Organizational financial performance: identifying and testing multiple dimensions.
Carton, Robert B. ; Hofer, Charles W.
INTRODUCTION
There is little dispute that one of the core purposes of both
entrepreneurship and strategic management theory and research is the
improvement of organizational performance (Eisenhardt & Zbaracki,
1992; Venkatraman & Ramanujam, 1986). However, there is no consensus
regarding the best, or even sufficient, measures of organizational
performance.
It has long been reported that different measures of organizational
effectiveness and performance have been used in entrepreneurship and
management studies with little or no thoughtful discussion of why the
measures used in the studies were chosen (Cameron, 1986). Little
attention has been paid to the limitations that these measures may
impose on the interpretation or generalizability of the results of the
research. The most frequently used measures of organizational
performance are financial. However, no study has successfully proposed
and empirically tested a generalizable multi-dimensional model of
organizational financial performance. This is a particularly challenging
issue since changing environmental conditions may dictate that different
performance dimension priorities exist at different times. For instance,
during economic recessions, liquidity may be more crucial than
profitability, while during economic booms, profitability and growth may
take precedence.
This research examines the multi-dimensional structure of
organizational financial performance and seeks to empirically identify
distinct financial performance constructs and appropriate measures of
those constructs.
THE RATIONALE FOR THIS RESEARCH
The topic of this research is particularly important for several
reasons. First, a multidimensional model of organizational financial
performance has not previously been explicitly studied. However, in
1987, Venkatraman and Ramanujam empirically demonstrated that growth and
profitability were distinctly different measures of performance, but did
not attempt to propose a specific model for financial performance
measurement. In 1996, Murphy, Trailer and Hill examined the dependent
measures used in entrepreneurship research and through exploratory
factor analysis found nine distinct financial performance constructs
among the more than 50 different dependent financial performance
variables reported upon. And, in 1998, Robinson empirically tested the
relationship between four separate independent variables (stage of the
life cycle, industry concentration, entry barriers, and product
differentiation) with eight different measures of financial performance
used in new venture research and found significantly different results
between each independent variable and the eight different dependent
financial performance variables. This further demonstrated the existence
of multiple dimensions of organizational financial performance. Finally,
part of this research involved an analysis of 1,045 articles published
in the leading entrepreneurship and management journals between 1996 and
2001. Of these 1,045 articles, 138 purported to use overall
organizational performance as the dependent variable. Over 70% of these
138 articles used a financial performance measure as the dependent
variable. Further, 46% of these 138 articles used only a single measure,
25% used two measures, and 18% used three measures to represent
organizational performance. Overall, a total of 88 different dependent
measures were used to represent overall organizational performance in
these 138 articles, generally without any support for the validity of
the measures utilized. In short, it can be inferred from these
statistics that there is no consensus in the entrepreneurship and
strategic management research conducted over the 5 years with respect to
valid measures of organizational performance. However, it is also clear
from this prior research that organizational financial performance is
definitely a multi-dimensional construct.
Second, a generalizable and more powerful model for measuring
organizational financial performance has significant implications for
future research and for reexamining the findings of prior research where
less powerful dependent variable measurement models were used. Such a
model can help resolve multiple inconsistent theories where differing
dependent variables were used.
Finally, a multi-dimensional model of organizational financial
performance can significantly improve organizational stakeholders'
understanding of the effectiveness of management. It allows for the
measurement of value creation and for reasonable comparison across
companies that have chosen different routes to creating value.
THE IMPORTANCE OF EFFECTIVE MEASUREMENT
The development of valid operationalizations of the key concepts
and constructs used to form of both independent and dependent variables
in the models used in entrepreneurship and strategic management research
is fundamental to the description and corroboration of theoretical
relationships being tested in research, and is the essence of the
measurement stream of research. Put differently, the validity of
research studies that use arbitrary dependent measures to represent
overall organizational performance is highly questionable. Researchers
need to know that the effects they are studying will reasonably be
represented by the changes in the dependent variables. The use of
different measures as proxies for "performance" makes
extension from one study to the next dubious. Peter (1979) clearly
summed up the importance of construct measurement as follows:
Valid measurement is the sine qua non of science. In a general
sense, validity refers to the degree to which instruments truly measure
the constructs which they are intended to measure. If the measures used
in a discipline have not been demonstrated to have a high degree of
validity, that discipline in not a science (page 6).
Measurement is the "careful, deliberate, observations of the
real world for the purpose of describing objects and events in terms of
the attributes composing a variable" (Babbie, 1998:116). For a
variable to be clearly and equally understood by many different
individuals, it must be accurate, precise, quasi-invariant across
observers, provide discrimination from other variables, and be stable
over time. As a result, it is problematic that overall organization
performance has been "measured" in scores of research studies
by dozens of variables that are generally not strongly correlated over
time.
PRIOR STUDIES OF ORGANIZATIONAL PERFORMANCE
Despite the importance of accurately measuring organizational
financial performance, only seven studies have addressed the question of
how organizational financial performance is or should be measured.
Perhaps more importantly, none of these studies seem to have
significantly influenced how organizational financial performance is
actually measured in most of the empirical research that uses this
construct as a dependent measure. These seven studies fall into three
braod categories. They are (1) Studies that describe the measures
actually used in entrepreneurship and strategic management research.
These include the research of Brush and VanderWerf (1992), Murphy,
Trailer and Hill (1996), and Carton (2004). (2) Studies that focus on
the need for multi-dimensional measures of organizational financial
performance. These include the research of Rawley & Lipson (1985),
Chakravarthy (1986), and Venkatraman and Ramanujam (1987). And (3),
studies that seek to determine the "best" measure(s) of
organizational financial performance, i.e., Robinson (1996).
Studies of the Measures used in Entrepreneurship and Strategic
Management Research
Brush and VanderWerf (1992) examined thirty-four different studies
in the entrepreneurship literature that explicitly used firm performance
as the dependent variable. They found that thirtyfive different measures
of performance were used in those studies indicating that researchers
perceived many different dimensions of performance and that there was no
agreement on what measures actually represent organizational
performance. The most frequently used measures of performance were
changes in sales, organizational survival, changes in number of
employees, and profitability. Brush and VanderWerf state that they did
not attempt to sort out the problem of which performance measures to
use. However, they note that the fact that 35 different performance
measures were used in just 34 studies indicates that more work needs to
be done to identify measures that make sense for use across studies.
Murphy, Trailer and Hill (1996) examined the variables used to
measure organizational performance in entrepreneurship research in the
years 1987 through 1993. They identified 51 articles published in the
top entrepreneurship and strategic management journals that explicitly
used firm performance as a dependent variable. They found, consistent
with Brush & VanderWerf (1992) and Cooper (1993), that there was no
consistency in the variables used to measure new venture performance. In
total, they identified 71 different dependent variables used to measure
performance in the 51 articles.
Murphy et al. then examined 19 financial variables from a sample of
995 public firms with 500 or fewer employees. They found that less than
half of the intercorrelations between performance measures were
significant, indicating that these variables measured different
dimensions of performance. More than 25% of the significant correlations
of performance measures were negative. Murphy, et al. concluded that the
"... relationship between a given independent variable and
performance is likely to depend upon the particular performance measure
used." They further concluded ".research finding support for
an effect on one performance variable cannot justify the assumption that
the effect is similar across other measures of performance (1996:
21)." Their study also found that the performance measures tested
failed to meet the requirements of convergent and discriminant validity
necessary to validate a one-dimensional performance construct (Campbell
& Fiske, 1959).
Murphy et al. performed an exploratory factor analysis on the 19
variables, which yielded 9 factors that explained over 70% of the
variance in the performance measures. In an attempt to fully examine the
results obtained by Murphy et al., we performed a confirmatory factor
analysis of the nine factors identified in their research, using LISREL
8.12 as suggested by Bollen (1989). The covariance data reported in the
1996 study was used as the basis for the analysis. The results of this
analysis, reported in Carton (2004), indicated that the factors
determined by Murphy et al. did not fit the data. The model yielded
Chi-square = 1292 with 127 degrees of freedom (p<0.0001), Bentler and
Bonnet normed fit index of .727 and a non-normed fit index of .657. This
indicates that while the variables did load into factors, these factors
may not be supported by the data. This is possible since exploratory
factor analysis will force variables into the number of factors
specified, even though the factors may not have the best possible fit
for the data. Even if the data fit the model, because the analysis was
an exploratory factor analysis, there was no theoretical support for the
identified constructs. Consequently, the variables within each factor,
as determined by the exploratory factor analysis, fail the test for
convergent validity using confirmatory factor analysis, and they also
did not have any face validity.
Studies of the Multi-Dimensional Character of Organizational
Financial Performance
In 1985, Rawley and Lipson examined the relationships among several
combinations of performance measures to demonstrate that different
common measures of financial performance did not represent the same
attributes. Of these comparisons, the only overall performance measures
that they found to be related to each other at statistically significant
levels were the Q ratio versus cash flow return on investment
("CFROF) adjusted for the Capital Asset Pricing Model
("CAPM") discount rate, and market-to-book value versus return
on investment adjusted for inflation. The Q ratio was proposed by
Callard & Kleinman (1985) as a substitute for Tobin's Q, and is
calculated as the ratio of the value of individual business units
divided by the inflation adjusted purchase cost of assets. The other
measures that they compared were clearly discriminant and do not measure
the same construct. Table 1 summarizes their findings for the S&P
400 companies for the period 1982 through 1984.
Chakravarthy's 1986 comparative study of seven
"exemplar" firms with seven "maladapted" firms in
the computer industry developed an eight variable discriminant function
for the two groups. None of the individual profitability measures tested
in this research was capable of discriminating between the two groups.
The discriminant function developed included multiple dimensions of
performance, again indicating the importance of multivariate measures of
organizational performance.
Using confirmatory factor analysis ("CFA"), Venkatraman
and Ramanujam (1987) empirically examined the degree of convergence
across methods of measuring organizational financial performance and in
so doing, demonstrated that sales growth, profit growth, and
profitability were discriminate measures of different dimensions of
organizational financial performance. They selected these measures based
upon a review of the different performance dimensions typically used by
different disciplines done by Hofer (1983) and Woo and Willard's
(1983) findings of key dimensions of performance based upon an analysis
of PIMS data. The implication of this finding is that in isolation, none
of the three variables individually measure the organizational financial
performance construct. Consequently, the findings from a study that uses
sales growth to represent organizational financial performance should
not be equated to findings from a study that uses either profit growth
or profitability to represent business economic performance.
Studies of the "Best" Measure(s) of Organizational
Financial Performance
Robinson (1995) examined ten different new venture performance
measures to determine which individual measure was the most effective in
accurately assessing long-term economic value creation. The ten measures
studies were change in sales, sales level, return on sales, return on
invested capital, return on equity, return on assets, net profit,
earnings before interest and taxes, earnings multiples, and shareholder
value created. All ten performance measures were tested individually for
their relationship with multiple independent variables that had been
found in prior literature to have positive relationships to new venture
performance. The shareholder value created measure (also commonly known
as return to shareholders) was determined to be the most effective
measure for effectively differentiating among new venture strategies,
the second most effective measure for differentiating among the
structure of the new venture's entered industry, and the most
effective measure in differentiating among the interactions between new
venture strategies and the structure of the industry the new venture
entered. The fact that the different performance measures of overall new
venture performance resulted in significantly different r-squares
implies that the variables do not measure the same things.
Summary
In summary, it is clear from the prior empirical studies that there
has been no consistency in the measures used to represent the construct
of overall organizational performance in strategic management or
entrepreneurship research. Researchers compound the problem by confusing
determinants of performance with measures of performance ENRf8(Cameron,
1986). Further, prior empirical research has demonstrated that there are
multiple dimensions to the performance construct. While Robinson (1995)
found that return to shareholders was the most powerful individual
performance with respect to new venture performance among companies that
have undergone initial public offerings, these findings cannot be
reasonably generalized to studies that use different samples. In short,
there continues to be no conclusive research that has identified a
"best" measure of overall organizational performance, nor has
a measurement model that accurately represents the construct yet been
developed.
PROBLEMS WITH THE MULTI-DIMENSIONAL VIEW OF PERFORMANCE
As noted above, there is no consensus reflected in prior
entrepreneurship and strategic management research regarding the best or
even a sufficient set of measures of organizational performance.
However, most theorists have concluded that organizational performance
is multidimensional in character--a conclusion that is supported by all
seven of the studies that have explicitly addressed some aspects of the
question of how organizational financial performance is or should be
measured. Given these facts, the question immediately arises: "Why
don't more researchers use multi-dimensional measures of
organizational financial performance?" Two important reasons are
the facts that: (1) performance has often been characterized by the
purposes of the research being performed, and (2) there are many
different views on the most desirable outcomes of organizational
effectiveness. However, there is at least one additional reason for this
situation, namely the lack of any reasonably accurate
"individual" measures of organizational financial performance.
The "best" macro-measure of organizational financial
performance in the view of most accounting, entrepreneurship, finance,
and strategic management scholars is that of "shareholder wealth
creation" (Rappaport 1986). But, none of the traditional individual
measures of organizational financial performance is an effective
surrogate for "shareholder wealth creation." The
"best" of these traditional individual measures is Return on
Assets (ROA), and its correlation (R2) with shareholder wealth has been
found to be only 0.10 (Carton & Hofer, 2006), far below anything
that could be considered a "statistically significant"
relationship.
The primary objective of this research is to address this issue by
developing a new multidimensional model of organizational financial
performance. The secondary objective of this research is to develop a
model that is far "more robust" in its ability to
"explain" changes in "Shareholder Wealth" creation.
In the process, we show that the simultaneous consideration of these
multiple dimensions is more appropriate for drawing conclusions about
the effectiveness of managerial actions than is considering each
individual performance dimension separately.
DEVELOPING AND TESTING A MODEL OF ORGANIZATIONAL FINANCIAL
PERFORMANCE: OUR METHODOLOGY
The Macro-Design of This Study
Since no prior research has empirically established the domain of
organizational financial performance, this research was by necessity
exploratory in nature. The process used in this research to infer a
multi-dimensional model of organizational financial performance involved
four phases.
First, data on shareholder valuation and other financial
performance indices was collected from a sample of 1,500 public
companies. These data points were then separated into high, medium, and
low performance groups based on both annual and three-year financial
performance as indicated by their shareholder returns over the specified
period of time since shareholder return Robinson (1995) had found were
the most robust measure of organizational performance.
Second, the financial performance measures most commonly used in
past research and/or discussed in the entrepreneurship and strategic
management literatures were then used to compare the high and low
performance groups in order to identify the specific measures that
differentiated the two groups at statistically significant levels. Also,
each measure that was included in this research was evaluated both with
respect to its static value and its change in value over the period of
interest in order to determine the relative information content of
static vs. change scores.
Third, the financial performance variables (both static measures
and change score measures) that differentiated between the high and low
performing groups of companies in phase 2 were then grouped into
different financial performance constructs.
Finally, the validity of these different financial performance
constructs was tested.
Developing Our Sample of 3,819 Data Points
In order to secure data on shareholder valuations, it was necessary
to gather data from a sample of publicly owned firms. The population of
public companies used for this research was companies in the Standard
and Poor's 1500 on December 31, 2002 (a combination of the Standard
and Poor's 500, the Standard and Poor's Mid Cap 400, and the
Standard and Poor's Small Cap 600 indices). These companies include
a wide cross-section of industries, organizational sizes, and
organizational ages. Four years of financial data was collected on each
of these firms from the Standard and Poor's Compustat[R] database.
[Note that the fourth year was needed to be able to calculate the change
scores.] One and three-year performance variables (both static measures
and change score measures) were calculated from this data. Our maximum
potential sample was, therefore, 4,500 individual firm years of data and
1,500 three-year data points. Any firm-year or firm-period of data that
was incomplete was eliminated from the final sample. Also any firm-year
or firm-period of data that contained significant outliers was also
eliminated from the final sample. The final sample included 2,894
individual firm-years of data and 925 three-year periods of data.
Developing a Sample of High and Low Financial Performing Companies
Next, the individual annual firm years and three-year periods in
the sample were classified as having high, medium, or low financial
performance based upon their returns to shareholders. A firm was
classified as having high financial performance if its return to
shareholders was one standard deviation or more above the mean. A firm
was classified as having low financial performance if its return to
shareholders was one standard deviation or more below the mean. All
other firms were classified as having medium financial performance.
A total of 309 firm years were classified as having
"high" financial performance, 321 firm years were classified
as having "low" financial performance, and the remaining 2,264
firm years were classified as having "medium" financial
performance. In the case of three-year period data, a total of 124
three-year periods were classified as having "high" financial
performance, 143 threeyear periods were classified as having
"low" financial performance, and the remaining 658 threeyear
periods were classified as having "medium" financial
performance.
Identifying Financial Measures that Differentiate High and Low
Performance Firms
The two samples of high and low performing companies were used to
compare 20 of the financial measures most commonly used in past
entrepreneurship and strategic management research and from the
literatures of these fields using t tests to determine if there was a
statistically significant difference between the high and low performing
companies with respect to each of these 20 measures. [Note: A total of
40 tests were performed since each measure was examined using both
static and change score data.]
Those static and change score measures that indicated a
statistically significant difference between the groups were retained
for further evaluation in this research. Those static and change score
measures that did not discriminate at statistically significant levels
between the high and low financial performance companies were not
utilized further.
Testing Our Multi-Dimensional Model of Overall Organizational
Financial Performance
Having identified a set of financial measures that effectively
discriminated with a high degree of statistical significance between
high, medium, and low performance companies, these measures were grouped
into several different financial constructs based on a review of
accounting and finance literatures (see Breley, Myers & Marcus,
2001; Copeland, Koller & Murrin, 2000; Penman, 2001). This
combination of theoretical financial performance constructs together
with the financial performance variables used to represent them became
the model of organizational financial performance tested in the next
stage of this research.
The validity of these financial performance constructs was tested
using confirmatory factor analysis ("CFA") through the use of
AMOS 5.0 structural equation modeling software (Bollen, 1989;
Venkatraman & Ramanujam, 1987). The CFA framework uses a maximum
likelihood approach to providing a statistical analysis of the entire
validity of a construct and a decomposition of the measurement variance
into its constituent components (Bagozzi, Yi, & Phillips, 1991). The
proposed financial performance constructs and their various measures
were tested for both discriminant and convergent validity. The financial
performance constructs or measures that were not found to be valid were
eliminated from our multi-dimensional model of organizational financial
performance in order to create a revised multi-dimensional model of
organizational financial performance that contained only constructs and
measures shown to possess both discriminant and convergent validity.
RESULTS
Financial Measures That Discriminated Between High and Low
Performing Companies
This research found that, for annual periods, 35 of the 40 (20
static and 20 change score) financial performance measures tested
discriminated between high and low performance companies at p < .10,
using market adjusted return to shareholders as the basis for
classification. For three-year periods, 32 of the 40 (20 static and 20
change score) financial performance measures tested discriminated
between high and low performance companies atp < .10. Table 2
summarizes the results of the t tests for the variables that did and did
not discriminate between high and low performance companies for both
annual and three-year periods.
One paradox of management research described by Cameron (1986) was
that most empirical studies tend to use measures and methods that
explain average performance, while the primary focus should be on
understanding what makes firms either very successful or very
unsuccessful. It is, therefore, essential to select performance metrics
that can discriminate sufficiently among companies that perform at
different levels of performance. This research identified 32 annual and
27 three-year financial performance measures that discriminated between
high and low performance companies atp < .01 with respect to
market-adjusted returns to shareholders. Of these measures, those that
provided the most information about the return to shareholders referent
with respect to the sample of all companies, not just high and low
performing companies, should provide the most statistical power for
research where shareholder value creation is the phenomenon of interest.
Tests of Our Annual Multi-Dimensional Model
Figure 1 depicts the initial set of financial performance
constructs that was tested for both convergent and discriminant validity
for annual data. AMOS 5.0 software was used to test whether the annual
data fit the proposed model. Constructs with only one observed measure
were constrained to exactly equal the value of that measure, as
suggested by Bollen (1989). The variance of each construct was
constrained to unity so that the parameters for each observed variable
could be freely estimated.
The assumptions of structural equation modeling require that the
estimated variance of each measure must be positive and the covariance
matrices must be positive definite. Three fit indices, as recommended by
Arbuckle and Wothke (1999), were chosen to determine if the data fit the
model. These three indices included (1) the comparative fit index (CFI;
Bentler, 1990), (2) the Tucker-Lewis coefficient (TLI; Bentler &
Bonnet, 1980), which is also known as the Bentler and Bonnet non-normed
fit index (NNFI), and (3) the root mean square error of approximation
(RMSEA; Browne & Cudeck, 1993). Arbuckle and Wothke (1999) suggest
that a value of .90 or greater for both the CFI and the TLI indicates a
reasonable fit of the data with a model and, an RMSEA value of about
0.08 or lower, but certainly no greater than 0.10 indicates an
acceptable error rate for a model. Accordingly, a model that met all
three criteria was considered to have an acceptable fit.
A random sample of 150 high performing firms and 150 low performing
firms was used to test the model. This annual data did not fit the model
as proposed in Figure 1, since the covariance matrix for the constructs
that comprise the model was not positive definite. This indicated that
there was a specification error in the model or that the sample size was
too small (Joreskog & Sorbom, 1996). Since our sample had 300
observations, it was reasonable to conclude that the model had a
specification error.
An examination of the correlations between financial performance
variables across financial performance constructs revealed that residual
income return on investment ("RI ROI") was more highly
correlated with the profitability construct than with residual income
("RI") in the economic value construct. The model
specification error was eliminated by moving the RI ROI measure to the
profitability construct. The revised model had [chi square] = 2213.4
with df = 334 and n = 300. The CFI for the model was 0.83, the TLI was
0.78, and the RMSEA was 0.14. All three tests indicated that the data
did not acceptably fit the model. Accordingly, the financial performance
measures included in the model were reexamined to determine which should
be kept and which should be excluded from a revised model.
[FIGURE 1 OMITTED]
Figure 2 depicts the revised annual model of organizational
financial performance that met the tests for convergent validity of the
measures with the separate constructs. As a result of these changes in
the specification of the model, the data fit the revised ten-construct
model. The revised model had [chi square] = 405.4 with df = 130 and n =
300. The CFI for the revised model was 0.95, the TLI was 0.93, and the
RMSEA was 0.08. All three measures indicated that the data fits the
model. Accordingly, the revised model met the requirements of convergent
validity (Arbuckle and Wothke, 1999; Bollen, 1989; Joreskog & Sorbom
1996).
[FIGURE 2 OMITTED]
Using a confirmatory factor analysis framework, discriminant
validity is achieved when the correlations between the separate
constructs are statistically significantly lower than unity (1.0)
(Bollen, 1989). Any correlations that appeared "high" were
tested by setting the correlation between the two constructs equal to
unity, and then testing the statistical significance of the change in
[chi square] between the two models.
For the revised model, only two constructs, growth and growth
change, were correlated in excess of 0.60. Consequently, the correlation
between the two constructs was constrained to unity and the model was
retested. The result was [chi square] = 414.0 with df = 131. The change
in [chi square] was 8.6 with df = 1 andp < 0.01. Thus, these
constructs were found to be discriminant for the revised model. All
other combinations of constructs had correlations below 0.52, which
indicated that all constructs met the requirements for discriminant
validity.
In summary, 10 of the original 14 constructs and 20 of the original
30 measures were retained in the annual financial performance
measurement model. It was also demonstrated that these constructs were
discriminant and that the measures of the constructs met the tests for
convergent validity.
Having shown that it is possible to create an annual financial
performance measurement model that contains 10 financial performance
constructs and 20 financial performance measures that possesses both
discriminant and convergent validity, the question naturally arises as
to whether this model might be useful in future research. A complete
answer to this question is beyond the scope of this paper. The simple
answer, however, is YES! In a forthcoming paper, the authors develop a
composite annual financial performance measure from this annual
financial performance measurement model that explains over 46% of the
variance in market adjusted returns to shareholders, a 350% improvement
over all existing financial performance measures.
Testing Our Three-Year Multi-Dimensional Model
Figure 3 presents a diagram of the initial set of financial
performance constructs and their corresponding measures for three-year
data. The convergent validity of the financial performance measures for
each financial performance construct was tested using confirmatory
factor analysis, as was done with the annual data model.
The three-year data did not fit the proposed multi-dimensional
financial performance model. The model had [chi square] = 1013.2 with df
= 307 and n = 120. The fit indices were CFI = 0.81, TLI = 0.74, and
RMSEA = 0.14. There were specification errors in the model, as indicated
by estimated negative residuals for four measures including ROI, CROI,
GR Sales, and COC. Consequently, adjustments to the initial model were
necessary, as was the case with the annual data model. Therefore, the
financial performance measures included in the three-year model were
reexamined in order to determine which should be retained in a revised
model.
As a result of the changes made in the model specification, the
data fit the revised, 10construct model depicted in Figure 4. This model
had [chi square] = 212.1 with df = 137 and n = 120.
Thus, the revised multi-dimensional financial performance model met
the requirements for convergent validity (Arbuckle and Wothke, 1999;
Bollen, 1989; Joreskog & Sorbom, 1996). As with the testing of the
financial performance measurement model for annual data, any
correlations between constructs that appeared "high" were
tested by setting the correlation between the two constructs equal to
unity, and then testing the statistical significance of the change in
[chi square] between the two models. Two pairs of constructs appeared to
be "highly" correlated. They were (1) the change in
profitability and the change in survival and (2) the cost of equity
capital and the change in survival.
[FIGURE 3 OMITTED]
[FIGURE 4 OMITTED]
First, the correlation between the change in profitability and the
change in survival was constrained to unity, and the model was retested.
The result was [chi square] = 251.1 with df = 138. The change in [chi
square] was 39 with df = 1 and p < 0.001. Therefore, the constructs
were deemed to be discriminant. Next, the correlation between the cost
of equity capital and the change in survival was constrained to unity
and the model was retested. The result was [chi square] = 230.0 with df
= 138. The change in [chi square] was 17.9 with df = 1 andp < 0.001.
Thus, for the revised model, the constructs were deemed to be
discriminant.
In summary, 10 of the original 14 constructs and 20 of the original
30 measures were retained in the three-year financial performance
measurement model. It was demonstrated that these 10 constructs were
discriminant and that the revised set of 20 financial performance
measures of these 10 financial performance constructs met the tests for
convergent validity.
Having shown that it is possible to create a three-year financial
performance measurement model that contains 10 financial performance
constructs and 20 financial performance measures that possesses possess
both discriminant and convergent validity, again the question naturally
arises as to whether this model might be useful in future research. And
again, a complete answer to this question is beyond the scope of this
paper, but the simple answer is YES! In the forthcoming paper mentioned
above, the authors develop a composite three-year financial performance
measure from this three-year financial performance measurement model
that explains over 62% of the variance in market adjusted returns to
shareholders, a 520% improvement over all existing financial performance
measures.
SUMMARY AND CONCLUSIONS
For over fifty years, management scholars have suggested that
organizational performance is a multi-dimensional construct. However,
this is the first study that has undertaken to empirically identify both
the distinct dimensions of organizational financial performance and the
financial measures that represent these dimensions. The major
contribution of the research described in this paper is the development
of two (a one-year and a three-year) multi-dimensional models of
organizational financial performance that possess both discriminant and
convergent validity. Even more important, though, is the fact that these
models can be used to develop composite (one-year and three-year)
financial performance measures that correlate with shareholder wealth
350% (the one-year) to 520% (the three-year) better than any measures in
current use.
Limitations of this Research
Overall organizational performance is a multi-dimensional
construct. This research focused on only one of these dimensions, namely
financial performance. The operational and stakeholder dimensions of
overall organizational performance were not examined. As a consequence,
the relative importance of financial performance to the two other
organizational performance dimensions was not examined. A model of
overall organizational performance that includes all major performance
dimensions might require a different set of financial performance
dimensions or different measures of the identified financial performance
dimensions because of overlapping information across the higher order
constructs.
An additional limitation of the research design was the selection
of only two timeframes, one and three years. While these two timeframes
are those most frequently used in entrepreneurship and strategic
management research, 21% of the empirical studies summarized in this
research used other timeframes, most notably single point in time
measures and five-year measures. Also, the annual and three-year models
of financial performance developed in this research were not the same,
implying that different dimensions of financial performance are more or
less important at different times. Accordingly, the generalization of
the results of this research to other timeframes will require additional
testing.
A final limitation of the research design of the study was the
omission of risk from consideration in the financial performance model.
While financial risk was indirectly included through the calculation of
the cost of equity capital, strategic risk was not considered in any
way. Bromiley (1990) suggested that risk should be a component of the
analysis of organizational performance. However, it should be up to the
discretion of the researcher whether it is appropriate to capture the
effects of risk as an independent measure or as a component of the
dependent measure.
The sample utilized in this study limits the generalization of the
findings in several ways. These limitations include: (1) only U.S.
companies were included in the sample; (2) only publicly traded
companies were included; (3) financial services firms such as banks and
insurance companies were disproportionately eliminated from the sample
because they did not report sufficient information to calculate the
measures tested in this research; (4) only one three-year period was
used to develop and test the three-year financial performance model; (5)
the annual data was from the same timeframe as the three-year data; (6)
one primary source was used to gather most of the financial data; and
(7) the three-year sample did not include any companies that went out of
business during the three-year period.
Implications for Future Research
This research empirically demonstrated that organizational
financial performance is a multidimensional construct. Consequently, if
the unit of analysis for a study is the entire organization, and if the
phenomenon of interest is organizational financial performance, it is
incumbent upon the researcher to consider the effects of the independent
variables on multiple performance dimensions simultaneously.
However, if only one dimension of organizational performance is
examined in a study, then it is inappropriate to claim to be studying
the effects of the independent variables on "overall organizational
performance". In such situations, it would be more appropriate to
specify the specific dimensions of organizational performance being
studied, which would provide better context and understanding to readers
of the research.
This research identified 10 separate performance dimensions for
both annual and three-year timeframes as well as multiple measures of
these constructs. The theories being tested should help guide
researchers in selecting the dimensions of organizational performance
that they should examine. Specifically, those dimensions of
organizational performance that are hypothesized to create value for the
constituents of interest should be used as the dependent variables. In
general, dependent measures representing the phenomenon of interest
should be selected with great care and a full explanation for the
criteria used for their selection should be provided for users of the
research.
Implications for Management Practice
Potentially the most important finding for practioners is the fact
that the changes in performance metrics are separate constructs from
static measures of performance. Top management already places some
emphasis on changes in performance metrics as indicated by the content
of the management discussion and analysis (MD&A) sections in annual
reports and SEC filings. Typically, changes in performance
period-over-period are the primary focus of the MD&A section.
However, while changes may be explained post hoc, management planning
should focus on actions that will accomplish needed organizational
changes.
Those performance metrics that have been shown in this research to
discriminate between high and low performing companies should receive
the primary attention of management and users of financial statements.
Those companies that attain and sustain competitive advantage in the
market do not strive to be average. Therefore, the metrics they use to
gauge performance should focus on outcomes that set them apart from the
competition.
REFERENCES
Arbuckle, J. L., & Wothke, W. 1999. Amos 4.0 User's Guide.
Chicago: SmallWaters Corporation.
Babbie, E. 1998. The Practice of Social Research (8th ed.). New
York: Wadsworth Publishing.
Bagozzi, R. P., Yi, Y., & Phillips, L. W. 1991. Assessing
construct validity in organizational research. Administrative Science
Quarterly, 36: 421-458.
Bentler, P. M. 1990. Comparative fit indexes in structural models.
Psychological Bulletin, 107: 238-246.
Bentler, P. M., & Bonet, D. G. 1980. Significance tests and
goodness of fit in the analyssis of covariance structures. Psychological
Bulletin, 88: 588-606.
Bollen, K. A. 1989. Structural Equations with Latent Variables. New
York: John Wiley & Sons.
Brealey, R. A., Myers, S. C., & Marcus, A. J. 2001.
Fundamentals of Corporate Finance (3rd ed.). New York: McGrawHill.
Bromiley, P. 1990. On the use of finance theory in strategic
management, Advances in Strategic Management, Vol. 6: 71-98. New York:
JAI Press.
Brown, M. W., & Cudeck, R. 1993. Alternative ways of assessing
model fit. In K. A. Bollen, & J. S. Long (Eds.), Testing Structural
Equation Models: 136-162. Newbury Park, CA.: Sage.
Brush, C. G., & Vanderwerf, P. A. 1992. A comparison of methods
and sources for obtaining estimates of new venture performance. Journal
of Business Venturing, 7(March): 157-170.
Callard, C. G., & Kleinman, D. C. 1985. Inflation-adjusted
accounting: Does it matter? Financial Analysts Journal, May-June: 51-59.
Cameron, K. 1986. Effectiveness as paradox: Consensus and conflict
in conceptions of organizational effectiveness. Management Science,
32(5): 539-553.
Campbell, D. T., & Fiske, D. W. 1959. Convergent and
discriminant validation by the multitrait-multimethod matrix.
Psychological Bulletin, 56: 81-105.
Carton, R. B., 2004. Measuring organizational performance: an
exploratory study, Unpublished Doctoral Dissertation, The University of
Georgia, Athens, GA.
Carton, R. B. & Hofer, C. W. 2006Measuring Organizational
Performance: Metrics for Entrepreneurship and Strategic Management
Research, Northampton, MA: Edqard Elgar Publishing.
Chakravarthy, B. S. 1986. Measuring strategic performance.
Strategic Management Journal, 7: 437-458.
Cooper, A. C. 1993. Challenges in predicting new firm performance.
Journal of Business Venturing, 8(3): 241-253.
Copeland, T., Koller, T., & Murrin, J. 2000. Valuation:
Measuring and Managing the Value of Companies (Third ed.). New York:
John Wiley & Sons, Inc.
Eisenhardt, K. M., & Zbaracki, M. J. 1992. Strategic Decision
Making. Strategic Management Journal, 13: 17-37.
Hofer, C. W. 1983. ROVA: A new measure for assessing organizational
performance. In R. Lamb (Ed.), Advances in Strategic Management, Vol. 2:
43-55. New York: JAI Press.
Joreskog, K., & Sorbom, D. 1996. Lisrel 8 User's Reference
Guide (2nd ed.). Chicago: SSI Scientific Software.
Murphy, G. B., Trailer, J. W., & Hill, R. C. 1996. Measuring
performance in entrepreneurship research. Journal of Business Research,
36: 15-23.
Penman, S. H. 2001. Financial Statement Analysis and Security
Valuation. New York: McGraw-Hill.
Peter, J. P. 1979. Reliability: A review of psychometric basics and
recent marketing practices. Journal of Marketing Research, 18: 133-145.
Rappaport, A. 1986. Creating Shareholder Value: The New Standardfor
Business Performance. New York: Free Press.
Rawley, T., & Lipson, M. 1985. Linking Corporate Return
Measures to Stock Prices. St. Charles, Ill.: HOLT Planning Associates.
Robinson, K. C. 1995. Measures of entrepreneurial value creation:
An investigation of the impact of strategy and industry structure on the
economic performance of independent new ventures. Unpublished Doctoral
Dissertation, The University of Georgia, Athens, GA.
Robinson, K. C. 1998. An Examination of the Influence of Industry
Structure on Eight Alternative Measures of New Venture Performance for
High Potential Independent New Ventures. Journal of Business Venturing,
14: 165187.
Venkatraman, N., & Ramanujam, V. 1986. Measurement of business
performance in strategy research: A comparison of approaches. Academy of
Management Review, 11(4): 801-814.
Venkatraman, N., & Ramanujam, V. 1987. Measurement ofbusiness
economic performance: An examination of method convergence. Journal of
Management, 13(1): 109-122.
Woo, C. Y., & Willard, G. 1983. Performance Representation in
Strategic Management Research: Discussions and Recommendations. Paper
presented at the Academy of Management, Dallas.
Robert B. Carton, Western Carolina University
Charles W. Hofer, Kennesaw State University
Table 1: Summary of Relationships Among Selected Performance Measures
from Rawley and Lipson (1985)
Variables Compared [R.sup.2]
Price-to-earnings ratio vs. EPS growth .12
Price-to-book ratio vs. ROE less CAPM cost of equity .19
Price-to-book ratio vs. Return on capital employed--CAPM .34
cost of capital
Q ratio vs. CFROI less discount rate .65
Market-to-book ratio vs. ROI less inflation .71
Table 2: Measures That Discriminate Between High and Low Performing
Companies by Construct
Variable Annual Three-Year
Profitability ** **
Return on Assets ** **
Return on Equity ** **
Return on Sales ** **
Return on Investment ** **
EBITDA Return on Investment ** **
Operating Margin ** **
Growth
Growth Rate of Sales ** **
Growth Rate of Operating Expenses ** *
Growth Rate of Total Assets ** **
Growth Rate of Employees ** *
Cash Flow
Growth Rate of Operating Cash Flow * *
Operating Cash Flow to Equity ** **
Free Cash Flow to Equity NS NS
Liquidity
Liabilities to Total Assets NS NS
Survival
Altaian's Z Score ** NS
Efficiency
Asset Turnover + NS
Economic Value
Residual Income ** +
Residual Income Return on Investment ** **
Cost of Equity Capital
Cost of Equity Capital ** **
Market
Price to Book Ratio ** NS
Variable Annual Three-Year
Profitability ** **
Return on Assets ** **
Return on Equity ** **
Return on Sales ** **
Return on Investment ** **
EBITDA Return on Investment ** **
Operating Margin ** **
Growth
Growth Rate of Sales ** *
Growth Rate of Operating Expenses ** **
Growth Rate of Total Assets ** **
Growth Rate of Employees ** NS
Cash Flow
Growth Rate of Operating Cash Flow * NS
Operating Cash Flow to Equity NS **
Free Cash Flow to Equity NS NS
Liquidity
Liabilities to Total Assets NS **
Survival
Altaian's Z Score ** **
Efficiency
Asset Turnover ** **
Economic Value
Residual Income ** **
Residual Income Return on Investment ** **
Cost of Equity Capital
Cost of Equity Capital ** **
Market
Price to Book Ratio ** **
** p < .01 * p < .05 + p < .10 NS p > .10