return (%) 0.037 0.011 0.009 0.014 0.025 0.059 Daily Standard Deviation (%) 1.280 0.012 0.614 0.648 0.990 1.223 Skewness -‐ 0.039 1.812 -‐ 0.289 0.074 -‐ 0.868 -‐ 0.034 Kurtosis 10.488 6.303 7.004 8.527 12.353 11.386
N 4405 4405 4405 4405 4405 4405
Figures 1 and 2 show the absolute and the cumulative returns, of the total and the market portfolio, respectively. The total portfolio contains all stocks available for inclusion in an ACSI-based portfolio. As can be seen in Figure 2, the yearly return of the ACSI Universe portfolio (13.72% on average) is considerably higher than the market return (8.88% on average).
Figure 1: Yearly Returns of the ACSI Universe and the Market
Figure 2: Cumulative Yearly Returns of the ACSI Universe and the Market
1http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/ -‐60,00% -‐40,00% -‐20,00% 0,00% 20,00% 40,00% 60,00% 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
Market Return ACSI Universe Return
Results
We constructed portfolios by ranking the companies from high to low based on their ACSI scores. In Table 2, the returns and the standard deviations of specific selections from the ACSI universe are presented. In Table 3, we present the return and the standard deviation of the high minus low customer satisfaction portfolios (e.g. high – low 10% is the top 10% portfolio minus the worst 10% portfolio).
Table 2a shows that all ACSI-based portfolios have a positive average daily return. Most notable is the 0.067% daily average return of the 5th 20% portfolio. This average return is higher than the average daily returns of the other portfolios. Besides, the standard deviation of the 5th 20% portfolio is the lowest of all portfolios. Table 2b confirms that the 1st 20% is not performing the best. The top 20% portfolio has a cumulative return of 234% over the period of 17 years, whereas the 5th 20% portfolio has a cumulative return of 267%. However, the annualized standard deviation of the top 20% is the lowest (16.83%). Surprising is that all the ACSI based quintile portfolios outperform the market return. Table 3 also shows that the portfolios with higher ACSI scores not necessarily give higher returns. The average daily returns of the high minus low portfolios are zero or negative.
Table 2: Descriptive Statistics of the ACSI based portfolios
ACSI Universe (100%) Best 50% Worst 50% 1st 20% 2nd 20% 3rd
20% 4
th
20% 5th 20% Market A. Descriptive Statistics on Daily Basis
Daily Average Return (%) 0.059 0.056 0.061 0.056 0.063 0.051 0.056 0.067 0.037 Daily Standard Deviation (%) 1.223 1.194 1.308 1.150 1.319 1.329 1.363 0.089 1.280 Skewness -‐0.034 -‐0.061 -‐0.003 0.026 0.092 0.010 -‐0.081 0.089 -‐0.039 Kurtosis 11.386 11.919 10.167 9.125 12.169 11.250 8.639 11.494 10.488 Jarque-‐Bera 12909 14605 9427 6887 15436 12492 5851 13248 10293 N 4405 4405 4405 4405 4405 4405 4405 4405 4405
B. Annualized Descriptive statistics
Cumulative Returns (%) 233.26 228.19 238.33 233.70 263.21 193.38 209.44 267.42 150.99 Annualized Mean Return (%) 13.72 13.42 14.02 13.75 15.48 11.38 12.32 15.73 8.88 Annualized Standard Deviation (%) 18.16 18.49 18.09 16.83 20.97 19.12 20.95 18.34 18.79 N 17 17 17 17 17 17 17 17 17
Table 3: Descriptive Statistics of the ACSI based Hi-‐Low portfolios
Hi-‐Low 10% Hi-‐Low 20% Hi-‐Low 30% Hi-‐Low 40% Hi-‐Low 50% Daily Average Return (%) -‐ 0.001 -‐ 0.011 0.000 -‐0.003 -‐ 0.006 Daily Standard Deviation 1.256 0.918 2.262 1.281 0.539
Skewness 0.161 0.144 0.205 0.110 0.007
Kurtosis 9.536 9.857 8.680 8.584 7.608
Jarque-‐Bera 7861 8646 5952 5730 3897
N 4405 4405 4405 4405 4405
.
Table 4a: Four-‐Factor Model Results
Alpha Market Beta SMB Beta HML Beta MOM Beta Adj. R-‐square F-‐value ARCH-‐LM ACSI Universe (100%) (3.76) 0.021 *** (191.39) 0.885 *** (13.35) 0.122 *** (39.21) 0.357 *** (-‐16.00) -‐ 0.098 *** 0.909 11049.37 *** 81.64 *** Best 50% (2.68) 0.018 *** (146.68) 0.841 *** (11.43) 0.129 *** (33.45) 0.377 *** (-‐10.01) -‐ 0.076 *** 0.854 6442.18 *** 104.75 *** Worst 50% (3.47) 0.023 *** (165.98) 0.930 *** (10.36) 0.115 *** (30.53) 0.336 *** (-‐16.19) -‐ 0.121 *** 0.884 8375.36 *** 99.28 *** 1st 20% 0.021 (2.24) ** (93.78) 0.748 *** (1.44) 0.023 (19.56) 0.307 *** -‐ 0.007 (-‐0.06) 0.696 2516.25 *** 41.41 *** 2nd 20% 0.025 (2.75) *** (116.68) 0.879 *** (11.31) 0.169 *** (29.79) 0.442 *** (-‐12.04) -‐ 0.121 *** 0.793 4227.50 *** 97.78 *** 3rd 20% 0.012 (1.36) (130.98) 0.921 *** (13.70) 0.190 *** (27.18) 0.376 *** -‐ 0.085 (-‐9.07) *** 0.823 5112.71 *** 103.16 *** 4th 20% 0.019 (2.09) ** (120.76) 0.924 *** (8.37) 0.127 *** (15.28) 0.230 *** (-‐12.41) -‐ 0.126 *** 0.800 4415.94 *** 151.95 *** 5th 20% 0.028 (2.77) *** (114.31) 0.954 *** (6.52) 0.107 *** (25.65) 0.421 *** (-‐13.80) -‐ 0.153 *** 0.788 4092.25 *** 100.26 ***
Table 4b: Four-‐Factor Model Results
Alpha Market Beta SMB Beta HML Beta MOM Beta Adj. R-‐square F-‐value ARCH-‐LM Hi – Low 10% (0.17) 0.003 (-‐23.53) -‐ 0.334 *** -‐ 0.196 (-‐6.98) *** -‐ 0.275 (-‐9.85) *** (10.57) 0.199 *** 0.192 262.78 *** 44.57 ***
Hi – Low 20% -‐ 0.006 (-‐0.50) (-‐19.24) -‐ 0.206 *** -‐ 0.085 (-‐4.01) *** -‐ 0.114 (-‐5.42) *** (10.26) 0.146 *** 0.140 179.91 *** 51.88 ***
Hi – Low 30% (0.13) 0.004 (-‐14.99) -‐ 0.408 *** -‐ 0.129 (-‐2.39) ** (1.25) 0.067 (7.50) 0.272 *** 0.082 99.86 *** 53.89 ***
Hi – Low 40% -‐ 0.001 (-‐0.04) (-‐16.36) -‐ 0.251 *** -‐ 0.043 (-‐1.42) (3.24) 0.097 *** (7.45) 0.152 *** 0.094 115.59 *** 75.70 ***
Hi – Low 50% -‐ 0.005 (-‐0.62) (-‐13.60) -‐ 0.089 *** (1.14) 0.015 (3.16) 0.041 *** (5.08) 0.044 *** 0.064 76.02 *** 84.92 ***
As the results in Table 4a indicate, abnormal returns can be achieved by investing in firms that are in the ACSI universe (i.e. except from the 3rd 20% portfolio, all other portfolios have a positive and significant Jensen's alpha). Moreover, the betas are significant for all risk factors, with a positive sign for the market factor, the size factor, and the book-to-market factor, and a negative sign for the momentum factor. In Table 4b, no significant alphas are shown. Portfolios of firms with higher ACSI scores not necessarily give higher returns compared to portfolios with low ACSI-scores. However, the betas of the high minus low portfolios give significant results. For the high minus low 20% portfolio, the market beta, SMB beta, HML beta and MOM beta are -0.206 (p<0.01), -0.085 (p<0.01), -0.114 (p<0.01) and 0.146 (p<0.01), respectively. The negative sign for market, size, and market-to-book beta suggest that the systematic risk for these risk factors are lower for portfolios with higher ACSI scores compared to firms with lower ACSI scores. The momentum beta, however, has a positive sign, which suggests that better ACSI performing portfolios have more risk for the momentum investing than lower ACSI performing portfolios. The adjusted R-squares of the estimated models in Table 4a are all between 0.666 and 0.859, indicating that the included variables explain a substantial portion of variance in the dependent variable.
four-
University of Groningen 13
factor model are shown in Tables 5a-b. The results for the GARCH(1,1) variant of the three-factor model are shown in Appendix 3. All further models within this study have been estimated using the GARCH(1,1) approach.
Table 5a: Four-‐Factor Model Results (GARCH)
Alpha MarketBeta SMB Beta HML Beta MOM Beta Adj. R-‐square AIC ACSI Universe (100%) (3.88) 0.017 *** (243.80) 0.901 *** (17.03) 0.136 *** (34.33) 0.287 *** (-‐13.45) -‐ 0.071 *** 0.907 0.561 Best 50% (3.29) 0.017 *** (181.76) 0.852 *** (13.11) 0.125 *** (26.85) 0.259 *** -‐ 0.044 (-‐6.62) *** 0.848 0.974 Worst 50% (3.29) 0.019 *** (0.00) 0.948 *** (14.11) 0.140 *** (26.31) 0.284 *** (-‐15.04) -‐ 0.102 *** 0.882 1.035 1st 20% 0.014 (1.86) * (121.30) 0.799 *** (5.88) 0.080 *** (12.58) 0.145 *** (4.53) 0.037 0.679 1.630 2nd 20% 0.020 (3.05) *** (151.30) 0.883 *** (13.75) 0.165 *** (26.14) 0.324 *** (-‐14.40) -‐ 0.122 *** 0.790 1.516 3rd 20% 0.016 (2.32) ** (154.89) 0.918 *** (14.42) 0.180 *** (26.58) 0.310 *** -‐ 0.044 (-‐5.13) *** 0.820 1.448 4th 20% 0.015 (2.07) ** (143.68) 0.954 *** (11.24) 0.151 *** (15.87) 0.214 *** (-‐13.27) -‐ 0.120 *** 0.799 1.616 5th 20% 0.026 (3.21) *** (148.65) 0.955 *** (6.82) 0.095 *** (25.81) 0.377 *** (-‐13.25) -‐ 0.132 *** 0.787 1.812
Table 5b: Four-‐Factor Model Results (GARCH)
Alpha MarketBeta SMB Beta HML Beta MOM Beta Adj. R-‐square AIC Hi – Low 10% (0.32) 0.004 (-‐27.43) -‐ 0.315 *** -‐ 0.170 (-‐7.03) *** (-‐14.13) -‐ 0.349 *** (9.77) 0.176 *** 0.190 2.875
Hi – Low 20% -‐ 0.006 (-‐0.56) (-‐18.93) -‐ 0.163 *** -‐ 0.067 (-‐3.63) *** -‐ 0.174 (-‐9.33) *** (10.75) 0.145 *** 0.145 2.339
Hi – Low 30% (0.12) 0.003 (-‐15.12) -‐ 0.340 *** -‐ 0.151 (-‐3.32) *** -‐ 0.139 (-‐3.01) *** (7.83) 0.262 *** 0.077 4.183
Hi – Low 40% (0.28) 0.004 (-‐17.15) -‐ 0.235 *** -‐ 0.065 (-‐2.55) ** -‐ 0.051 (-‐1.96) * (7.42) 0.143 *** 0.088 3.036
Hi – Low 50% -‐ 0.001 (-‐0.11) (-‐15.68) -‐ 0.092 *** -‐ 0.009 (-‐0.82) -‐ 0.008 (-‐0.76) (5.93) 0.048 *** 0.060 1.342
The findings of the GARCH(1,1) variant of the four-factor model are in line with the results of the four-factor model based on OLS. As can been seen in Table 5a, all portfolios have a positive and significant alpha. Even the alpha of 3rd 20% portfolio is significant (p<0.05). The signs and the levels of significance for the risk factors are in unison with the four-factor model based on OLS (Tables 4a-b). The results for the high minus low portfolios are shown in Table 5b. Overall, the market, size, and market-to-book beta are significantly negative. The momentum beta, however, has a positive sign, which indicates that better ACSI performing portfolios have more risk for momentum investing as compared to lower ACSI performing portfolios. The results for the alphas are not significant and give no clear indication for the sign (i.e. the alpha is positive for the high minus low 10% portfolio but negative for the high minus low 20% portfolio).
Table 6a: Carhart Results for Sample Split I
Alpha Market Beta SMB Beta HML Beta MOM Beta Period A Period B Period A Period B Period A Period B Period A Period B Period A Period B Hi – Low 10% (1.45) 0.030 -‐ 0.009 (-‐0.51) (-‐22.21) -‐ 0.454 *** (-‐14.35) -‐ 0.211 *** -‐ 0.282 (-‐7.36) *** -‐ 0.187 (-‐5.32) *** (-‐11.23) -‐ 0.429 *** (-‐12.72) -‐ 0.549 *** (8.42) 0.215 *** (5.92) 0.161 ***
Hi – Low 20% (0.55) 0.010 -‐ 0.008 (-‐0.63) (-‐21.17) -‐0.350 *** (-‐10.11) -‐ 0.097 *** (-‐11.66) -‐0.354 *** (1.97) 0.048 ** -‐ 0.294 (-‐9.54) *** (-‐13.55) -‐ 0.390 *** (9.31) 0.191 *** (4.84) 0.088 ***
Hi – Low 30% (1.73) 0.077 * -‐ 0.028 (-‐0.84) (-‐20.88) -‐ 0.846 *** -‐ 0.150 (-‐5.52) *** (-‐12.24) -‐ 0.903 *** (2.38) 0.141 ** -‐ 0.525 (-‐6.66) *** -‐ 0.649 (-‐9.15) *** (8.74) 0.449 *** (2.66) 0.119 ***
Hi – Low 40% (1.24) 0.032 -‐ 0.009 (-‐0.48) (-‐23.00) -‐ 0.532 *** -‐ 0.080 (-‐4.66) *** (-‐10.08) -‐ 0.432 *** (0.78) 0.025 -‐ 0.200 (-‐4.50) *** (-‐10.92) -‐ 0.447 *** (8.25) 0.245 *** (2.43) 0.061 **
Hi – Low 50% (0.72) 0.008 -‐ 0.005 (-‐0.61) (-‐18.89) -‐ 0.199 *** -‐ 0.026 (-‐3.36) *** -‐ 0.128 (-‐6.78) *** (0.36) 0.005 -‐ 0.055 (-‐2.85) *** -‐ 0.171 (-‐9.49) *** (7.32) 0.093 *** (1.22) 0.013
Period A is the sample period from 4/01/1996 – 12/31/2004. Period B is the sample period from 1/03/2005 –9/30/2013. Table 6b: Carhart Results for Sample Split II
Alpha Market Beta SMB Beta HML Beta MOM Beta Period C Period D Period C Period D Period C Period D Period C Period D Period C Period D Hi – Low 10% (0.37) 0.006 (0.03) 0.000 (-‐21.35) -‐ 0.379 *** (-‐11.66) -‐ 0.188 *** -‐ 0.213 (-‐6.64) *** -‐ 0.137 (-‐3.22) *** (-‐10.12) -‐ 0.342 *** (-‐12.63) -‐ 0.664 *** (8.74) 0.203 *** (4.21) 0.138 ***
Hi – Low 20% (0.01) 0.000 -‐ 0.018 (-‐1.05) (-‐19.69) -‐ 0.282 *** -‐ 0.064 (-‐5.86) *** -‐ 0.235 (-‐9.58) *** (3.68) 0.108 *** -‐ 0.254 (-‐9.54) *** (-‐11.84) -‐ 0.397 *** (10.21) 0.184 *** (5.95) 0.125 ***
Hi – Low 30% (1.50) 0.052 -‐ 0.047 (-‐1.13) (-‐20.75) -‐ 0.718 *** -‐ 0.055 (-‐1.80) * (-‐11.09) -‐ 0.686 *** (4.93) 0.325 *** -‐ 0.468 (-‐6.72) *** -‐ 0.614 (-‐7.71) *** (8.46) 0.388 *** (4.78) 0.248 ***
Hi – Low 40% (1.11) 0.022 -‐ 0.023 (-‐0.99) (-‐22.03) -‐ 0.431 *** -‐ 0.043 (-‐2.22) ** -‐ 0.293 (-‐8.16) *** (2.71) 0.100 *** -‐ 0.133 (-‐3.46) *** (-‐10.40) -‐ 0.483 *** (8.35) 0.219 *** (3.08) 0.089 ***
Hi – Low 50% (0.76) 0.007 -‐ 0.010 (-‐0.96) (-‐19.69) -‐ 0.171 *** -‐ 0.007 (-‐0.76) -‐ 0.086 (-‐5.57) *** (1.95) 0.032 * -‐ 0.045 (-‐2.72) *** -‐ 0.190 (-‐9.38) *** (7.32) 0.083 *** (1.83) 0.022 *
Period C is the sample period from 4/01/1996 – 7/30/2007. Period D is the sample period from 8/01/2007–9/30/2013.
As can be seen in Tables 6a-b, there is no change in significance and sign of the risk factors from period A to period B (Table 6a), and from period C to period D (Table 6b). The alpha is only significant for the high minus low 30% portfolio over a time horizon of 4/01/1996 – 12/31/2004 (period A). However, the sign of alpha changes from positive to negative in both splits. As shown in Table 6a, the alpha for the high minus low 20% portfolio changes from 0.010 in period A to -0.008 in period B. In Table 6b, the alpha for the high minus low 20% portfolio changes from 0.000 in period C to -0.018 in period D. It should be noted that there is no change in the significance of the alpha.
Table 7a: Results Carhart Extended with Change in CS Variable
Alpha MarketBeta SMB Beta HML Beta MOM Beta ΔCS Adj. R-‐square AIC ACSI Universe (100%) (3.88) 0.017 *** (242.70) 0.901 *** (16.99) 0.136 *** (34.32) 0.287 *** (-‐13.45) -‐ 0.071 *** (0.00) 0.000 0.907 0.562 Best 50% (3.27) 0.017 *** (181.71) 0.852 *** (13.12) 0.125 *** (26.85) 0.259 *** -‐ 0.044 (-‐6.60) *** (0.28) 0.003 0.848 0.974 Worst 50% (3.27) 0.019 *** (196.58) 0.948 *** (14.10) 0.140 *** (26.16) 0.284 *** (-‐15.01) -‐ 0.102 *** (0.59) 0.006 0.882 1.035 1st 20% 0.014 (1.86) * (121.31) 0.799 *** (5.86) 0.080 *** (12.57) 0.145 *** (4.53) 0.037 *** (0.29) 0.004 0.679 1.630 2nd 20% 0.021 (3.08) *** (151.20) 0.883 *** (13.74) 0.165 *** (26.13) 0.323 *** (-‐14.41) -‐ 0.122 *** -‐ 0.004 (-‐0.36) 0.790 1.517 3rd 20% 0.015 (2.26) (154.74) 0.918 *** (14.37) 0.180 *** (26.58) 0.310 *** -‐ 0.044 (-‐5.11) *** (1.01) 0.012 0.820 1.448 4th 20% 0.016 (2.08) ** (143.03) 0.954 *** (11.25) 0.151 *** (15.87) 0.214 *** (-‐13.26) -‐ 0.120 *** -‐ 0.002 (-‐0.12) 0.799 1.617 5th 20% 0.026 (3.20) *** (148.52) 0.955 *** (6.82) 0.095 *** (25.78) 0.377 *** (-‐13.24) -‐ 0.132 *** (0.24) 0.004 0.787 1.813
Table 7b: Results Carhart Extended with Change in CS Variable
Alpha MarketBeta SMB Beta HML Beta MOM Beta ΔCS Adj. R-‐square AIC Hi – Low 10% (0.33) 0.004 (-‐27.45) -‐ 0.315 *** -‐ 0.170 (-‐7.04) *** (-‐14.14) -‐ 0.349 *** (9.77) 0.176 *** (1.00) 0.024 0.190 2.875
Hi – Low 20% -‐ 0.006 (-‐0.55) (-‐18.93) -‐ 0.163 *** -‐ 0.067 (-‐3.64) *** -‐ 0.174 (-‐9.32) *** (10.73) 0.145 *** (0.36) 0.007 0.133 2.339
Hi – Low 30% (0.16) 0.004 (-‐15.13) -‐ 0.340 *** -‐ 0.150 (-‐3.30) *** -‐ 0.138 (-‐2.99) *** (7.84) 0.262 *** -‐ 0.028 (-‐0.56) 0.077 4.184
Hi – Low 40% 0.004 (0.25) (-‐17.13) -‐ 0.234 *** -‐ 0.065 (-‐2.57) ** -‐ 0.052 (-‐1.97) ** (7.42) 0.143 *** (0.46) 0.013 0.088 3.036
Hi – Low 50% -‐ 0.001 (-‐0.12) (-‐15.68) -‐ 0.092 *** -‐ 0.009 (-‐0.82) -‐ 0.009 (-‐0.76) (5.92) 0.048 *** (0.139) 0.002 0.059 1.342
As can be seen in Tables 7a-b, the beta of the change in customer satisfaction (ΔCS) gives no significant results for any of the tested portfolios. Therefore, change in customer satisfaction has no strong predictive power for stock returns. The four-factor model extended with the lagged customer satisfaction variable (ΔCS lagged) gives comparable results (see Appendix 4).
The second part of this study focused on whether economic situation has exerted any influence on the market value of customer satisfaction. The approach of Henrikkson and Merton (1981) has been followed to specify the method for testing the influence of economic situation. Within this method, the alphas and betas are specified for the up-market, as well as for the down-market days. The results for up and down market following equation (7) are provided in Table 8.
Table 8: Up-‐ and Down-‐market Beta (CAPM), GARCH
Alpha+ Market Beta+ Alpha-‐ Market Beta-‐ AIC
Hi – Low 10% (0.30) 0.007 (-‐13.15) -‐ 0.271 *** -‐ 0.058 (-‐2.34) ** (-‐18.26) -‐ 0.338 *** 2.930
Hi – Low 20% -‐ 0.009 (-‐0.44) -‐ 0.135 (-‐9.27) *** -‐ 0.043 (-‐2.27) ** (-‐13.67) -‐ 0.198 *** 2.376
Hi – Low 30% -‐ 0.071 (-‐1.42) -‐ 0.243 (-‐6.07) *** -‐ 0.028 (-‐0.57) (-‐11.10) -‐ 0.420 *** 4.195
Hi – Low 40% -‐ 0.016 (-‐0.57) -‐ 0.200 (-‐8.82) *** -‐ 0.042 (-‐1.50) (-‐13.23) -‐ 0.303 *** 3.046
Table 8 shows that in a down-market situation, the alpha is significant for the high minus low 10% and 20%. The down-market alpha of the high minus low 10% portfolio is -0.058 and is significant (t-value is -2.34 and p<0.05). This may suggest that portfolios with the lower ACSI performing firms outperform the best ACSI performing firms in a down-market situation. However, these results do not hold for the model including the size, the market-to-book ratio, and the momentum factors, as specified in equation (8). These results are shown in Table 9.
Table 9: Up-‐ and Down-‐market Beta Model (Carhart), GARCH
Alpha+ Market Beta+ BetaSMB + BetaHML + MOM Beta+ Alpha-‐ Market Beta-‐ BetaSMB -‐ BetaHML -‐ MOM Beta-‐ AIC
Hi – Low 10% (0.23) 0.005 (-‐15.60) -‐ 0.306 *** -‐ 0.165 (-‐4.99) *** -‐ 0.319 (-‐9.21) *** (8.57) 0.198 *** -‐ 0.023 (-‐0.91) (-‐16.84) -‐ 0.343 *** -‐ 0.175 (-‐4.70) *** (-‐10.35) -‐ 0.373 *** (5.18) 0.149 *** 2.876 Hi – Low 20% -‐ 0.021 (-‐1.11) -‐ 0.136 (-‐9.19) *** -‐ 0.077 (-‐3.00) *** -‐ 0.143 (-‐5.56) *** (9.66) 0.168 *** -‐ 0.031 (-‐1.56) (-‐12.45) -‐ 0.197 *** -‐ 0.051 (-‐1.98) ** -‐ 0.198 (-‐7.86) *** (5.85) 0.122 *** 2.340 Hi – Low 30% -‐ 0.097 (-‐1.95) * -‐ 0.215 (-‐5.06) *** -‐ 0.195 (-‐3.07) *** -‐ 0.060 (-‐0.88) (7.43) 0.331 *** -‐ 0.018 (-‐0.36) (-‐10.45) -‐ 0.408 *** -‐ 0.070 (-‐1.10) -‐ 0.200 (-‐3.11) *** (3.98) 0.198 *** 2.262 Hi – Low 40% -‐ 0.035 (-‐1.22) -‐ 0.174 (-‐6.76) *** -‐ 0.091 (-‐2.70) *** -‐ 0.011 (-‐0.29) (7.07) 0.179 *** -‐ 0.038 (-‐1.34) (-‐12.99) -‐ 0.296 *** -‐ 0.021 (-‐0.55) -‐ 0.082 (-‐2.23) ** (3.71) 0.108 *** 3.035 Hi – Low 50% -‐ 0.018 (-‐1.65) -‐ 0.067 (-‐5.51) *** -‐ 0.018 (0.17) (3.52) 0.002 (5.36) 0.068 *** -‐ 0.014 (0.71) -‐ 0.113 (-‐6.99) *** (1.83) 0.008 * -‐ 0.014 (1.01) (1.90) 0.028 ** 1.342
Table 9 indicates that the alphas for the most high minus low portfolios are insignificant and negative. Only the up-market alpha for the high minus low 10% portfolio is positive (0.005). To test whether the various risk factors differ during up- and down-market periods, an ANOVA F-test has been performed. The results are shown in Table 10.
Table 10: ANOVA F-‐test
Market -‐ Rf SMB HML MOM
Up Down Up Down Up Down Up Down Mean 0.849 -‐ 0.906 0.018 -‐ 0.002 -‐ 0.067 0.106 -‐ 0.076 0.139 Standard Deviation 0.909 0.961 0.620 0.607 0.627 0.660 1.005 0.961 Observations 2341 2064 2341 2064 2341 2064 2341 2064 F-‐test 0.008 *** 0.326 0.019 ** 0.040 **
Table 11: Wald Test for Model Coefficients (GARCH)
Alpha Market Beta SMB Beta HML Beta MOM Beta Hi – Low 10% 0.687 1.730 0.039 1.096 1.822 Hi – Low 20% 0.131 8.005 *** 0.554 1.630 3.061 * Hi – Low 30% 1.245 11.720 *** 1.986 2.308 4.090 ** Hi – Low 40% 0.005 13.714 *** 2.069 1.960 3.471 * Hi – Low 50% 0.053 9.658 *** 1.424 0.581 6.265 **
As Table 11 indicates, the coefficients for the market and the momentum factors differ significantly between up- and down-market times. The coefficients for the alpha, SMB factor, and HML factor do not differ significantly. The significance of the market and momentum beta justify the use of the up- and down-market beta specification of the Carhart model.
Conclusion and Discussion
This paper examines the relationship between customer satisfaction and stock returns. Portfolios have been formed based on the customer satisfaction scores of firms. The four-factor model of Carhart (1997) has been used to examine the abnormal returns of the portfolios. To test whether the economic situations have an impact on the relationship between customer satisfaction and stock returns, the methodology of Henrikkson and Merton (1981) has been followed.
Investment in ACSI-based portfolios results in significant returns. All quintile portfolios generate positive excess returns. The alphas for all quintile portfolios are significantly positive. Results hold a GARCH(1,1) specification of the models. Contradicting Fornell (2006), there is no clear indication that the top 20% outperform the other portfolios. The results are in line with the results of O’Sullivan et al. (2009). Their evidence is less favorable for the portfolios formed on the level of customer satisfaction. The high minus low portfolios also indicate that the top customer satisfaction portfolios do not outperform the other portfolios.
The sample splits confirm the findings of the general four-factor model, namely, in that customer satisfaction does not affect the alpha, but lowers systematic risk. The insignificance of the alphas suggests market efficiency, since customer satisfaction information is publicly available. Our conclusion is that customer satisfaction lowers systematic risk.
This study rejects the results of Fornell et al. (2010) who states that the ACSI index has proven to be a strong predictor and an important indicator of economic performance for the macro economy. Our study has examined whether the change in the ACSI scores have strong predictive power for stock returns. As indicated by the levels of significance, as well the change in customer satisfaction, the lagged change in customer satisfaction have no strong predictive power. Moreover, the signs of the betas for the (lagged) change in customer satisfaction are not consistent for the different portfolios. This allows us to conclude that the (lagged) change in customer satisfaction has no strong predictive power for stock returns.
down market situations. This is in line with the Carhart (1997) model following the GARCH(1,1) specification.
For further research, the inclusion of customer satisfaction information of various countries would be an interesting consideration. All studies focusing on the relationship between customer satisfaction and stock prices used the freely available customer satisfaction data provided by ACSI. Further research can involve the NCSI-UK, which is based on customers from the United Kingdom. Thereby, it would be interesting to identify ACSI’s selection criteria for a firm’s inclusion in their database. Since all ACSI quintile portfolios outweigh the market return, their selection criteria could be an important factor.
A limitation of this research is its reliance on the publicly available ACSI data. These data tend to include only larger firms. Therefore, it would be interesting to measure customer satisfaction by using other approaches in future studies.
Altogether, customer satisfaction causes no excess returns but lowers systematic risk. These findings hold over different time intervals, as well as over periods of up and down market. In conclusion, there is no clear indication for the relation between customer satisfaction and stock returns, and the mispricing of customer satisfaction.