The Joint Influence of Evaluation Mode and Benchmark Signal on Environmental Accounting-Relevant Decisions

Abstract

The assessment of environmental alternatives occurs in one of two evaluation modes: joint (JE) or separate (SE) evaluation. This study explores the combined influence of evaluation mode and the attractiveness of environmental alternatives on decisions based on non-financial environmental accounting information. General Evaluability Theory [GET; Hsee, C. K., and J. A. Zhang. 2010. “General Evaluability Theory.” Perspectives on Psychological Science 5: 343–355] predicts greater decision dependence on benchmark performance signals received in SE than JE mode. However, GET is silent on the influence of the alternatives’ attractiveness on evaluations, e.g. when available environmental alternatives all perform either superior or inferior to a benchmark case. To address this condition, we propose supplementing GET with the ‘negativity bias’, which predicts that negative signals (e.g. worse than benchmark values) receive more decision weight compared to positive signals (e.g. better than benchmark values) [Rozin, P., and E. B. Royzman. 2001. “Negativity Bias, Negativity Dominance, and Contagion.” Personality and Social Psychology Review 5 (4): 296–320]. Accordingly, this study's experiment (n = 77) manipulated evaluation mode (between-participants; JE/SE) and the alternatives’ performances relative to a benchmark (within-participants; available alternatives all better, or worse, than benchmark). Participants made investment allocations (as a proxy for performance ratings) to available factories based on factory environmental accounting performance. Participants invested less (more) in factories when factory performance was inferior (superior) to benchmarks. However, consistent with a combined GET and negativity bias prediction, this difference was more (less) pronounced in SE (JE) mode. Overall, results suggest that joining GET with a negativity bias accurately captures the joint influences of evaluation mode and benchmark signals on decisions based upon environmental accounting information. Further, decision-makers seem to adopt an asymmetric decision heuristic in evaluating environmental accounting information. Specifically, they avoid increasing decision weight on bad environmental performance information in JE compared to SE mode, but decision weights are indifferent across evaluation mode for good environmental performance information.

Acknowledgements

We thank David Atwood, Glenn Blomquist, Kim Ikuta, Sean Peffer, Bob Ramsay, and Milton Shen for comments and analyses of earlier drafts, as well as reviewers and participants at the 2012 AAA Management Accounting Section Research Conference and the 2015 Accounting, Behavior and Organizations Section Midyear Meeting.

Notes

1. This assumes a decision-maker who is not a specialist in environmental issues (e.g. an environmental engineer), which is likely when measurements reflect non-traditional and unfamiliar environmental accounting information. Managers are not always familiar with the measures they analyse for evaluations. This unfamiliarity is particularly acute when managers do not participate in developing performance evaluation measures (Lipe and Salterio 2000).

2. Hsee and Zhang (2004) develop the distinction bias framework to explain consumer behavior. Specifically, they observe that consumers’ predictions made from the analysis of an alternative's values in JE mode overestimate the difference between the alternatives compared to when the alternative is actually experienced in SE mode by a consumer. This overestimate occurs when the alternatives are quantitatively different; when the alternatives’ values are qualitatively different, there is enough context to reduce the extent of this overestimate. The current study analyses quantitative differences among alternatives differently from Hsee and Zhang (2004), and uses a different decision setting; the current study does not focus on consumer comparisons of their consumption predictions in JE mode with actual consumption outcomes in SE mode. In the present setting, comparisons between evaluation modes in this study are based on predictions that cannot be readily realised by personal experiences because factories are not ‘consumed' in the same sense as consumer products. In other words, all evaluations in this study rely on the gathering and cognitive processing of information that is not experienced personally (e.g. analysis of a factory's performance as opposed to analysing personally derived pleasure from viewing a state-of-the-art television). Hence, the decision choices of the present study are linked to, and better reflect, managerial and environmental accounting decision settings.

3. The theoretical framework suggests that similar results should obtain regardless of whether some or all of the environmental evaluative measures contain benchmark data. This study proceeds with the more realistic practitioner setting of only some benchmark data available for environmental accounting information. The supplemental analysis section will address a setting with complete benchmark information availability. The experiment was programmed in z-Tree (Fischbacher 2007).

4. For example, participant instructions for the measure ‘Annual energy savings', ‘The more kilowatt hours of energy saved each year, the better the factory's performance'. One may claim that this is a form of evaluability information, but this information serves as an experimental control to avoid participant misinterpretations of the correlation between the measures’ values and factory performance. If this confound remained, the manipulation would be less effective, potentially leading to anomalous or no results.

5. The willingness-to-invest measure is a form of the WTP measures, which are often employed in marketing and psychology settings. WTP measures help in analyses involving environmental accounting information (Epstein 2008). See Breidert, Hahsler, and Reutterer (2006) for more discussion of this measure and its applications in marketing research.

6. For example, a participant viewing data sets in the order of data sets 1, 2, 3, and 4 would view one factory alternative from each data set before viewing the second factory alternative from data set 1, data set 2, etc.

7. Experimental instructions stated that there would be ‘multiple drawings for a chance to win gift cards worth $25 each'. Actual odds of winning were about 3.3% for each participant.

8. There were few cases of statistically significant differences: there was an interaction effect between gender and level of evaluability information (p = .019) and between the number of current accounting classes taken and evaluation mode (p = .045), yet no interaction effect between mode and completed accounting classes. Factor analysis grouped some of the environmental attitudes measures into one component, which interacted with level of evaluability information (p = .034), but assessed environmental attitudes did not vary across conditions (p = .255). We found no meaningful relationships from these results.

9. Similar to Hsee and Leclerc (1998), Willemsen and Keren's (2004) attributes were not nearly as complex as the attributes used in this study; examples included the size of a college dorm room and commuting time from the room to the campus. Another difference to note between the two studies is that this study's alternatives performed either better or worse (but not both) than a reference value. Willemsen and Keren (2004) tested multiple performance measures which contained both types of performance signals at the same time.

Additional information

Funding

For financial support, we thank the Von Allmen School of Accountancy and the Gatton College of Business & Economics at the University of Kentucky, and the University of Alabama in Huntsville.