An empirical analysis of manufacturing overhead cost drivers - Essay Example

An empirical analysis of manufacturing overhead cost drivers 1.0. Introduction and overview This work is a critical analysis of the work of Rajiv Banker, Gordon Potter, and Roger Schroeder on “An empirical analysis of manufacturing overhead cost drivers.” The work was published in the Journal of Accounting & Economics in 1995. When the authors wrote their work, Rajiv Banker and Roger Schroeder were with the Carlson School of Management in the University of Minnesota. Gordon Potter were with the School of Hotel Administration of the Cornell University. The Journal of Accounting and Economics received the work in April 1992 and the final version by July 1994 before the work was finally published in 1995. These suggest that although the work was published in 1995, the data that were the basis for the work of Banker et al. (1994), were taken much earlier than 1995. As of 2000, Rajiv Banker is a distinguished co-author of a book on management accounting, Management Accounting, with Anthony Atkinson, Robert Kaplan and S. Mark Young. The Prentice Hall’s “Learning on the Internet Partnership” documented that Banker is the Director of Accounting Programs at the University of Texas at Dallas. Further, according to the Prentice Hall’s “Learning on the Internet Partnership”, Banker holds the Ashbell Smith Chair in Accounting and Information Management of the University as of 2000. The Prentice Hall’s “Learning on the Internet Partnership” also document that Banker graduated from the University of Bombay at the top of his class and acquired a doctorate degree in business administration from Harvard University. Even as of year 2000, the “Learning on the Internet Partnership” has already recorded that Banker wrote at least 100 articles in leading research journals in accounting and related disciplines. According to a Cornell University website, Gordon Potter is still with the School of Hotel Administration of the Cornell University. The same website revealed that Cornell has a Ph.D. in Accounting from the University of Wisconsin-Madison, an M.S. in Accounting from the University of Wisconsin-Whitewater, and a B.A. Economics from Rutgers College-Rutgers University. The economics background of Gordon Potter may partly explain why the research methodology employed by Banker et al. (1995) are similar to those used in economics or in econometrics. In Cornell University, Potter has been an Associate Professor since 1993 until today (no indication in the Cornell University website that he was promoted in the University of Cornell since 1993). However, Potter has taught in the University of Wisconsin from 1982 to 1986. His resume in the Cornell University website does not reflect a job description entry from 1986 to 1993, possibly indicating that he took his PhD schooling during the period. Potter’s 5-page resume in the Cornell University website indicates Potter’s numerous achievements that included authorship of several papers. Roger Schroeder is a distinguished author in a textbook called Operations Management. According to the website of the Sage Publication, the ISI has classified him as the most highly cited scholar in the world in the field of Operations Management. The Sage Publication also reported that he holds the Frank A. Donaldson Chair in Operations Management at the Carlson University School of Management of the University of Minnesota. The Sage Publication website identified Schroeder as the author of over 100 articles in academic journals. The topic addressed by the work of Banker et al. (1995), overhead costs, is an important issue for manufacturers. This is indicated by the abundance of internet entries on the concern. A quick check on the internet indicates that there are at least 3.6 million entries on the worldwide web on the importance of reducing overhead costs for manufacturers. In 1995, during which the Banker et al. (1995) was published, overhead costs is one of the primary concern of manufacturers and manufacturers had viewed that overhead costs, compared to operational costs, are costs variable in which manufacturers have less discretion (Colin et al. 1995, p. 16). Thus, reducing overhead costs is an important concern for management. Understanding the fundamentals of overhead costs may allow manufacturers to have more control or discretion over their production operation. Basic economics suggest that overheads can be increased or decreased depending on whether the industry is experiencing diseconomy or economy of scale. Overhead can increase given increasing marginal cost of production. . On the other hand, improvements in productivity can decrease the overhead as production efficiency improves. Related to this, the work of Banker et al. (1995) on the manufacturing overhead cost drivers is highly relevant. Banker et al. (1995) touches on the variables that managers, using the tools of accounting, can use to monitor production and on which policies can be based towards improving work and efficiency. The article title of Banker et al. (1995) fully reflects the concern of the article and is, therefore, appropriate. Banker et al. (1995, p. 116) compared their work to earlier works on overhead costs such as with Miller and Vollmann. In summary, they pointed out that while earlier works have tended to show that while earlier studies tended to show empirically that overhead varies with transactions and not with volume, their study sought to validate the observation through their empirical study (Banker et al. 1995, p. 116). 2.0. Overall description/summary of the article The work of Banker et al. (1995) examined the “empirical validity of the claim that overhead costs are driven not by production volume but by transactions resulting from production complexity” (p. 115). The study addressed its research problem using data from 32 manufacturing plants from the electronics, machinery and automobile component industries (Banker et al. 1995, p. 115). The 32 manufacturing plants were selected based on the following process (Banker et al. 1995, p. 115): Sixty plants were randomly selected from a directory of manufacturing plants covering three industries: electronics, machinery, and automobile. Only 42 out of 60 agreed to participate in the study. Out of the 42, however, only 34 were willing to reveal their cost figures. Thirty-two became the final figure because 2 of the 34 plants have incomplete information. Measures of transaction were measured by the following variables (Banker et al. 1995, p. 115): number of engineering change orders number of or purchasing and production planning personnel, shop floor area, and number of quality control and improvement personnel. According to Banker et al. (1995, p. 117 last paragraph), their research methodology used “direct labor dollars” as the measure of the production volume. Further, Banker et al. (1995, p. 118) asserted that their measures of transaction are not directly related to volume. Because of this, Banker et al. (1995, p. 118) noted that their study “provide evidence that manufacturing overhead costs are associated with transactions that are not directly related to volume.” Banker et al. (1995, p. 118) pointed out that based on the work of Miller and Vollmann, there are four types of transactions: logistical, balancing, quality, and change transactions. Logistical transactions are related with receiving and moving materials in the plant (Banker et al. 1995, p. 118). Balancing transactions are related with coordination activities to ensure that “supply of materials, labor, and capacity equals demand” (Banker et al. 1995, p. 118). However, in the study and without offering any neither an explanation nor a justification, Banker et al. (1995, p. 119) used the number of product lines as the measure of balancing transactions. Quality transactions are those related “to insure that goods are produced to customer requirements” (Banker et al. 1995, p. 119). Finally, change transactions are those related with revising “manufacturing systems for alterations in product or process design” (Banker et al., 1995, p. 119). Table 1 (see Annex 1) enumerates the transaction variables utilized by Banker et al. (1995) study. The transaction variables cover logistical, balancing, quality, and change variables. Unfortunately, Banker et al. (1995) did not clarify how the logistical transaction variable is measured but the measure is definitely related to the square feet of shop floor devoted to production (Banker et al. 1995, p. 115, 117-118, 127). AREAPP or the square feet per floor is the “surrogate measure for logistical transaction” (Banker et al. 1995, p. 118). The logic behind the measure is that “the greater the area over which batches of materials need to be moved and stored, the greater the demand for support resources to perform logistical transactions” (Miller and Vollmann as cited by Banker et al. 1995, p. 118). As reflected on Table 1, the balancing transaction variable or PPPPER refers to number of purchasing and production planning personnel. The quality transaction variable or QUALPER refers to the number of quality control and improvement personnel. NECO refers to the number of change orders. The logic behind the use of the transactions assumes that these variables are important in shaping or determining the cost of overhead. Other than the variables associated with transactions costs, the other variables utilized for the regressions are DIRLAB, PLTEQP and NADMFN. DIRLAB is direct labor costs and the variable is supposedly measuring production volume. PLTEQP is the book value of plant and equipment in thousand dollars. Variable NADMFM is the number of administrative functions in the plant. The transaction variables reflected on Table 1 plus the production volume variable (represented by direct spending for labor in thousand dollars) and other independent variables together comprise the complete list of independent variables in Table 2 (see Annex 2) utilized by Banker et al. (1995). As pointed out earlier, not enough justification were offered for the choice of the other variables like PLTEQP. MOH is the dependent variable which is the annual manufacturing overhead costs in thousands of dollars. The statistical or econometric or regression models devised by Banker et al. (1995) assumes without adequate justification nor support the following multivariate regression model or the functional relationship between manufacturing costs represented by h, volume represented by v, and transactional variables represented by zk (Banker et al. 1995, p. 122): (equation 1) Taking the natural logarithm of the regression model mentioned yields the following (Banker et al. 1995, p. 122): . (equation 2) Equation 2 is the log-linear model consistent with Brooks (2008, p. 176). Following Brooks (2008, p. 176), the applicable interpretation for the log-linear model is that a unit increase in the independent variable will cause a 100x % or a increase in h. In the preceding equation, the capital letters represent the natural logarithm of the variables mentioned earlier. The preceding regression model was the subject of the estimation procedures of Banker et al. (1995). Like in all regressions, the significance of volume and transactional variables are assessed based on the significance of the coefficients associated with the independent variables. The fundamental results obtained by Banker et al. (1995) are recorded in Table 3 (see annex 3). Note that there are several versions as the work of Banker et al. (1995) tried several versions of the regression by removing one or several variables or adding a few ones. This approach is called data mining that was criticized by Gujarati (2004, p. 74). As indicated by Table 3, Banker et al. (1995) produced five empirical regression models for the basic regression function given by in Banker et al. (1995, p. 122). Using other estimation techniques, however, the estimates for the coefficients of model reflected on Table 3 are provided by Table 4 (see Annex 4). Banker et al. (1995, p. 134) also produced a regression model in which the dependent variable is the natural log of total manufacturing costs. This is in Table 5 (see Annex 5). Banker et al. (1995, p. 115) interpreted their results as indicative that overhead costs are strongly correlated with both the number of manufacturing transactions and the volume of production although most of the variation in overhead costs is explained by the measures of manufacturing transactions rather than by the production volume (Banker et al. 1995, p. 115). 3.0. Significance of the work of Banker et al. (1995) The work of Banker et al. (1995) can be interpreted as a work that tested how the economic law of increasing marginal costs and management innovation or lack of innovation can increase the overhead. According to the economic law of increasing marginal costs, when plant size is fixed, there comes a point when the additional costs of producing an additional unit of good increases as more of a good is increased. However, what is suggested in the work of Banker et al. (1995, p. 115), while marginal costs may increase as more of a good is produced, management can soften or even eliminate the increase in marginal costs via management operation or decreasing the number of transactions as more of the good is produced. This is equivalent to increasing efficiency in work as production volume is increased. Thus, while increasing the volume of production may increase the cost of overhead costs, reducing the number of transactions or increasing work efficiency can also reduce the overhead costs. Another way of looking at the significance of Banker and colleagues is that their work attempts to sustain the literature on management accounting that “production and support activities other than direct labor drive manufacturing overhead costs” (Banker et al. 1995, p. 116). Banker et al. (1995, p. 116) that the work of Miller and Vollman pointed out, however, that “the real driving force behind manufacturing overhead costs is not production volume but transactions dealing with logistics, balance, quality, and change” (Banker et al. 1995, p. 116, citing the work of Miller and Vollman). The work of Banker et al. (1995, p. 116) obviously attempted to provide evidence that both production volume and transactions drive overhead costs although the overhead costs are more responsive to changes in the transactions costs. Further, as pointed out by Banker (1995, p. 116), another significance of the work of Banker et al. (1995) is that the work highlights, “the need for an empirical investigation of the possible existence of multiple cost drivers.” Banker (1995, p. 116) pointed out that, “one consequence of distorted product costs is the likelihood that manufacturers will over-emphasize less profitable product lines.” Thus, Banker et al. considered their work as an attempt to have a better look at the drivers of the costs of overhead. 4.0. Critical analysis We can identify several areas of concern in the work of Banker et al. (1995). First, one broad area of concern is associated with economic methodology. The Banker et al. (1995, p. 121) focused on the superiority of their multivariate approach over the correlation analysis approach of Foster and Gupta (1990). As pointed out by Banker et al. (1995, p. 121), Foster and Gupta chose correlation analysis because they did not deem it appropriate to specify “a functional form relating volume and transactional variables to manufacturing overhead (MOH) costs.” The Foster and Gupta argument is important and their concern is highly legitimate because Gujarati (2004, p. 506) pointed out that if the model is not “correctly” specified then the problem of model specification error or model specification bias is encountered. In order to remove the model specification error or model specification bias, Gujarati (2004, p. 507) recommended that models must be developed that are consistent with theory. Unfortunately, the work of Banker et al. (1995) offered zero defense that their regression models based on theory. In econometrics on which the work of Banker et al. (1995, p. 122) is founded, this is unacceptable and does not constitute a valid work especially if the classical regression perspective is adopted. According to Gujarati (2004, p. 509), one type of error is the omission of relevant variables in the model. According to Gujarati (2004, p. 509), this type of error will result to a measurement bias. Another type of error is the opposite or the inclusion of irrelevant variables (Gujarati 2004, p. 208). Again, according to Gujarati (2004, p. 513), this error will lead to estimations errors. Inclusion of irrelevant variables and omission of relevant variables are also called errors related to overfitting or underfitting a model. Second, another area of concern is in the use of cross-section analysis. It seems standard in econometrics or regression techniques that panel data could yield better results than cross-section studies. Thus, the use of cross section over panel data studies must be justified. For instance, the inavailability of the data over time could have been used as good arguments on why the study employed cross-section over panel data. Third, the study pointed out that there is a potential for “spurious correlations to generate associations that are not based on causal relations” (Banker et al. 1995, p. 120). In so doing, what is incorrectly suggested is that time series studies are less prone to spurious correlations. In truth, however, time series studies are more vulnerable to yield spurious correlations because apart from the usual possibility of coincidence, two or more variables may have a common or similar trend that would be reflected as either negative or positive correlations between or among the variables. Not only would the causation measure would be spurious but the correlation that would be suggested by the correlation measures itself would be spurious. Fourth, the regression of Banker et al. (1995) is probably spurious. Overhead may be fixed within a production range. Outside of a production range, it can be increasing if there is diseconomy in size or it could be decreasing as a result of economies of scale. Lumping together 32 firms for an overhead function can be equivalent to lumping together firms with diseconomies to scale with firms that have acquired economies of scale relative to firm size. The variable used by Banker et al. (1995) for production volume is labor dollars. This is clear in Banker et al. (1995, p. 117, last paragraph, last sentence), “we use direct labor dollars to measure volume.” However, laborers have different wages rates. Some firms are labor-intensive while others are not. Some industries are skills-intensive, others industries are not. Some skills are very costly, some are only costly, and other skills are not costly. Thus, it is really doubtful whether direct labor dollars are really capturing production volume across several industries. Moreover, the PLTEQP or net investment (book value) varaible in plant and equipment is not adequately justified in the Banker’s regression Fifth, the regression techniques used by Banker et al. bear the signs of being massaged by data-mining techniques. In data-mining, one tries to find a regression with the best fit by adding or removing variables. A ridiculous way of data-mining is to add as independent variables one’s height, bust line, weight, or vital statistics and see if the fit of the regression improves and then offer justifications why the variable should be or can be included. Data mining results to specification bias or specification error that is criticized by Gujarati (2004, p. 74). Brooks (2008, p. 106) called data mining as “data-snooping.” According to Brooks, in data-mining or snooping, the true significance level will be considerably higher than the nominal significance assumed (Brooks 2008, p. 106). In other words, what seems significant is actually not significant. The significant coefficients are usually marked with *, **, and *** to indicate significance at the .10, .05, and .01 levels. Thus, the level of significance will be invalid with data mining or data snooping. According to Brooks (2008, p. 106), “a relationship observed in the examination period that is purely the result of data mining, and is therefore spurious, is very unlikely to be repeated for the out-of-sample period.” Therefore, according to Brooks (2008, p. 106), “models that are the product of data mining are likely to fit very poorly and to give inaccurate forecast for the out-of-sample period.” Finally or sixth, another area of concern is in the diagnostic tests appropriate for the statistic techniques used by Banker et al. (1995). For example, the study should have the possibility of collinearity among the independent variables. This means that the independent variables are somehow related with one another. According to Brooks (2008, p. 170), if the independent or explanatory variables have not relationship with one another, adding or removing an explanatory variable would not cause the values of the coefficients of the other variables to change. However, as we have seen in Table 3 and Table 5, the values of the coefficient change when variables are added or removed in the regression equation. This means that there is multicollinearity among the independent or explanatory variables. It is not difficult to test for collinearity among the independent or explanatory variables because softwares are available that can check for multi-collinearity among the dependent variables. 5.0. Implications of Banker et al. (1995) for management accounting As suggested by the title of Collier’s book in 2003, management accounting or the relevance of accounting for manager is to interpret accounting information for decision-making. The concern of Banker et al (1995) is to assess the empirical validity of the view that overhead costs are driven not by production volume but by transactions resulting from production complexity. As pointed out, the key finding derived by Banker et al. (1995) from their study is that while this is the case or that transactions can drive the cost of overhead, production volume also plays a role although not as well as the transaction variables. Nevertheless, regardless of whether the Banker et al. empirical study resulted to validly derived conclusions based on the data, the implication of the Banker et al. (1995) study is that managers must monitor not only production volume but also costs related to transactions if the concern is to reduce the cost of overhead. However, with or without the Banker et al. study, accountants and managers will probably monitor the same. Thus, at minimum, the Banker et al (1995) provides a reminder on the role of production volume and transactions in raising the overhead. Another way of defining management accounting is that it is an area of study instructing how managers can use accounting techniques, measures, or data for management. Defined this way, the work of Banker et al. (1995) identified some of the key variables that can increase overhead and these are task complexity and production volume. At the same time, Banker et al. (1995) pointed out how overhead costs can be reduced even if production volume is high. Based on the study of Banker et al. (1995), the key towards reducing overhead despite a high volume of production is via increasing efficiency or productivity or via reducing the number of various types of transactions. 6.0. Suggestions for improving the work of Banker et al. (1995) There are several ways of improving the study. First, at the fundamental level, the basic issue that the work of Banker et al. (1995) has to address is the theoretical foundation of their regression models. As we have pointed out earlier, regression techniques require a strong basis in theory for the exclusion and inclusion of explanatory variables in the regression model. Founding a regression model more strongly on theory would prevent the errors of over-specification and under-fitting (Gujarati 2004, p. 510). This is also in line or consistent with the prescription of Gujarati (2004, p. 74) and the criticisms of Brooks (2008, p. 106) against data mining. Second, perhaps one of the ways to improve the work of Banker et al (1995) is to factor in the role of computerization. Recall that the work of Banker et al. (1995) was done in 1995 when computerization and internet technology were yet very young. Today, with the advances in computer technology, it has been possible to implement several tasks simultaneous or do “multi-tasking”. It is therefore possible that with computerization the role between transaction and overhead has been considerably weakened. Another possibility is that computerization can moderate the relationship or the correlation between transactions and overhead. Third, the work of Banker et al. (1995) may be a step in the wrong direction. Perhaps, the better way to approach the research agenda of Banker et al. (1995) is via the route of building theories rather than “empirical studies.” The opinion being forwarded by this work is that what seems to be more important is to identify the costs drivers of the firm, including he costs drivers on overhead. The route of theory-building being suggested by this work is that studies must concentrate on developing ways instead of developing mathematical models of the costs components of a firm or the costs functions of the firm. Through such models, it will be possible to anticipate how the movement of one or several variables may affect the overhead and the other costs of a firm. In other words, studies on the role played by changes in specific variables on overhead and other total costs must focus on developing techniques on how the specific and overall costs of the firm are changes in one, two, several, or many variables simultaneously. With the development of techniques on how all the costs concerns of the firms can be modeled, the task of empirical validation could take place at the firm level and the improvement of the cost models can go on indefinitely. In other words, the theory-oriented works must concentrate on developing the techniques for developing cost models that management accounting practitioners can use at the firm level to anticipate how various cost figures, like the overhead, will be affected with the changes in prices and other variables. Meanwhile, the academic journals can study the mean, standard deviation, and percentile distribution on the percentages of contribution of various cost variables on the cost figures of a firm. Of course, it is also possible to develop large studies in which cost models can be developed both at the firm and industry levels. The work of Banker et al. (1995) may be useful but it may be more useful to develop techniques that would enable us to predict costs at the firm level. After all, we all know the production volume and all costs including transaction costs will surely contribute to costs anyway and firms probably differ in cost profile. For example, a firm may be extremely labor intensive and, thus, costs related to labor would influence the cost figures and even the overhead figures (via the size of supervisors or personnel managers that will be required) more than the changes in the values of the other variables. However, other firms may be transaction-intensive. This is the type of firms that will be most vulnerable to cost changes in transaction costs. The vulnerability will be in total as well as in specific costs like overhead costs. 7.0. Reflective summary The work of Banker et al. (1995) sought to examine “the empirical validity of the claim that overhead costs are driven not by production volume but by transactions resulting from production complexity” (p. 115). Using regression techniques, Banker et al. (1995) succeeded to provide evidence that indeed the view that overhead costs are affected not only by transactions costs but by volume variables as well. The empirical evidence for the Banker et al. (1995) perspective on overhead costs or on the drivers of overhead costs are in the statistical tables of the of the work of Banker et al. (1995). In summary, however, there are serious questions of validity in the work in the work of Banker et al. (1995). At the same time, even if the Banker et al. (1995) regression models are valid, the most important insight that the Banker et al. (1995) regressions can make is that productivity can reduce the overhead costs, something that is fundamental anyway in any management accounting theory and practice. Thus, the Banker et al. (1995) finding has a limited relevance in management accounting. The Banker et al. (1995) regressions can be interpreted in this manner: 1. Overhead need not increase even if volume of production increases by reducing the costs of transactions. 2. Overhead need not increase even if transaction costs increase if greater quantities can be manufactured at less cost. Of course, we must recall that Banker et al. (1995) “simply” made the study to offer an empirical validation of the role of transaction volume variables on the cost of overhead. However, the empirical validation will only be valid if the research method is also valid. The Banker et al. (1995) provides evidence that both transaction and volume variables can affect the cost of overhead but the evidence that the study provides are not conclusive if not useless. In this work, we have pointed out how the methodology of Banker et al. (1995) can be improved. At the same time, we have identified that perhaps computerization can be an intervening variable if the work of Banker et al. (1995) would be replicated today. We have also pointed out that a more useful endeavor is to focus on the development of techniques that would allow management accounting professionals to create costs models that can be used to anticipate the possible impact of how various changes in the relevant variables can affect the cost figures of a firm. The opinion expressed by this work is that the perhaps the management accounting professionals should concentrate on the endeavor. References Banker, R., Potter, G., and Schroeder, R., 1995. An empirical analysis of manufacturing overhead cost drivers. Journal of Accounting & Economics, 19, 115-137. Brook, C., 2008. Introductory econometrics for finance. 2nd Ed. Cambridge University Press. Campbell, J., Lo, A., and MacKinlay, C., 2008. The econometrics of financial markets. Princeton: Princeton University Press. Colin, A., Bowman, C., and Newton, J., 1995. Manager’s perceptions of the importance of supply, overhead and operating costs. International Journal of Operations & Production Management, 15 (3), 16-28. Collier, P., 2003. Accounting for managers: Interpreting accounting information for decision-making. West Sussex: John Wiley & Sons Ltd. Foster, G. and Gupta, M., 1990. Manufacturing overhead cost driver analysis. Journal of Accounting & Economics, Jan., 309-337. Gujarati, D., 2004. Basic Econometrics. 4th Ed. New York & London: McGraw Hill Companies. Miller, J. and Vollmann, T., 1985. The hidden factory. Harvard Business Review, Sept-Oct., 142-150. Annex 1: Table 1. Transaction variables utilized by Banker et al. (1995) . Source: Banker et al. (1995, p. 123) Annex 2: Table 2. Complete list of variables utilized by Banker et al. (1995) Source: Banker et al. (1995, p. 127) Annex 3: Table 3. Parameter estimates relating overhead costs to volume and transaction variables Source: Banker et al. (1995, p. 129) Annex 4: Table 4. Estimates of coefficients of the regression model using other estimation techniques Source: Banker et al. (1995, p. 132) Annex 5: Table 5. Model in which values of the dependent variable are in logarithm Source: Banker et al. (1995, p. 134) Read More