Enterprise resource planning systems: comparing firm performance of adopters and nonadopters

Transcription

1 International Journal of Accounting Information Systems 4 (2003) Enterprise resource planning systems: comparing firm performance of adopters and nonadopters James E. Hunton a, *, Barbara Lippincott b, Jacqueline L. Reck b a Department of Accounting, Bentley College, 175 Forest Street, Waltham, MA , USA b School of Accountancy, University of South Florida, Tampa, FL, USA Received 15 April 2002; received in revised form 10 October 2002; accepted 18 October 2002 Abstract The current study examined the longitudinal impact of ERP adoption on firm performance by matching 63 firms identified by Hayes et al. [J. Inf. Syst. 15 (2001) 3] with peer firms that had not adopted ERP systems. Results indicate that return on assets (ROA), return on investment (ROI), and asset turnover (ATO) were significantly better over a 3-year period for adopters, as compared to nonadopters. Interestingly, our results are consistent with Poston and Grabski [Int. J. Account. Inf. Syst. 2 (2001) 271] who reported no pre- to post-adoption improvement in financial performance for ERP firms. Rather, significant differences arise in the current study because the financial performance of nonadopters decreased over time while it held steady for adopters. We also report a significant interaction between firm size and financial health for ERP adopters with respect to ROA, ROI, and return on sales (ROS). Specifically, we found a positive (negative) relationship between financial health and performance for small (large) firms. Study findings shed new light on the productivity paradox associated with ERP systems and suggest that ERP adoption helps firms gain a competitive advantage over nonadopters. D 2003 Elsevier Science Inc. All rights reserved. Keywords: Enterprise resource planning, ERP; Firm performance; Productivity paradox; Longitudinal study * Corresponding author. Tel.: address: jhunton@bentley.edu (J.E. Hunton) /03/$ see front matter D 2003 Elsevier Science Inc. All rights reserved. doi: /s (03)

2 166 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Introduction Enterprise resource planning (ERP) reflects an innovative business strategy, as ERP adoption involves business process improvement, best practices implementation, intraenterprise integration and inter-enterprise coupling. ERP systems are designed to facilitate the ERP concept by replacing disparate patchworks of legacy systems across business organizations with synchronized suites of enterprise-wide applications. Potential benefits of an ERP system include productivity and quality improvements in key areas, such as product reliability, customer service, and knowledge management. As a result, ERP systems are expected to enhance market value and firm performance through efficiency and effectiveness gains. Hayes et al. (2001) offered evidence that the capital market placed incremental value on firms that adopted ERP systems, as investors reacted positively to ERP implementation announcements. Similarly, a behavioral study by Hunton et al. (submitted for publication) found that financial analysts significantly increased mean earnings forecast revisions when a firm announced plans to implement an ERP system. While both studies indicated that capital market participants believed ERP adoption would improve future firm performance, the extent to which expected returns eventually materialize remains unknown. To investigate this issue, Poston and Grabski (2001) examined the effect of ERP systems on firm performance over a 3-year period. They found a significant decrease in the ratio of employees to revenues in all 3 years, and a reduction in the ratio of cost of goods sold to revenues in year 3. However, they reported no significant improvement in the ratio of selling, general and administrative expenses to revenues, or residual income (net operating income minus imputed interest). Hence, they suggested a contradiction while ERP systems appear to yield efficiency gains in some areas, higher offsetting cost-to-revenue increases elsewhere leave residual income unaffected. Other researchers have also observed little or no performance effects associated with increasing information technology (IT) expenditures, a phenomenon that is often referred to as the productivity paradox (e.g., Grover et al., 1998; Harris 1994; Pinsonneault, 1998). Robertson and Gatignon (1986) and Hitt and Brynjolfsson (1996) suggested another way to look at the productivity paradox; that is, to the extent that increased spending on IT yields efficiency and effectiveness improvements, firms will pass on financial gains to consumers through decreased prices in a competitive marketplace. To investigate this possibility, we examined the longitudinal impact of ERP adoption on firms by comparing financial performance indicators of adopters and nonadopters. To the extent that adopters realize and transfer financial rewards, performance of adopters might not change using a pre- to postadoption analysis; however, the performance of nonadopters would be expected to decline by comparison. As expected, our results indicate that ERP adopters performed significantly better than non-erp adopters, primarily due to declining performance of non-erp adopters. Further, we examined whether the financial performance of ERP-adopting firms was affected by the interaction of firm size and financial health. Theory and research evidence indicates that the performance of small/healthy firms will be greater than small/unhealthy firms, and large/unhealthy firms will be greater than large/healthy firms (Hayes et al., 2001;

3 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Hunton et al., submitted for publication). In accordance with the hypotheses, we found a significant interaction between firm health and size. Support for these propositions sheds new light on the productivity paradox by suggesting that performance gains and losses resulting from increased IT expenditures may be masked in aggregate, but such effects could be more precisely identified when firms are discriminated along key dimensions. The next section provides theory and develops the study hypotheses. Section 3 presents the sample selection technique and research method. The final two sections analyze study results and discuss the research findings. 2. Hypotheses formulation 2.1. ERP systems and innovation While many studies have attempted to find a positive relationship between IT investments and firm performance, for the most part, research findings have yielded nonsignificant results (e.g., Weill, 1992; Mahmood and Mann, 1993; Hitt and Brynjolfsson, 1996). As a result, researchers have yet to provide compelling evidence that IT investments equate to measurable, positive value for business organizations. However, Dos Santos et al. (1993) suggested that a more refined analysis of IT investments could provide clarity in this regard. Dos Santos et al. (1993) argued that non-innovative technologies (those that maintain the status quo) are not likely to improve a firm s market value or financial performance, whereas innovative technologies (those that improve business processes) are expected to enhance value and performance. To empirically test their proposition, they observed the market response to IT investment announcements and found no overall effect; however, further subanalysis revealed that the market reacted positively to innovative investment announcements (Dos Santos et al., 1993). Similarly, Peffers and Dos Santos (1996) reported a positive relationship between innovative IT investments and firm performance. Hence, the lack of market and performance effects in prior studies may be due to a failure to discriminate between innovative and non-innovative investments. Drucker (1988) and Huber (1990) suggested that information technologies are considered innovative if they facilitate key business process improvements, such as 1. more accurate, comprehensive, timely, and available organizational intelligence from internal and external information sources at greatly reduced costs, 2. greater speed and accuracy in identifying problems and opportunities, 3. fewer intermediate human nodes within the organizational information-processing network, 4. reduced number of organizational levels involved in authorizing and making decisions, and 5. less time being consumed in the decision-making process. According to O Leary (2000), ERP systems are designed to support business process improvements of this nature, thereby enhancing information quality, decision making and

4 168 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) firm performance. In concept, the realization of such business process improvements is facilitated by ERP systems due to the integrated nature of enterprise-wide information via relational databases. To the extent that such improvements are not realized, much of the responsibility rests with inadequate planning and implementation of the ERP system, as the technology is capable of supporting the innovative process improvements mentioned above. Hence, while we recognize that some ERP system implementations have not coincided with business process improvements, we nevertheless categorize an ERP system as an innovative technology based on its potential in this regard. Accordingly, we expected to find significant positive effects of ERP adoption on market value and firm performance. Research evidence offered by Hayes et al. (2001) reinforces the notion that ERP systems are perceived to be innovative IT investments, as they observed a positive reaction from investors when firms announced that they were planning to implement an ERP system. In a similar study by Hunton et al. (submitted for publication), financial analysts significantly increased mean earnings forecast revisions when they learned that a firm was planning to implement an ERP system. The positive reaction of capital market participants to ERP adoption announcements reflects initial beliefs about the potential impact of ERP systems on future firm performance. However, whether ERP systems positively impact performance in the long run remains largely unanswered ERP systems and firm performance A recent study by Poston and Grabski (2001) investigated the impact of ERP system implementation on firm performance. They examined the post-implementation performance of 50 ERP-adopting firms over a 3-year post-implementation time horizon, after controlling for pre-implementation performance. They found no significant improvement in residual income (net operating income less imputed interest for cost of capital) or in the ratio of selling, general and administrative expenses to revenue throughout the 3-year window. However, they reported a significant decrease in the ratio of employees to revenue in each of the 3 years and a significant improvement in the ratio of cost of goods sold to revenue in year 3. Overall, they noted that ERP firms exhibited efficiency gains in some areas, but increased costs elsewhere seemed to offset such gains. Other researchers have also indicated little or no relation between IT investment and financial performance, which is often referred to as the productivity paradox (Harris, 1994). However, as suggested by Dos Santos et al. (1993), delineating between innovative and noninnovative uses of IT could offer clarity in this regard. How then could it be that Poston and Grabski (2001) examined the performance of companies that adopted an innovative IT investment (ERP system), yet found no significant gain in financial performance? While there are likely many answers to this question, one possible explanation suggested by Hitt and Brynjolfsson (1996) is that any financial gain associated with ERP adoption is passed through to customers in the form of lower prices. Robertson and Gatignon (1986) offered a similar explanation when they examined the impact of competitive factors on innovative technology diffusion. Through analytic modeling, Eliashberg and Jeuland (1986) discussed and Eliashberg and Chatterjee (1985, 1986) demonstrated that prices drop

5 immediately after the adoption of innovative technologies and demand increases as a result of price sensitivity. They further indicated that the financial performance of adopters might or might not improve significantly, depending on a host of exogenous factors such as competitive intensity, industry heterogeneity, demand uncertainty, and adoption rate of competitor firms; nevertheless, the performance of nonadopters would be expected to deteriorate by comparison in a competitive marketplace. If we view ERP adoption through this lens, we would not necessarily expect to see pre- to post-adoption financial gains for ERP firms. Instead, we would anticipate the financial performance of nonadopters to decline relative to adopters. Hence, we offer the following hypothesis (alternate form): H1: Longitudinal financial performance of firms that have not adopted ERP systems will be significantly lower than ERP-adopting firms Financial performance indicators J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) One way to evaluate firm performance is through financial statement analysis, which uses traditional accounting measures that are based on relationships among financial statement items. In the current study, we used four measures of performance. The first measure, return on assets (ROA), is frequently used by researchers as a measure of firm performance (e.g., Balakrishnan et al., 1996; Barber and Lyon, 1996; Barua et al., 1995; Bharadwaj, 2000; Hitt and Brynjolfsson, 1996; Weill, 1992). Since ROA incorporates both firm profitability and efficiency (Skousen et al., 1998), it tends to be a useful overall performance indicator. We purposely focused on ROA because it has been proposed that the benefits of ERP systems include improved efficiency and profitability (Brakely, 1999; Schaeffer, 1996; Stein, 1998; Vaughan, 1996; Wah, 2000). The combined effects of profitability and efficiency represented by ROA can be separated into return on sales (ROS) and asset turnover (ATO) two secondary measures of performance used in the current study. ROS, represented as income per dollar of sales, is a measure of the firm s profitability or margin. ATO, represented by the sales generated per dollar of assets, is a measure of asset efficiency. The last performance measure used in this study, return on investment (ROI) is included as a check on the robustness of the results using ROA, and because it has been cited as a key performance measure in the ERP implementation literature (e.g., Mabert et al., 2000; Stedman, 1999; Stein, 1998). We next examine the interaction of firm size and financial health on the financial performance of firms that adopt ERP systems Interaction of firm size and health: large firm effect The adoption of an ERP system by small firms is a significant undertaking, particularly with respect to the consumption of financial resources. For instance, Mabert et al. (2000) found that implementation costs, as a percent of revenue, range from 0.82% for very large firms (revenues > $5000 million) to 13.65% for very small firms (revenues <$50 million).

6 170 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Some small firms reported that total ERP implementation costs were nearly half of their annual revenues. Based on this analysis, Mabert et al. (2000, p. 56) stated...although larger firms incur a greater implementation cost because of their operating size, a comparison of revenues shows that there are economies of scale working in favor of the larger firms...[t]he move to ERP by smaller firms represents a major commitment in time and money. Additionally, ERP implementations can span a considerable amount of time. For instance, implementation periods from 1 to 2 years are common, and some implementation periods can stretch to 4 or 5 years (Cooke and Peterson, 1998; Mabert et al., 2000). Further, considerably large cost overruns are not unusual throughout extended ERP implementation periods. Overall firm performance can also decrease during implementation, due to temporary coordination and control disruptions (Koch, 1996; Wah, 2000). Thus, it can take several years to achieve a positive payback and desired ROI (Davenport, 2000; Stedman, 1999). Asa result, a firm must have sufficient resources, or access to resources, to withstand the strain of ERP implementation. While firm size is a key contextual factor to consider when postulating the impact of ERP on firm performance, financial health is also important since healthiness can affect a company s access to additional resources from external stakeholders, such as creditors and investors (Hayn, 1995). In general, large firms possess higher levels of resources and hold greater ability to attract additional resources than small firms. Thus, large firms can more easily absorb and withstand ERP implementation costs. However, research evidence offered by Khurana and Lippincott (2000) indicated that potential performance improvements are greater for relatively unhealthy, as compared to healthy, large firms. Consider a large financially healthy company that is efficient (e.g., high employee productivity and low cost producer) and effective (e.g., large market share and high-quality products) in its industry. Expected performance gains due to ERP adoption for this firm are less than anticipated gains of a relatively unhealthy counterpart, as the latter has greater potential for further improvement. Hayes et al. (2001) and Hunton et al. (submitted for publication) offered similar arguments with respect to the reaction of capital market participants to large firm ERP implementation announcements. Hayes et al. (2001) reported that the market reacted more strongly to ERP implementation announcements from large/unhealthy firms (standardized cumulative abnormal returns=0.272), as compared to large/healthy firms (standardized cumulative abnormal returns=0.061); however, the difference was not significant due, in part, to the low power of the small sample. Hunton et al. (submitted for publication) indicated that the mean earnings forecast revision (for 2001) was greater for large/unhealthy firms ($0.18) than large/healthy firms ($0.06). While they did not report the significance of this comparison, we analyzed their data and found that the difference was significant (t=3.01, P=.02). Based on extant theory and research evidence, for large firms, we expect a negative relationship between firm health and performance due to ERP adoption, as posited below (alternate form): H2a: For relatively large ERP-adopting firms, there will be a significant negative association between firm health and performance.

7 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Interaction of firm size and health: small firm effect Small firms also stand to benefit from ERP adoption, particularly if the firms use ERP systems to facilitate business process improvements. In this manner, small/healthy firms can become larger, and small/unhealthy firms can become healthier. With respect to relative health, on the surface, one would expect the logic just presented to hold for small firms, thereby suggesting that ERP-related performance gains should increase as financial health declines. However, small firms have greater variability than large firms with respect to liquidity and solvency (Huff et al., 1999). Such variability, which equates to increased loan and investment risk, coupled with an economy of scale disadvantage, means that small/ unhealthy firms may run into considerable obstacles with regard to acquiring needed financial and intellectual resources to complete full ERP integration. This situation places small/healthy firms in a better position to have or obtain implementation resources than small/unhealthy firms. Another issue facing small/unhealthy firms is that they might be forced into a partial implementation of the ERP system (due to cost considerations and resource acquisition constraints), thereby foregoing the acclaimed innovative benefit of using a fully integrated system. Hayes et al. (2001) and Hunton et al. (submitted for publication) also posited that capital market participants would react more positively to ERP announcements from small/healthy firms, as compared to small/unhealthy firms. In support of this proposition, Hayes et al. (2001) indicated a significantly stronger market reaction to ERP implementation announcements from small/healthy firms (standardized cumulative abnormal returns=0.283) than small/unhealthy firms (standardized cumulative abnormal returns= 0.270), and Hunton et al. (submitted for publication) indicated that the mean earnings forecast revision (for 2001) was significantly greater for small/healthy firms ($0.08) than small/unhealthy firms ($0.02). Hence, we posit that firm performance will be positively related to firm health for small ERPadopting firms (alternate form): H2b: For relatively small ERP-adopting firms, there will be a significant positive association between financial health and performance. 3. Sample selection and method 3.1. Sample selection We obtained data on ERP-adopting firms from Hayes et al. (2001) 1 and information concerning firm performance from Compustat. Since it may take several years for the benefits of ERP adoption to accrue, we constrained our sample to include ERP announcing firms for 1 Hayes et al. (2001) included in their sample all firms announcing ERP implementation in the time period January 1, 1990 through December 31, Announcing firms were identified through Lexis/Nexis Academic Universe s (News) Wire Service Reports.

8 172 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) which at least 3 years of financial data were available on Compustat subsequent to the ERP adoption announcement. As a result of this constraint, the sample included no ERP adoption announcements after For the inter-temporal comparisons, 3 years of pre-announcement performance information and 3 years of post-announcement performance information had to be available on Compustat. As a result of the constraints, the final sample resulted in 63 firms that announced ERP adoption. For tests of hypothesis 2 (H2a and H2b), we lost an additional three observations due to missing information on financial health (z-score). In our tests of the effect of ERP adoption on firm performance, we controlled for macroeconomic conditions that could influence test results. This was accomplished by matching ERP firms to a control sample of non-erp firms. 3 As has been done in prior studies of firm performance (e.g., Balakrishnan et al., 1996; Barber and Lyon, 1996), we matched firms on SIC code and firm size. The total sample of ERP and non-erp firms was 126 for the full sample (H1) and 120 for the reduced sample (H2a and H2b). To ensure that no ERP-adopting firms were included in the control sample, we conducted a telephone survey to determine whether the identified firms had indeed implemented an ERP system. A total of 21 of the 63 firms (33.33%) responded to the survey. With respect to the nonresponding firms, we searched through Lexis-Nexis and found that none of the control firms had a news wire disclosure concerning ERP adoption (see Hayes et al., 2001 for the search technique). Additionally, through Lexis-Nexis, we searched the SEC database and annual reports using the name of the firm and the following search string: ERP or enterprise resource planning or QAD or SAP R/3 or Oracle or Peoplesoft or J. D. Edwards or SSA or Baan or Geac or Lawson or Hyperion. With regard to foreign firms, we conducted an additional search of Lexis-Nexis using International Company Reports. One telephone survey respondent indicated ERP adoption. The Lexis-Nexis search resulted in three firms that possibly had adopted ERP. As a result, we selected four new control firms. The new control firms were subjected to the Lexis-Nexis search procedure to ensure no ERP disclosures, and none were found Measuring performance As mentioned earlier, we used four measures of performance (ROA, ROS, ATO, and ROI). Descriptions of the performance variables, along with descriptions of the metrics we used to determine firm size and health, are shown on Table 1. Performance was divided into two time periods pre-adoption and post-adoption. The fiscal year of the ERP announcement, identified as year zero (t 0 ), served as the baseline year for aligning the ERP-adopting and non-adopting firms. The pre-adoption period encompassed 3 years (t 3 to t 1 ), and the post-adoption period covered 3 years (t +1 to t +3 ). 2 The sample size for firms having 4 years of post-adoption information was too small to allow for testing. 3 When industry averages were calculated for the sample firms, it was found that the industry size varied considerably from the sample firms. In some instances, the industry average was much smaller than the sample firm, and in other instances, the industry average was much larger than the sample firm. As a result, a matched-pair design was used rather than an industry average control variable.

9 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Table 1 Description of variables Variable Description ROA Return on assets is income before extraordinary items (available for common stockholders), divided by the average of the beginning and ending total assets. Compustat then multiplies the ratio by 100 ROS Return on sales is income before extraordinary items (available for common stockholders), divided by net sales for the period ATO Asset turnover is net sales for the period, divided by the average of the beginning and ending total assets ROI Return on investment is income before extraordinary items (available for common stockholders), divided by the sum of total long-term debt, preferred stock, minority interest and total common equity. Compustat then multiplies the ratio by 100 Size The natural log of total assets Health Altman s z-score, as represented by the Compustat variable ZSCORE For H1, we tested for differences between pre- and post-adoption for ERP firms and non- ERP firms. Additionally, we conducted parametric and nonparametric tests and a regression analysis of performance differences between ERP and non-erp-adopting firms. The regression model allowed us to control for the firms pre-adoption performance. In the regression model, we regressed performance measures on pre-adoption financial performance and a dummy variable representing ERP versus non-erp adoption as follows: Financial Ratio ¼ a 0 þ a 1 Pre-Ratio þ a 2 Non-ERP Adoption þ e ð1þ where, Financial Ratio=post-adoption performance, as measured by average performance over the time period t +1 to t +3 for the ratios ROA, ROS, ATO, and ROI, Pre-Ratio=preadoption performance, as measured by average performance over the time period t 3 to t 1 for the ratios ROA, ROS, ATO, and ROI, Non-ERP Adoption=1 if the firm was a non-erp adopter, and 0 if an ERP adopter, e=error term. To be included in the sample, a firm must have at least one value in the time period t 3 to t 1. For firms with values in more than 1 year (t 3 to t 1 ), we averaged the values and used the average as the measure of the pre-ratio. The regression results helped to ensure that differences between ERP and non-erp-adopting firms were not due to an omitted variables problem. 4 To test hypothesis 2 (H2a and H2b), we regressed the financial ratios for ERP firms on the firms size, financial health, and the interaction of firm size and financial health, along with the control variable reflecting pre-adoption performance: Financial Ratio ¼ b 0 þ b 1 Pre-Ratio þ b 2 Size þ b 3 Health þ b 4 ðsize HealthÞþu ð2þ where, Financial Ratio=post-adoption performance, Pre-Ratio=pre-adoption performance, as measured by performance in the time period t 3 to t 1, Size=measured as the log of total 4 Barber and Lyon (1996) indicate that in addition to controlling for industry and size, it is important that preevent performance be controlled in models testing for abnormal performance. The Pre-Ratio variable used in the study represents such a lagged performance measure.

10 174 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) assets at time t 0, Health=financial health, measured using Altman s z-score at time t 0, and u=error term. Since we matched firms on size during the implementation year (t 0 ), it was selected as the most appropriate year for measuring the size variable in H2a and H2b. 5 For consistency, we also measured financial health at time t 0. However, if financial health was measured at time t 1, the results presented in the next section remain qualitatively the same Results 4.1. Descriptive statistics The ERP sample was distributed over 23 two-digit SIC classifications (see Table 2, panel A), indicating little clustering by industry. We matched 62 of the firms using at least a twodigit SIC classification, with one firm matched on a one-digit SIC classification. 7 As indicated by Table 2, panel B, over 58% of ERP implementation announcements occurred in This result is not surprising given the relatively recent development of ERP systems. 8 Table 3, panel A, indicates that, on average, ERP firms were slightly larger than non-erp firms when measured by assets or sales. However, we did not find a statistically significant difference when we measured assets or sales in either the year of the announcement (t 0 ) or the year preceding the announcement (t 1 ), as a was greater than.10 for both years. Table 3 also provides raw descriptive statistics for years t 3 through t 1 with respect to the performance measures (ROA, ROS, ATO, and ROI). While the raw measures provided on Table 3 are not used in the study tests, they do provide an indication that there was no trend in performance measures prior to ERP adoption and that the measures were skewed due to extreme observations. To reduce the influence of extreme performance measure observations, we restated outliers to the 5th and 95th percentiles for the pre- and post-performance 3-year distributions used in the subsequent tests. 9 5 An analysis was also conducted using the log of sales as the size variable. The levels of significance for the ERP firms remained unchanged using sales as the size variable. For the non-erp firms, none of the inferences reported using total assets were changed. 6 When using the z-score at t 1, the interaction term for ROI was not significant at a< Baber and Lyon (1996) indicate that the power of the tests is not significantly affected by the use of a fourdigit match versus a two-digit match. 8 All tests reported in this study were rerun controlling for event clustering in While results were slightly stronger after controlling for the clustering, the reported inferences remain unchanged. 9 Restating to the 95th and 5th percentiles is a procedure followed by Balakrishnan et al. (1996). Restatement is recommended by Barber and Lyon (1996) as a method of reducing conservatism of the parametric t statistic. That is, resetting the outliers reduces the differences between the parametric and nonparametric statistics. Tests were rerun using the 1st and 99th percentiles. The highly nonnormal distributions at the 1st and 99th percentiles greatly increased the conservatism of the parametric statistics.

11 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Table 2 Sample characteristics Panel A: SIC classification of ERP sample firms Two-digit SIC Full sample firms (H1) Total Panel B: Year in which ERP sample firm announcements occurred Announcement year Number of firms for the full sample Reduced sample firms (H2a and H2b) a Number of firms for the reduced sample Total a For analysis of hypothesis 2(H2a and H2b), three observations are deleted due to missing z-scores Hypothesis 1 The results of the tests of H1 are reported on Table 4. The last column of Table 4 presents the difference in pre- and post-adoption performance. As expected, no significant difference occurred between pre- and post-performance for ERP firms. However, over the same period, non-erp firms experienced a significant decline in ROA (t=2.239; one-sided P=.014), ROI

12 176 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Table 3 Descriptive statistics for ERP firms and non-erp firms Panel A: Matching ERP firms and non-erp firms on firm size (n=63) Item Sample Mean Median Standard deviation t Statistic P value a Time=t 1 Total assets ERP adopter 10, ,448 nonadopter , Net sales ERP adopter ,967 nonadopter , Time=t 0 Total assets ERP adopter 11, ,888 nonadopter , Net sales ERP adopter 10, ,521 nonadopter , Panel B: Descriptive information concerning firm performance Time Item Sample Mean Median Standard n deviation t 3 ROA ERP adopter nonadopter ROI ERP adopter nonadopter ROS ERP adopter nonadopter ATO ERP adopter nonadopter t 2 ROA ERP adopter nonadopter ROI ERP adopter nonadopter ROS ERP adopter nonadopter ATO ERP adopter nonadopter t 1 ROA ERP adopter nonadopter ROI ERP adopter nonadopter ROS ERP adopter nonadopter ATO ERP adopter nonadopter a A nonparametric Wilcoxon test provides P values which are larger than those reported below, indicating no significant difference at (a<.10) between the ERP firms and non-erp firms.

13 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Pre vs. Post3 c Table 4 Performance comparison between and within ERP firms and non-erp firms a Panel A: Pre- and post-adoption means [medians] for ERP firms (n=63) Financial Pre Post1 Post2 Post3 ratio b ROA [4.637] [5.031] d [4.715] d d [4.692] d t=0.731; P=.468 ROS [0.050] [0.045] [0.053] [0.054] t=0.351; P=.727 ATO [1.041] [1.105] [1.091] [1.051] t=1.016; P=.314 ROI [8.614] [8.957] d d [8.411] e e [8.395] e t=0.518; P=.606 Panel B: Pre- and post-adoption means [medians] for non-erp firms (n=63) Financial ratio Pre Post1 Post2 Post3 Pre vs. Post3 f ROA [5.280] [3.725] [3.151] [3.501] t=2.239; P=.014 ROS [0.047] [0.043] [0.045] [0.045] t=1.018; P=.157 ATO [1.018] [0.959] [0.936] [0.989] t=2.976; P=.002 ROI [7.861] [5.462] [5.153] [5.731] t=2.397; P=.010 a The Pre information represents the average for the time periods 3 through 1. The Post information reflects performance in year +1 for Post1, the average of performance in years +1 and +2 for Post2, and the average of performance in years +1 through +3 for Post3. Due to non-normality, means for the Pre sample and the Post samples were winsorized at the 5th and 95th percentile. b For a description of the financial ratios, see Table 1. c P values are for two-sided tests of the mean difference between Pre performance and Post3 performance. A two-sided test is used since we made no prediction concerning direction of performance for ERP firms. d Indicates that the ERP firm ratio is significantly larger than the non-erp firm ratio using a one-sided P<.10 based on t tests of means, or Wilcoxon tests of medians. e Indicates that the ERP firm ratio is significantly larger than the non-erp firm ratio using a one-sided P <.05 based on t tests of means, or Wilcoxon tests of medians. f P values are for one-sided tests of the mean difference between Pre performance and Post3 performance. A one-sided test is used since we anticipated a significant decline in performance for non-erp firms. (t=2.397; one-sided P=.010), and ATO (t=2.976; one-sided P=.002). While non-erp firm ROS also experienced a noticeable decline, it was not significant at conventional levels. 10 A test (not shown) comparing pre-adoption performance to base year (t 0 ) performance indicated a moderately significant (t=1.680; P=.098) increase in ROS for ERP firms and a significant (t=2.473; one-sided P=.008) decrease in ATO for non-erp firms. No other performance differences were significant at conventional levels. The results indicate that subsequent to ERP adoption, ERP firms performance was relatively unchanged, while non-erp firms performance significantly declined. Table 4 also shows that performance of ERP firms differed little from non-erp firms until two years after adoption, when non-erp firms performed significantly ( P<.10) worse than ERP firms on ROA and ROI. 3 years after adoption, non-erp firms also performed 10 A Wilcoxon matched-pairs test provided qualitatively similar results; however, the results were not as strong as those provided by the t test.

14 178 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Table 5 Regression results for performance variables, controlling for pre-adoption results (n=126) Financial ratio for third year (t +3 )=Intercept+Pre-Ratio+Non-ERP adopter a Financial ratio Intercept Pre-Ratio Non-ERP adopter Adjusted R 2 ROA (3.82) [<.001] (4.90) [<.001] ( 1.49) [.069].161 ROS (3.37) [.001] (6.58) [<.001] ( 0.99) [.162].252 ATO (3.70) [<.001] (24.98) [<.001] ( 1.92) [.029].834 ROI (4.27) [<.001] (3.60) [<.001] ( 1.82) [.036].103 a For each independent variable, the coefficient value, (t statistic) and [ P values] are provided. The P values for the Non-ERP Adopter variable are one-sided, all others are two-sided. The Pre-ratio reflects the average of the financial ratio from time period t 3 through t 1. Non-ERP Adopter is 1 if the firm is not an ERP adopter and 0 if the firm is an ERP adopter. The dependent variable and the pre-financial variable are winsorized. Cook s D indicates no influential observations that could be unduly affecting results. Variable descriptions are provided in Table 1. significantly worse than ERP firms on ROA ( P<.10) and ROI ( P<.05). Although ROS appeared to remain stable for ERP firms while decreasing for non-erp firms, the difference was not significant. Additionally, there was no significant difference in ATO between the ERP and non-erp firms, although ATO did decline between the pre- and post-adoption periods for non-erp firms relative to ERP firms. The results reported on Table 4 suggest that performance benefits accruing from ERP adoption may take several years to realize. Therefore, the remainder of the tests focused on firm performance 3 years after adoption. The results on Table 4 lend partial support to H1, as non-erp firms experienced a significant decline in performance; however, when compared to ERP firms, the declines were only significantly lower for ROA and ROI. As an additional test of H1, we regressed performance in the third year after adoption on ERP firms versus non-erp firms, controlling for pre-adoption performance. 11 For ROA and ROI, the results were essentially unchanged after controlling for pre-adoption performance (see Table 5), as ROA (t= 1.49; P=.069) and ROI (t= 1.82; P=.036) remained significantly lower for non-erp-adopting firms. In addition, after controlling for pre-adoption performance, ATO was significantly lower (t= 1.92; P=.029) for non-erp firms than ERP firms, indicating that there may be efficiency benefits associated with ERP adoption. 12 Overall, test results partially supported the first hypothesis Hypothesis 2 We also hypothesized that ERP firm performance may be influenced by the interaction of firm size and financial health. Test results (Table 6) indicated a generally positive association 11 Regression allows for control of pre-adoption performance, which Barber and Lyon (1996) indicate is necessary to minimize misspecification of test statistics. 12 The large adjusted R 2 for the ATO model was related to the high correlation (.90) between the ATO variable and the Pre-Ratio. An analysis indicated that the Pre-Ratio explained.79 of the variance associated with ATO. 13 Regression results for performance in year +1 and year +2 (not shown) substantially support the results found on Table 4.

15 Table 6 Within ERP adopters regressions related to the effect of firm size and health on financial performance (n=60) a Financial ratio ROA Intercept Pre-Ratio Size Health Size*health Adjusted R ( 3.18) [.002] ROS ( 3.67) [<.001] ATO (0.63) [.528] ROI ( 3.20) [.002] J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) (2.86) [.006] (3.38) [.001] (19.06) [<.001] (2.04) [.046] (3.68) [<.001] (4.19) [<.001] (0.50) [.619] (3.83) [<.001] (3.16) [.003] (3.77) [<.001] (0.18) [.860] (3.14) [.003] ( 2.24) [.029] ( 2.74) [.008] ( 0.57) [.573] ( 2.31) [.025] a For each independent variable the coefficient value, (t statistic) and [two-sided P values] are provided. Pre- Ratio is the average of the financial ratio from the time periods 3 through 1. The Financial Ratio is the average of the ratio from time periods +1 through +3. Variable definitions are provided in Table 1. The dependent variables and Pre-Ratio variables have been winsorized between ERP firm performance and pre-ratio (control variable), firm size, and financial health. Firm size significantly and positively affected performance as measured by ROA (t=3.68; P<.001), ROI (t=3.83; P<.001), and ROS (t=4.19; P<.001). Additionally, firm health significantly and positively influenced ROA (t=3.16; P=.003), ROI (t=3.14; P=.003), and ROS (t=3.77; P<.001). Neither size nor health significantly affected the ATO of ERP adopters. As expected, there was a significant interaction between firm size and financial health relative to ROA (t= 2.24; P=.029), ROI (t= 2.31; P=.025), and ROS (t= 2.74; P=.008) for ERP adopters. Moderated regression analysis (Sharma et al., 1981) indicated that the presence of the interaction term significantly (a=.05) increased the explanatory power of the ROA, ROI, and ROS regressions. Interpretation of the interaction term indicates that as firm size and financial health increased, there was a decrease in firm performance. However, to better illustrate the precise nature of the interaction, we performed a subgroup analysis on those models from Table 6 with significant interactions (Sharma et al., 1981). To conduct the analysis, we first divided the sample at the mean into large and small firms. 14 Then, we regressed performance measures on the control variable and financial health (see Table 7). Panel A of Table 7 indicates that for larger firms, financial health significantly and negatively affected ROI (t= 1.67; one-sided P=.059). While financial health also negatively affected ROA and ROS, the results were not significant at conventional levels. The subgroup analysis for large firms lends partial support to the overall interaction found on Table 6 and H2a. Panel B of Table 7 indicates that for small firms, financial health 14 Since we had no basis for determining how to define relative firm size, selection of the mean was arbitrary. However, analyses were also conducted using the median of total assets and by dividing the sample into one-third large and two-thirds small firms. As the analysis moved from the mean (about 1/4 of the sample reflects large firms) to the median, the association between health and performance for large firms became less negative and increasingly insignificant. Small firm analysis was unchanged.

16 180 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Table 7 An analysis of firm size a and financial health b for ERP adopters Panel A: large ERP adopters (n=16) c Financial ratio Intercept Pre-Ratio ( ) d Health Adjusted R 2 ROA (3.63) [.003] (4.15) [.001] ( 0.99) [.171].801 ROS (1.22) [.243] (9.25) [<.001] ( 0.96) [.177].901 ROI (4.26) [.001] (5.12) [.000] ( 1.67) [.059].786 Panel B: Small ERP adopters (n=44) c Financial ratio Intercept Pre-Ratio (+) d Health Adjusted R 2 ROA (1.01) [.320] (1.98) [.055] (2.72) [.005].374 ROS ( 0.29) [.777] (2.86) [.007] (2.89) [.003].419 ROI (1.45) [.154] (1.17) [.247] (2.50) [.008].227 a Firms are divided into small and large based on the mean of total assets. b Variable descriptions are provided in Table 1. c Mean (standard deviation) [ P value]. d Hypothesized direction of results. significantly and positively impacted ROA (t=2.72; one-sided P=.005), ROI (t=2.50; onesided P=.008) and ROS (t=2.89; one-sided P=.003). The result supports H2b, indicating that for smaller ERP firms, as financial health improved, so did financial performance Post hoc analyses While Hypotheses 2a and 2b focus on how firm size and financial health affect the performance of ERP firms, no hypotheses concerning size and health were provided for non- ERP firms. However, we did conduct non-erp adopter tests (not reported) to ensure that the results obtained for ERP adopters were associated with ERP adoption and not a spurious correlation. We found no significant positive association between performance and firm size, firm health, or the interaction of firm health and firm size. The lack of significant associations in the non-erp firms lends some support to the arguments made in hypotheses H2a and H2b. While Altman s z-score has been used in prior studies (e.g., Barron et al., 1999; Hayes et al., 2001; Miller and Skinner, 1998) as a proxy for financial health, there is no theoretical basis for defining financial health. Therefore, to test the robustness of our results using z- scores as a proxy of financial health, a second financial health variable was tested. In the second set of tests, we defined financial health as the difference between income before extraordinary items at year zero (the year of implementation) minus the average of income before extraordinary items in the 3 years prior to year zero (t 3 through t 1 ). This second variable yielded levels of significance comparable to those using z-scores for the health and interaction variables. However, using this alternative metric for financial health, there was no longer a significant (a<.10) main effect for size in the models (ROA, ROI, ROS) reported on

17 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Table 6. Since we did not hypothesize a main effect for size, the results of our analyses were not affected. Because small samples can be sensitive to departures from regression assumptions, we conducted several tests to ensure the robustness of the reported results. To test for heteroscedasticity, we used White s (1980) test. Results indicated that heteroscedasticity was a problem for two of the regression equations examined on Table 6. White s correction 15 yielded essentially the same levels of significance as those reported earlier; therefore, the uncorrected statistics were reported for all regression equations. We checked to determine if the ERP vendor was significantly associated with firm financial performance. Vendors were divided into two groups SAP and Peoplesoft versus all other vendors (in accordance with Hayes et al., 2001). Including a vendor control variable slightly increased the test statistics reported for the interaction terms in Table 6; however, the overall inferences remained unchanged. Cook s D test revealed two potentially influential observations related to the ERP firm regressions reported on Table 6, panel A (Neter et al., 1990). Eliminating the two influential observations did not influence the explanatory power of the regressions, and the results were essentially the same. Finally, neither variance inflation factors (VIF) (Neter et al., 1990) nor multicollinearity diagnostics (Belsley et al., 1980) revealed that multicollinearity was significant among the pre-ratio, size, and health variables. 5. Discussion In this study, we investigated the longitudinal impact of ERP adoption on firm performance. To accomplish this objective, we compared the financial performance of 63 ERP adopters to 63 nonadopters in a matched-pair design. We examined the following performance metrics: ROA, ROS, ATO, and ROI. We predicted that firm performance would be greater for adopters than nonadopters, primarily because the financial performance of nonadopters would decline by comparison. Overall, research findings suggest that three (ROA, ATO, and ROI) of the four performance metrics support this hypothesis. More specifically, study results indicate that ROA, ROI, and ATO were significantly lower for nonadopters than adopters, the third year after ERP implementation. Additionally, the average 3-year ROA and ROI were significantly greater for adopters when compared to nonadopters. Subanalysis revealed that the performance metrics for adopters did not change significantly from pre- to post-adoption, but the metrics declined for nonadopters over the same time period. In addition, we investigated the interactive effect of firm size and financial health on the performance of ERP adopters. We found a significant interaction between size and health for three of the financial measures (ROA, ROI, and ROS). Our results suggest that large/ unhealthy adopters experience better ROI than large/healthy adopters. We also found that 15 White s correction adjusts the t statistic to account for misspecification due to heteroscedasticity. For additional information on the correction, see White (1980).

18 182 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) small/healthy firms that adopt ERP systems demonstrate better performance (ROA, ROI, and ROS) than small/unhealthy firms. Accordingly, the second hypothesis (H2a and H2b) was partially supported. Our significant findings are consistently related to two measures of firm performance ROA and ROI. We also examined two components of ROA (ROS and ATO) that are often described as measures of profitability (ROS) and efficiency (ATO). We found that ATO decreased significantly for non-erp adopters, and there was some indication that ROS decreased for nonadopters; however, the difference was not significant. Since ROA is computed by multiplying these two measures, study results indicate that improvements in firm performance are the result of the combined efficiency and profitability gains. With respect to pre- to post-erp adoption gains, our results are similar to Poston and Grabski (2001), in that both studies found limited evidence of efficiency gains, but little evidence of gains elsewhere. However, a comparison of ERP adopters and nonadopters sheds light on the productivity paradox, suggesting that financial gains arising from ERP adoption may be passed on to customers in the form of lower prices; hence, the performance of nonadopters declines by comparison. We recognize that we do not directly test whether ERP gains are being transferred to customers, as micro-level archival data of this nature is not available. Thus, future researchers might test this theory via survey or case data, as well as examine competing reasons that might explain why the relative performance of nonadopters declines when compared to ERP adopters. Our findings are also consistent with Hayes et al. (2001) and Hunton et al. (submitted for publication) in that potential ERP adopters should be aware of performance improvement limitations that may result from a combination of size and health. That is, large/unhealthy firms can expect greater performance gains than their large/healthy counterparts, as large/ unhealthy firms have more room for potential efficiency and effectiveness gains. Additionally, small/healthy firms can anticipate greater future benefit from ERP adoption than small/ unhealthy firms, primarily because small/healthy firms can likely acquire needed resources to complete full integration and thereby become larger players in the marketplace. Unfortunately, small/unhealthy firms might suboptimize the potential impact of ERP systems, as they may be forced into partial implementations due to resource constraints. As a result, business process innovation may not occur to the extent desired which can hamper further improvement in financial health. However, the disadvantage of small/unhealthy firms may dissipate in the near future, as ERP vendors are now making it more affordable for medium- and small-size firms to implement ERP systems by scaling back the complexity of their systems and assisting their clients in implementation activities. One relatively recent method of increasing affordability is the use of application service providers (ASPs), where an ASP firm offers to host and maintain the ERP system, and the service recipient pays a fee to use the system. This strategy minimizes the technical and financial burden on medium- and small-size firms, while offering to them the full range of ERP functionality. The spread of ERP systems across business organizations has implications for accounting researchers and professionals, since the systems automate a wide array of business processes and decrease the need for many accounting functions currently being performed by

19 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) accountants. As a result, new internal control and auditing questions arise regarding the unique risks inherent with ERP systems, as compared to legacy accounting systems. Accountants and auditors must also deal with ways to report financial information in a more timely manner than in the past (e.g., near-continuous financial reporting) and attest to the reliability of such information. Additionally, eliminating some of the currently performed accounting functions will eliminate many lower level accounting positions, while raising the sophistication of tasks performed by upper level accountants. For instance, degreed accountants can be expected to perform deeper financial analyses, provide more value-added advice regarding business process improvements, develop internal controls commensurate with ERP risks, and make tactical and strategic decisions aimed at improving firm performance. These issues, and more, provide the backdrop for investigating a host of interesting and relevant issues related to ERP adoption. References Balakrishnan R, Linsmeier TJ, Venkatachalam M. Financial benefits from JIT adoption: effects of customer concentration and cost structure. Account Rev 1996;1(2): Barber BM, Lyon JD. Detecting abnormal operating performance: the empirical power and specification of test statistics. J Financ Econ 1996;41(3): Barron OE, Kile CO, O Keefe TB. MD&A quality as measured by the SEC and analysts earnings forecasts. Contemp Account Res 1999;16(1): Barua A, Kriebel CH, Mukhopadhyay T. Information technologies and business value: an analytic and empirical investigation. Inf Syst Res 1995;6(1):3 23. Belsley DA, Kuh E, Welsch RE. Regression diagnostics: identifying influential data and sources of collinearity. New York: Wiley; Bharadwaj AS. A resource-based perspective on information technology capability and firm performance: an empirical investigation. MIS Quart 2000;24(1): Brakely HH. What makes ERP effective? Manuf Syst 1999;17(3):120+. Cooke DP, Peterson WJ. SAP implementation: strategies and results. New York (NY): The Conference Board; Davenport T. Putting the enterprise into the enterprise system. Harv Bus Rev 2000;76(4): Dos Santos B, Peffers K, Mauer DC. The impact of information technology investment announcements on the market. Inf Syst Res 1993;4:1 23. Drucker PF. The coming of the new organization. Harv Bus Rev 1988;66(1): Eliashberg J, Chatterjee R. Analytical models of competition with implications for marketing issues, findings, and outlook. J Mark Res 1985;22(August): Eliashberg J, Chatterjee R. Stochastic issues in modeling the innovation diffusion process. In: Mahajan V, Wind Y, editors. Innovation diffusion models of new product acceptance. Cambridge (MA): Ballinger Press; p Eliashberg J, Jeuland AP. The impact of competitive entry in a developing market upon dynamic pricing strategies. Mark Sci 1986;5(1): Grover V, Teng J, Segars AH, Fielder K. The influence of information technology diffusion and business process change on perceived productivity: the IS executive s perspective. Inf Manage 1998;34(3): Harris DH. Organizational linkages: understanding the productivity paradox. Washington (DC): National Academy Press; Hayes DC, Hunton JE, Reck JL. Market reaction to ERP implementation announcements. J Inf Syst 2001; 15(1):3 18.

20 184 J.E. Hunton et al. / Int. J. Account. Inf. Syst. 4 (2003) Hayn C. The information content of losses. J Account Econ 1995;20: Hitt LM, Brynjolfsson E. Productivity, business profitability, and consumer surplus: three different measures of information technology value. MIS Q 1996;20(2): Huber GP. A theory of the effects of advanced information technologies on organizational design, intelligence, and decision making. Acad Manage Rev 1990;15(1): Huff PL, Harper Jr RM, Eikner AE. Are there differences in liquidity and solvency measures based on company size? Am Bus Rev 1999;17(2): Hunton JE, McEwen RA, Wier B. Analysts reactions to ERP announcements. J Inf Syst [submitted for publication]. Khurana I, Lippincott B. Restructuring and firm value: the effects of firm profitability and restructuring purpose. J Bus Finance Account 2000;27(9): Koch C. The integration nightmare: sounding the alarm. CIO Mag 1996;(November 15):6 10. Mabert VA, Soni A, Venketaramanan MA. Enterprise resource planning survey of U.S. manufacturing firms. Prod Inventory Manage J 2000;52 58 [second quarter]. Mahmood MA, Mann GJ. Measuring the organizational impact of information technology investment: an exploratory study. J Manage Inf Syst 1993;10(1): Miller GS, Skinner DJ. Determinants of the valuation allowance for deferred tax assets under SFAS No Account Rev 1998;73(2): Neter J, Wasserman W, Kutner MH. Applied linear statistical models. Homewood (IL): Irwin; O Leary D. Enterprise resource planning systems: systems, life cycle, electronic commerce, and risk. Cambridge (MA): Cambridge Univ Press; Peffers K, Dos Santos BL. Performance effects of innovative IT applications over time. IEEE Trans Eng Manage 1996;43(4): Pinsonneault A. Information technology and the nature of managerial work: from productivity paradox to the Icarus Paradox. MIS Q 1998;22(3): Poston R, Grabski S. Financial impacts of enterprise resource planning implementations. Int J Account Inf Syst 2001;2: Robertson TS, Gatignon H. Competitive effects on technology diffusion. J Mark 1986;50(3):1 12. Schaeffer C. Performance measurement. IIE Solut 1996;28(3):20 7. Sharma S, Durand RM, Gur-Arie O. Identification and analysis of moderator variables. J Mark Res 1981; 18(August): Skousen KF, Stice EK, Stice JD. Intermediate Accounting. 13th ed. Cincinnati (OH): South-Western College Publishing; Stedman C. Survey: ERP costs more than measurable ROI. Computerworld 1999;33(14):6. Stein T. Drinking in enterprise resource returns. Informationweek 1998;679(Apr 27):72 5. Vaughan J. Enterprise applications. Software Mag 1996;16(5): Wah L. Give ERP a chance. Manage Rev Mar 2000;20 4. Weill P. The relationship between investment in information technology and firm performance: a study of the valve manufacturing sector. Inf Syst Res 1992;3(4): White H. A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica 1980;48: