1 of 30

The Financial Markets and Customer Satisfaction: �Re-examining the Value Implications �of Customer Satisfaction �From the Efficient Markets Perspective� � April 20, 2007

Robert Jacobson Natalie Mizik

University of Washington Columbia University

Bob Jacobson and Natalie Mizik �Customer Satisfaction

1

2 of 30

Linking Marketing Metrics �to Financial Valuation

MSI call for research on:

“Marketing Meets Wall Street”

Understanding which marketing metrics have value implications can help marketers determine how best to develop and implement strategies that positively impact firm performance

Bob Jacobson and Natalie Mizik �Customer Satisfaction

2

3 of 30

Linking Marketing Metrics to Financial Valuation: �New Research Domain for Marketing

Marketing does not have a long history studying the relationship between marketing metrics and financial market performance
A stream in the accounting research is providing a flawed path (e.g., ignoring autocorrelation in market value series and the role of risk)
Pervasive issues (because, perhaps, marketers are not comfortable with efficient markets – consumers certainly are not rational)

key implications of financial market theories have not been integrated into the analyses
central variables have not been adequately measured
analyses with improper statistical specifications have been undertaken

Bob Jacobson and Natalie Mizik �Customer Satisfaction

3

4 of 30

Study Objectives:

Provide a framework that can serve as a starting point for the analysis of financial implications of marketing metrics.

Provide a demonstration: assessing the financial market response to the American Customer Satisfaction Index (ACSI)

ACSI is one of the most extensively studied marketing metrics in the context of financial market outcomes

Bob Jacobson and Natalie Mizik �Customer Satisfaction

4

5 of 30

Linking Customer Satisfaction (CS) to Financial Market Valuation

Prior research has reported:

Conflicting findings with respect to the current-term financial market reaction to CS

Consistent findings (which are contradictory to the efficient markets) of CS associated with future-period stock returns

MSI Research Priority Topics: “Why are movements in customer satisfaction not immediately reflected in stock prices, even though a long-run relationship between the two exists?” (April 4, 2007)

Bob Jacobson and Natalie Mizik �Customer Satisfaction

5

6 of 30

Empirical Demonstration: �Re-examining the Value of CS

What is the basic relationship between CS and market valuation?

Does CS have incremental information content to accounting variables in explaining stock return?

Is there a market anomaly associated with CS?

Bob Jacobson and Natalie Mizik �Customer Satisfaction

6

7 of 30

Our results differ from those reported in previous research primarily

because:

We account for autocorrelation in the market value series
We model unanticipated components of CS
We allow for the fact that financial market participants also have access to accounting information (e.g., earnings)
We utilize risk-adjusted stock return
We account for context- and sector-specific factors

Bob Jacobson and Natalie Mizik �Customer Satisfaction

7

8 of 30

Efficient Markets as the Theoretical Foundation of Valuation Research

the theory of the efficient markets is just the theory of competitive equilibrium applied to asset markets (Leroy 1989)
the market price of a security is the equilibrium belief in the market about its intrinsic value.

Bob Jacobson and Natalie Mizik �Customer Satisfaction

[2]

[1]

(1)

(2)

8

9 of 30

CAPM

Fama and French (1992, 1993)

Bench-mark approaches

(e.g., Barber and Lyon 1997)

Bob Jacobson and Natalie Mizik �Customer Satisfaction

Models of Expected Return ( )

9

Fama (1970) noted that any analysis of abnormal (i.e., risk-adjusted) returns depends on the model for expected (i.e., normal) returns. The problem is that all models of expected returns are incomplete descriptions of returns for any sample period. That is, any asset pricing model is just an approximation and does not completely describe expected returns. Further, even if there were a true model, any sample period produces systematic deviations from the model’s predictions, that is, sample-specific patterns of average returns are due to chance. As a result, analyses of risk-adjusted returns are always contaminated by “bad-model” issues in estimating expected returns.

The problem is less serious in event studies that focus on short-run returns since daily expected returns are close to zero and thus will have relatively little effect on estimates of unexpected (abnormal) returns. But the problem grows with the return horizon, e.g., when abnormal returns are assessed over a number of months or years. Any deficiency in the expected return model that generates an erroneous abnormal return, however small, becomes compounded and exaggerated over a longer-term horizon.

10 of 30

Issues with Modeling Expected Return

Fama on “bad-model” issues in estimating expected returns:

analysis of abnormal returns depends on the model for expected returns.
all models of expected returns are incomplete descriptions of returns for any sample period
any sample period produces systematic deviations from the model’s predictions
The problem grows with the return horizon – i.e., any deficiency in the expected return model that generates an erroneous abnormal return, however small, becomes compounded and exaggerated over a longer-term

No agreement exists as to which method is best to compute and time-aggregate (BHAR, CAR, CumAR) the expected and abnormal returns

Bob Jacobson and Natalie Mizik �Customer Satisfaction

10

11 of 30

Explaining Changes in the Expectations of Future Performance ( )

Accounting metrics
Marketing metrics

Bob Jacobson and Natalie Mizik �Customer Satisfaction

11

12 of 30

The Dynamic Performance Impact of Marketing Assets:�Estimation Framework

Bob Jacobson and Natalie Mizik �Customer Satisfaction

12

13 of 30

Assessing Potential Market Anomalies

Empirical tests of capital market efficiency typically assess whether the stochastic process characterizing stock prices has the properties of a martingale.

A series is said to follow a martingale process if

The most direct empirical tests of the martingale model attempt to determine whether some variable in an agents’ information set is predictive of future returns.

Bob Jacobson and Natalie Mizik �Customer Satisfaction

13

The methods discussed so far rely on the efficient market framework. There is, however, mounting (albeit far from universally accepted) evidence in the literature inconsistent with the efficient markets hypothesis. Empirical tests of capital market efficiency typically assess whether the stochastic process characterizing stock prices has the properties of a martingale (a broader class of models that encompasses the random walk). A series is said to follow a martingale process if E(Pt+1| Wt)= Pt. The most direct empirical tests of the martingale model attempt to determine whether some variable in an agents’ information set is predictive of future returns.

Findings of inefficiency are always challenged and it is common for them to be explained by alternative risk considerations. Fama (1998, p. 288) contends that most anomalies “tend to disappear when reasonable alternative approaches are used to measure them.” He continues (p. 299), “All methods for estimating abnormal returns are subject to bad-model problems, and no method is likely to minimize bad-model problems for all classes of events.” As such, this again highlights the necessity of validating any findings suggesting a potential market inefficiency with multiple measures of abnormal returns.

14 of 30

Previous Research

Market Valuation (contemporaneous response)

Ittner and Larcker (1998), Fornell, Mithas, Morgeson, Krishnan (2006)

link level of CS to market value and find sig results
do not address the issue of autocorrelation in the market value series
do not include income statement information (i.e., earnings)

Mittal, Anderson, Sayrak, and Tadikamalla (2005)

link stock return and the level of CS
do not use unanticipated components of CS in the stock return model estimation
find different results with Tobin’s Q, consistent with measurement bias in the independent variable

Anderson, Fornell, and Mazvancheryl (2004)

find of sig correlation between change is CS and Tobin’s Q growth – this suggests a within-period financial market adjustment in response to change in CS.

Market Anomaly (relation with future-period stock return)

Fornell, Mithas, Morgeson, Krishnan (2006)

find no contemporaneous relation in an announcement day, but substantial returns to a portfolio of firms with high level of CS the next year stock

Mittal, Anderson, Sayrak, and Tadikamalla (2005)

link stock return and the level of CS. The reported effect is contradictory to efficient markets (i.e., since CS is autocorrelated, it can be used to predict CS in the next time period and earn abnormal stock returns)

Morgan and Rego (2006)

find of sig correlation between CS at period (t) to stock return at period (t+2)

Bob Jacobson and Natalie Mizik �Customer Satisfaction

14

15 of 30

Data: a total of 903 observations (1994-2006)

CRSP database for stock returns
Annual COMPUSTAT for accounting data
American Customer Satisfaction Index (ACSI) database provided on www.theacsi.com for CS
Risk factors from Kenneth French’s data library posted on http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/

We use 3 different measures of abnormal returns

(BHAR, CAR, CumAR)

Bob Jacobson and Natalie Mizik �Customer Satisfaction

15

16 of 30

Industry Affiliations

Bob Jacobson and Natalie Mizik �Customer Satisfaction

We use firms in SIC 49 is our “public utility” grouping. We have designated firms in SICs 35, 59, and 73 as our “Computer/Internet” grouping

16

17 of 30

The “Levels Model”

Bob Jacobson and Natalie Mizik �Customer Satisfaction

17

18 of 30

Stock Return Models: Raw Return, �the Role of Earnings, the Role of Risk

Bob Jacobson and Natalie Mizik �Customer Satisfaction

“It seems to me that there should be some kind of intervening endogenous variable that represents the outcome of the perceptions, perhaps sales. In other words, the … measures affect sales which then affect stock price behavior.” JMR Editor Comments on

Mizik and Jacobson (2005) “The Information Content of Brand Asset Attributes”

18

19 of 30

The Dynamic Performance Impact of Marketing Assets:�Estimation Framework

Bob Jacobson and Natalie Mizik �Customer Satisfaction

19

20 of 30

Stock Return Models: Raw Return, �the Role of Earnings, the Role of Risk

Bob Jacobson and Natalie Mizik �Customer Satisfaction

“It seems to me that there should be some kind of intervening endogenous variable that represents the outcome of the perceptions, perhaps sales. In other words, the … measures affect sales which then affect stock price behavior.” JMR Editor Comments on

Mizik and Jacobson (2005) “The Information Content of Brand Asset Attributes”

20

21 of 30

Utilities vs. Non-Utilities

Bob Jacobson and Natalie Mizik �Customer Satisfaction

21

22 of 30

Assessing the Value-Relevance of CS:�Differential Effects in the Computer and Internet Sector and other Non-Utility Sectors

Bob Jacobson and Natalie Mizik �Customer Satisfaction

22

23 of 30

Assessing Customer Satisfaction Anomaly:�Performance of Quartile Portfolios Formed Based on Lagged Customer Satisfaction

Bob Jacobson and Natalie Mizik �Customer Satisfaction

23

24 of 30

Assessing Customer Satisfaction Anomaly: �Performance of Quartile Portfolios Formed Based on Lagged Customer Satisfaction Growth

Bob Jacobson and Natalie Mizik �Customer Satisfaction

24

25 of 30

Assessing Customer Satisfaction Anomaly: �Future Premium Returns to �Customer Satisfaction Industry Leadership

Bob Jacobson and Natalie Mizik �Customer Satisfaction

Results are presented as mean [t-statistic], number of observations

Differential is the stock return of the top firm based on lagged Satisfaction (Panel A) or lagged Satisfaction growth (Panel B) in the 2-digit SIC minus the mean stock return for that year of the other firms in the same 2-digit SIC. Results are presented as mean [t-statistic], number of observations

25

26 of 30

Conclusions on � Customer Satisfaction Value Relevance

We find statistically significant value relevance of Customer Satisfaction for Computer and Internet firms

We find no statistically significant value relevance of Customer Satisfaction for non-Computer and Internet firms

Bob Jacobson and Natalie Mizik �Customer Satisfaction

26

27 of 30

Conclusions on �Customer Satisfaction Anomalies

We find no market anomalies associated with the level, growth, or industry leadership in CS per se.

We find market anomalies for firms with high and low levels of CS and high growth in CS in the Computer and Internet sector over the period of study

Bob Jacobson and Natalie Mizik �Customer Satisfaction

27

28 of 30

Directions for Future research

What are the conditions under which CS provides information valuable to the market? (i.e., are the results on computer/internet industry a fluke or will they persist into the future or occur in other sectors?)

Different CS initiatives might have different implications for performance (level, timing, and variability of earnings)

Can we better isolate the effect for non-utilities? When does the market incorporate information about CS as other info is released?

What other measures is market using to form expectation of firm future performance? Info is released 6 weeks later. The market is using some info because we see a reaction before info is released.

Bob Jacobson and Natalie Mizik �Customer Satisfaction

28

29 of 30

Abnormal Stock Return Definitions

Bob Jacobson and Natalie Mizik �Customer Satisfaction

29

30 of 30

Appendix

Bob Jacobson and Natalie Mizik �Customer Satisfaction

30