We examine the relationship between strategic positioning of firms and their production efficiency. Firms with competitive advantages based on either cost leadership or differentiation are able to outperform their competitors. Firms pursuing a cost leadership strategy seek to be the lowest cost producer, primarily by minimizing inputs for a given level of output, thus concentrating on increasing the efficiency of their production processes. On the other hand, firms that pursue a differentiation strategy rely on innovation, brand development, marketing, and so forth to achieve competitive advantages; therefore such firms do not place high emphasis on production efficiency. Thus the importance of production efficiency for the success of a firm depends on the strategic positioning of the firm. We apply DEA to an archival data for a large sample of publicly listed firms to investigate the importance of production efficiency for firms based on their strategic positioning. We provide empirical evidence that firms pursuing a cost leadership strategy attribute higher importance to production efficiency, while firms pursuing differentiation strategy attribute less importance to production efficiency.
Porter [
Calthrop, Vice President of Bain International, writes that “Cost leadership is about cost per unit of input, not lowest cost per se” [
Our study highlights a crucial dichotomy in the differential importance of production efficiency to firms pursuing different business strategies. Production efficiency enhancements help reduce costs and achieve cost leadership in a market; hence production efficiency is very important to cost leaders. Hence, business managers who are looking to enhance production efficiency should first evaluate their business strategy, to ensure that such enhancements will yield the maximum effect. Finally, this study shows the versatility in the application of DEA in its ability to answer research questions in diverse areas of business.
The rest of the paper is organized as follows. In the next section we provide the background literature to motivate our hypotheses. In Section
Fried et al. [
A firm needs to possess competitive advantages over its competitors in order to outperform them. Porter [
Firms adopting a differentiation strategy seek to create value through customer loyalty, price inelasticity, and unique image which they achieve via brand image, advertising intensity, novelty, fashion, and exclusive distribution networks [
We employ data envelopment analysis (DEA) to measure the production efficiency of the firms, which is an integral part of testing our hypotheses. DEA is a tool used to measure the production efficiency of decision making units. It is a nonparametric linear programming method that has become very popular in management science and economics to estimate production efficiency. Introduced by Charnes et al. [
The linear program is solved for each observation
Balsam et al. [
SG&A/SALES is the ratio of the selling, general, and administrative expenses to net sales. This measure captures the firms efforts to differentiate itself from customers through marketing oriented activities such as brand development [
Factor analysis to confirm strategy constructs (
Variables  Confirmatory factor analysis  

Cost leadership factor loading ( 
Differentiation factor loading ( 
Composite reliability  Average variance extracted (AVE)  
SG&A/SALES  0.96 (150.60)  
R&D/SALES  0.70 (119.60)  
SALES/COGS  0.54 (79.62)  0.79  0.57  
SALES/CAPEX  0.86 (171.70)  
SALES/P&E  0.91 (183.00)  
EMPL/ASSETS  0.51 (99.36)  0.82  0.61  


Goodness of fit measures  Goodness of fit index  0.9584  
Goodness of fit index adjusted for degrees of freedom  0.8907  
Bentler’s comparative fit index  0.9415  
Bentler & Bonett’s nonnormed index  0.8903 
SG&A/SALES = average of SG&A/net sales from
R&D/SALES = average of R&D Exp/net sales from
SALES/COGS = average of net sales/cost of goods sold from
SALES/CAPEX = average of net sales/Capital expenditure from
SALES/P&E = average of net sales/net book value of plant and equipment from
EMPL/ASSETS = number of employees/average of total asset from
Prior research [
We use a twostage method to evaluate our hypotheses on the impact of firm strategy on production efficiency. We estimate the second stage using two methods, OLS and Tobit regressions per Banker and Natarajan [
Since the production efficiency is a variable between 0 and 1, some prior studies use Tobit regressions in the second stage of a DEA analysis [
Following Bulan et al. [
We use the log of Assets as a proxy for firm size. ROA is return of assets computed as annual net income scaled by total assets. Age is the firm age computed by determining the year in which the firm first appeared in the COMPUSTAT database. Competition is computed as the sales of the firm in question as a percentage of the total sales of that firm’s industry segment to control for the competitive pressure faced by the firm. Finally, industry dummies, based on the FamaFrench 12 industry classification, are used as proxy for the industry structure.
We use publicly available data from the Compustat database for this study. Our dataset contains all firmyears that are available on the Compustat dataset for which the relevant variables have valid figures (nonmissing and not invalid observations). The dataset spans 1993 to 2006 and contains 33,499 firmyear observations.
Table
Descriptive statistics (
Variable  Mean  Median  Std Dev  25th percentile  75th percentile 

Production efficiency  0.5601  0.5913  0.2190  0.4313  0.7142 
Size  5.5660  5.4381  1.8304  4.1967  6.7772 
ROA  0.0116  0.0387  0.1438  −0.0075  0.0784 
Age  28.41  28.00  13.71  16.00  37.00 
Competition  0.0010  0.0002  0.0040  0.0000  0.0006 
See the appendix for variable definitions.
We evaluate our two hypotheses by estimating the regression model (
Strategic positioning and production efficiency (
Panel A  Panel B  Panel C  

Coefficient estimate 


Coefficient estimate 


Coefficient estimate 



Intercept  −0.0106  −1.26  0.2095  −0.0114  −1.37  0.1704  −0.0121  −1.46  0.1442 
CL  0.0099  4.71  0.0000  0.0056  2.67  0.0076  
Diff  −0.0378  −10.30  0.0000  −0.0368  −10.10  0.0000  
Size  0.0408  27.32  0.0000  0.0405  28.29  0.0000  0.0411  28.08  0.0000 
ROA  0.1890  25.97  0.0000  0.1940  27.07  0.0000  0.1921  26.78  0.0000 
Age  0.0008  6.44  0.0000  0.0006  5.32  0.0000  0.0006  5.02  0.0000 
Competition  5.1546  3.05  0.0023  5.0460  3.03  0.0025  4.9718  3.02  0.0026 
Industry dummies included  Industry dummies included  Industry dummies included  
Year dummies included  Year dummies included  Year dummies included  


Adjusted 
0.7594  0.7693  0.7697  
Total Obs  33499  33499  33499 
See the appendix for variable definitions.
Panel A tabulates the results using only the cost leadership measure, Panel B tabulates the results using only the differentiation measure, and Panel C tabulates the results with both the differentiation and cost leadership measures. The results in Panel A show a significant and positive relationship between cost leadership (coefficient estimate is 0.0099;
The coefficient estimates for the control variables, in each of the panels of Table
Next, we estimate model (
Strategic positioning and production efficiency (
Panel A  Panel B  Panel C  

Coefficient estimate 


Coefficient estimate 


Coefficient estimate 



Intercept  −0.0112  −3.04  0.0024  −0.012  −3.33  0.0009  −0.0128  −3.53  0.0004 
CL  0.00987  12.8  <0.0001  0.00556  7.29  <0.0001  
Diff  −0.0377  −40.1  <0.0001  −0.0367  −38.63  <0.0001  
Size  0.04102  109.4  <0.0001  0.04063  113.83  <0.0001  0.04126  112.46  <0.0001 
ROA  0.18929  44.26  <0.0001  0.1943  46.48  <0.0001  0.19236  45.97  <0.0001 
Age  0.00076  16.47  <0.0001  0.00061  13.6  <0.0001  0.00058  12.82  <0.0001 
Competition  5.15574  32.89  <0.0001  5.04723  32.94  <0.0001  4.97243  32.41  <0.0001 
Industry dummies included  Industry dummies included  Industry dummies included  
Year dummies included  Year dummies included  Year dummies included  


Log likelihood  27461  26758  27487  
Total Obs  33499  33499  33499 
See the appendix for variable definitions.
The results of Table
Our results indicate that the cost leadership strategy is associated with higher production efficiency. We conduct additional analysis to examine whether the link between production efficiency and cost leadership exists in the entire spectrum of the cost leadership variable. Specifically, we evaluate whether our results hold for the entire sample or are driven by certain subsets of it by creating pairs of subsamples of our data based on the cost leadership variable to evaluate whether the positive link between the two variables exists in every subsample. Each subsample pair is created by dividing the original sample into the top
Cost leadership strategy and production efficiency. Low cost leadership firms compared to high cost leadership firms. Dependent variable: production efficiency.
Panel A: High cost leadership (top 33%)  Panel B: Low cost leadership (bottom 33%)  

Coefficient estimate 


Coefficient estimate 



Intercept  −0.0829  −3.94  0.0000  −0.0119  −0.83  0.4041 
CL  0.1110  3.83  0.0001  0.0077  3.09  0.0020 
Size  0.0454  27.45  0.0000  0.0465  23.47  0.0000 
ROA  0.0879  7.87  0.0000  0.2475  17.54  0.0000 
Age  0.0005  2.87  0.0042  0.0005  2.21  0.0269 
Competition  1.2515  2.50  0.0126  12.5400  7.54  0.0000 
Industry dummies included  Industry dummies included  
Year dummies included  Year dummies included  


Adjusted 
0.8151  0.7287  
Total Obs  11074  11101 
See the appendix for variable definitions.
Panels A and B show that cost leadership is positive and statistically significant for both the top 33rd percentile subsample and the bottom 33rd percentile subsample. The coefficient of the top 33rd subsample is larger than the bottom 33rd subsample, indicating that the impact of cost leadership is greater for the subsample where cost leadership is higher. These results hold for the 10th, 25th, and 50th percentile subsamples as well. Furthermore, untabulated results show that a replication of the analyses comparing the top and bottom 50th, 33rd, 25th, and 10th percentiles using a Tobit regression yields the same results.
Next, we formally evaluate whether
Bootstrap results of significance of the difference in the CL coefficient between high and low cost leadership samples.
Coefficient estimate 


Coefficient estimate 




Panel A: Top 50th percentile versus bottom 50th percentile  
CL  0.0999  406.88  0.0000  0.0064  154.24  0.0000 
Difference  0.0935  375.38  0.0000  


Panel B: Top 33rd percentile versus bottom 33rd percentile  
CL  0.1167  265.55  0.0000  0.0062  132.76  0.0000 
Difference  0.1104  249.97  0.0000  


Panel C: Top 25th percentile versus bottom 25th percentile  
CL  0.1294  214.16  0.0000  0.0063  125.70  0.0000 
Difference  0.1231  202.97  0.0000  


Panel D: Top 10th percentile versus bottom 10th percentile  
CL  0.3183  186.75  0.0000  0.0100  144.808  0.0000 
Difference  0.3083  180.75  0.0000 
See the appendix for variable definitions.
The results in Table
To summarize, the results in Tables
We use an inputminimization DEA model since it is more appropriate for investigating the production efficiency of cost leadership firms. However, as a sensitivity measure, we also compute “production efficiency scores” using an output maximization DEA model. Our results do not change qualitatively and are therefore robust to different specifications of the DEA model.
This paper presents a unique application of DEA in examining the importance of production efficiency in the context of strategic positioning of firms, thus providing insights into an aspect of evaluating the effectiveness of firms in achieving their strategic positioning. The importance of clearly defined strategy for long term success of firms is well documented [
We examine this contention using publicly available data. We compute proxies for a firm’s strategic positioning using Balsam et al. [
When evaluating our results, it is important to keep in mind the criticisms of the 2stage approach to evaluating production efficiency by Simar and Wilson [
Our paper makes several important contributions to the literature. First, we document the association between firm strategy, specifically the cost leadership strategy and production efficiency. Second, we highlight how estimated production efficiency can be used to evaluate one important aspect of a cost leadership strategy. One facet of a cost leadership strategy can be implemented
Our results are important to strategy formulators, managers, financial analysts, and investors alike. For the former group, our study provides additional insights on cost leadership and differentiation strategies and on ways in which such strategies can be implemented and how such implementation can be evaluated. To the latter group our study provides a means of expost evaluating the implementation effectiveness of a firm’s stated strategy.
Our study points to several avenues for future research. One such research would be to investigate the continuing role that production efficiency plays in firm strategy, as the discipline of strategy itself evolves beyond the “generic strategy” paradigm of Porter [
Cost of goods sold (COGS, Compustat Data 41).
Selling and distribution (SGA, Compustat Data 189).
Capital expenditure (Cap Ex, Compustat Data 128).
Sales revenue (Compustat Data 12).
Production efficiency: Production efficiency of the firm (measured using DEA).
CL: Construct to capture cost leadership. Continuous variable, based on the factor analysis of the
Diff: Construct to capture differentiation. Continuous variable based on the factor analysis of the
Size: Natural log of assets (Compustat data #6).
ROA: Return on assets (Net income (Compustat data #18)/total assets (Compustat data #6)).
Age: Current year minus the year in which the firm first appeared in Compustat.
Competition: Sales of the firm scaled by the sales of all of the firms in the same industry category.
The authors declare that there is no conflict of interests regarding the publication of this paper.
Please see the appendix for definitions of these input and output variables.
Prior studies [
We estimate production efficiency assuming Hicksian neutral technical change and follow the procedure specified in Chang et al. [
The bootstrapping procedure we use is as follows. Consider the top
High productive efficiency is a sine qua non for a successful cost leadership strategy. DEA enables us to determine the production efficiency frontier for each industry. Therefore, a firm which has implemented a cost leadership strategy could evaluate the success of its strategy implementation by evaluating how close its post implementation production efficiency is to the efficiency frontier of its industry category. However, it must be noted that a high production efficiency by itself does not indicate a totally successful cost leadership strategy implementation; other aspects of a successful implementation would be as capturing market share, judicious selection of product lineup access to low cost raw materials, and so forth.