EXPERIMENTAL RESEARCH ON MANAGEMENT ACCOUNTING

1
EXPERIMENTAL RESEARCH ON MANAGEMENT ACCOUNTING

• Michael D. Shields
• Michigan State University
• 2005

2
OUTLINE

• EXAMPLES
• INTRODUCTION
• IMPORTANT FEATURES OF MA EXPERIMENTS
• READING LIST

3
1. EXAMPLES

• Most experiments in MA test hypotheses from
  economics and/or psychology about individual or
  small group cognition (judgments, decisions)
  and/or motivation (effort, learning,
  performance).
• Some experimental studies investigate issues
  related to bargaining/negotiating, cross-cultural
  psychology, and personality.
• Next slides provide summaries of five studies.

4
Chow (1983)

• Do budgets and compensation contracts influence
  individuals choice of contracts and performance?
• Innovative study that provided the foundation for
  many experimental studies on incentive
  contracting, participative budgeting, performance
  measurement, and management audits (see Young and
  Lewis 1995).
• Independent variables
• Budget difficulty average or difficult
• Compensation contract fixed pay, piece rate,
  budget based
• Budget setting compensation contract and budget
  difficulty is self selected or imposed
• Dependent variables
• Choice of compensation contract and individual
  performance

5

• H1 Result When budgets are imposed, budget
  difficulty and compensation contract have
  additive effects on performance, controlling for
  performance capability.
• Highest performance with difficult budget and
  piece rate, and lowest performance with average
  budget and fixed pay.
• Result consistent with goal setting theory
  (additive) but not agency theory (interaction).
• H2 Result When individuals select their own
  compensation contract (fixed pay or budget
  based), they select based on their skill.
• Individuals with higher skill select contracts
  that have more budget-based incentives,
  consistent with agency theory.

6
Luft (1994)

• Does the choice of compensation contract depend
  on whether incentives are labeled as a bonus or
  penalty?
• 80 base 20 bonus if P gt B or 80 base if P lt
  B
• 100 base if P gt B or 100 base - 20 penalty if
  P lt B
• Economic analysis treats both contracts as
  equivalent ( expected pay), but most people
  prefer the bonus contract.
• Economically, but not psychologically,
  equivalent.
• Non-monetary payoffs Bonus has connotations of
  approval and reward, which many people receive
  utility from.
• Prospect theory People realize more subjective
  disutility from a realized loss (penalty) than
  from a lost gain (bonus).
• Implicit contract Contract sets an implicit base
  pay level, which determines future base pay.
  People prefer higher base pay because it implies
  higher future base pay levels.

7

• H1 Employees will be more likely to choose an
  incentive contract with a given set of payoffs
  for performance relative to a budget when the
  incentive is labeled a bonus rather than a
  penalty.
• Result prediction
• Responses to post-experiment questions indicate
  that most people explain their choice by 1 or
  more of the 3 explanations.
• H2 When employees judgments about their past
  performance is dependent on their memory, their
  preference for a bonus over a penalty incentive
  contract increases with experience.
• People encode and retrieve from memory
  experiences that are more memorable (e.g., big
  / payoffs) because they create strong emotions.
  As people gain experience, they are more likely
  to remember their big positive emotional
  experiences (e.g., receiving a bonus), which then
  creates preferences for bonus over penalty
  incentive contracts.
• Result prediction.

8
Vera-Munoz (1998)

• Does the use of opportunity costs in
  resource-allocation decisions depend on
  individuals accounting knowledge and decision
  context, controlling for their analytical
  ability?
• Financial accounting focuses on sunk, not
  opportunity, costs. When in a business context,
  individuals with financial accounting knowledge
  will activate this knowledge and focus their
  attention on information associated with it.
• H1 Number opportunity costs ignored in making
  resource-allocation decisions in a business
  context increases with financial accounting
  knowledge.

9

• H2 Number opportunity costs ignored by
  individuals who have a high level of financial
  accounting knowledge when making resource
  allocation decisions is larger in a business than
  in a personal context.
• Results predictions.
• Study illustrates a negative consequence of
  having financial accounting knowledge when making
  business decisions.
• Curse of knowledge It is difficult not to use
  knowledge that is activated by context (cue)
  regardless of whether that knowledge is relevant
  to a decision.

10
Frederickson, Peffer and Pratt (1999)

• Does an individuals experience in making
  decisions under a performance-evaluation system
  influence how he or she evaluates others
  decisions?
• Experience under performance evaluations based on
  outcomes (decisions) will direct attention toward
  outcomes (decision inputs/processes), causing
  outcomes (decisions) to be salient in knowledge
  representations.
• Frequency of outcome feedback reinforces
  outcome-based knowledge representations.
• Under outcome-based evaluations, outcome feedback
  reinforces a causal link between outcomes and
  evaluations.
• Under decision-based evaluations, outcome
  feedback is of no use in assessing decision
  performance, reinforcing the absence of a causal
  link between outcomes and evaluations.

11

• H Influence of outcome feedback on performance
  evaluations is an interactive function of
  experience with decision- or outcome-based
  performance evaluations and outcome feedback
  frequency.
• Relative size of the expected influence of
  outcome feedback on evaluations, from smallest to
  largest
• Experience w/ decision evaluations and frequent
  outcome feedback
• Experience w/ decision evaluations and infrequent
  outcome feedback
• Experience w/ outcome evaluations and infrequent
  outcome feedback
• Experience w/ outcome evaluations and frequent
  outcome feedback.
• Dependent variable difference in the performance
  evaluation of a decision in which the decision
  outcome is reported to be gt or lt a performance
  standard.
• If decision outcomes do not affect performance
  evaluations, then there is no difference in
  evaluations as decision outcomes vary.
• Results support hypothesis.

12
Luft and Shields (2001)

• Does an individuals subjective prediction
  performance about future profits arising from a
  current-period intangibles expenditure depend on
  whether the expenditure is expensed or
  capitalized?
• Individuals study accounting data to learn the
  expenditure-profit relation and then make
  predictions about how similar expenditures will
  affect future profits.
• Learning is less effective when an expenditure is
  expensed because expensing signals no expected
  future-profit effect, which focuses attention and
  hence learning on the contemporaneous relation
  rather than the lagged relation.
• Expense implies a negative relation between
  current expense and current profit. If expect
  negative relation, then focus less attention on
  learning the positive relation between current
  expense and future profit, resulting in a less
  accurate estimate of the magnitude of the lagged
  relation.

13

• H Individuals profit predictions will be less
  accurate and consistent, and their consensus and
  self-insight will be lower, when expenditures on
  intangibles are expensed than when they are
  capitalized.
• H Individuals cognitive model will less match
  the optimal (statistical) model when expenditures
  on intangibles are expensed than when they are
  capitalized.
• H In predicting profits, individuals will
  underweight the significant lagged effect of
  intangibles expenditures on future profits more
  when expenditures are expensed than capitalized.
• Results consistent with all hypotheses.

14
2. INTRODUCTION

• DEFINITION OF EXPERIMENT
• DISTINCTIVE FEATURES OF EXPERIMENTS
• BENEFITS OF EXPERIMENTS
• COSTS OF EXPERIMENTS
• VALIDITY OF EXPERIMENTAL EVIDENCE

15
A. DEFINITION

• EXPERIMENT MANIPULATES INDEPENDENT VARIABLES
  (IVs) and OBSERVES (MEASURES) their EFFECTS on
  DEPENDENT VARIABLES (DVs).

16
B. DISTINCTIVE FEATURES OF EXPERIMENTS

• Manipulation of IVs
• Random assignment of people to treatments (i.e.,
  levels of IVs)
• Control of variables other than the IVs DVs

17
C. BENEFITS OF EXPERIMENTS

• Controlled tests of theory performance
  predictions, process validity, boundary
  conditions, assumptions.
• Experiments are especially valuable when theories
  make conflicting predictions and/or empirical
  evidence is mixed.
• High internal validity Control for the effects
  on the DV of all variables except the IV.
• Provide policy input on preferred practices and
  standards.
• Ex ante experiments for policy makers.
• Provide evidence on practices that are not
  currently implemented.
• Emerging practices
• Practices currently prohibited by laws or
  regulations
• Practices not implemented due to competition,
  fad, fashion, self selection.

18
E. COSTS OF EXPERIMENTS

• Abstraction or simplification of the natural
  ecology can affect inferences that are made
• Experiments are generally bad for testing point
  predictions (but good for testing directional
  predictions).
• Experiments are generally bad for exploratory
  investigations (lower external validity).
• Limited ability to collect additional data after
  the experiment.
• Creating some experimental settings is expensive.

19
F. VALIDITY OF EXPERIMENTAL RESULTS

• Results of lab experiments and field research in
  organizational psychology have same sign of
  effects but sometimes differences in magnitudes
  of effects
• participative decision making, goal setting,
  feedback, financial incentives, job performance,
  and job satisfaction.
• Differences in intra-method results are as large
  as differences in inter-method results.
• For these variables, the substantive conclusions
  or policy implications based on either research
  method are similar.
• Locke. 1986. Generalizing From the Laboratory
  to Field Settings.

20
3. IMPORTANT FEATURES OF MA EXPERIMENTS

• THEORY
• REALISM
• CONTEXT
• PEOPLE
• INCENTIVES

21
A. THEORY

• Comparative advantage of experiments is testing
  theories predictions
• 2 common theoretical perspectives in MA
  experiments
• Economics theories about domains (tasks,
  contexts) (e.g., production functions, cost
  functions, market structures)
• Asymmetric information, communication, value of
  information, incentives, learning, market
  structure, risk preferences
• Psychology theories about individuals in domains
  (preferences, beliefs, ability, effort)
• Judgment and decision, communication, learning,
  motivation, negotiation, group dynamics

22
ATTRIBUTES OF GOOD THEORY IN EXPERIMENTAL MA
RESEARCH

• Captures important features of MA and its
  environment. Since MA is context (institution,
  task) specific, it is important to show how MA
  and non-MA are related (interaction).
• Externally valid the proposed concepts and
  causal relations must be believable and capture
  important features of MA that generalize beyond
  the specific experimental setting.
• Theory, not data, generalize.
• Stated as general as possible but still accounts
  for the conceptual relation to be tested.
• States expected causal relations, usually in
  terms of directional differences, and not point
  predictions, for generalizability.

23
B. REALISM

• Swieringa and Weick (1982)
• How realistic should MA experiments be?
• Realistic enough to test the theory with high
  external and internal validities and high
  research realism (engage and motivate people).
• How much mundane realism should be included?
•  
• Cost of experiments can be their "artificiality"
  because of
• people surrogation (accountants and managers who
  have
• no experience with the task in an experiment
  students)
• task surrogation (abstract, simplified, mundane).

24
HOW MUCH TASK REALISM?

• Tendency in designing experimental tasks to want
  to make them as real as possible, but unnecessary
  mundane realism creates complexity which can
  reduce internal and statistical- conclusion
  validities.
•  
• If want mundane realism, then how decide which
  stimuli in the natural ecology to include?
• Fly on the wall, water cooler, drapes, picture on
  desk?
•  
• Do not include (mundane) realism merely to mimic
  natural environments, but rather because it makes
  the experiment a more valid test of the theory.
• Internal, external, construct, and
  statistical-conclusion validities.
•  
• If a less realistic experimental setting
  increases the validity of the experiment, then be
  less realistic.

25
OPTIMAL LEVEL OF TASK REASLISM DEPENDS ON PURPOSE
OF EXPERIMENT

• Testing theory artificiality can be beneficial
  as it provides a basis for valid tests of
  hypotheses by controlling for the effects of
  other variables.
• Experimental psychology typically requires more
  realistic settings because variables in theories
  (e.g., judgment, knowledge, attitude) are related
  to peoples' prior experiences (education,
  training, work).
• Experimental economics frequently provides less
  realistic settings because its focuses on how
  people make decisions using information and
  incentives given in an experiment.
•  
• Testing settings that do not exist in natural
  environments (due to self-selection, competition,
  regulation, law, fad, fashion) research tasks
  will by definition be artificial.

26
REALISTIC or ARTIFICIAL IVs?

• Realistic or artificial values of the IVs?
• Realistic values usually are near the mean of an
  IVs distribution and thus they are not likely to
  be sufficiently different from each other to
  observe (measure) their effect.
• Artificial values (extremes of distribution,
  present/absent) create stronger effects that are
  more detectable.
•  
• Realistic or artificial relations between IVs?
• Realistic relations will frequently have
  significant correlations among the IVs, which
  makes accurately observing the effects of IVs on
  DVs difficult due to multicollinearity.
• Artificial relations can present participants who
  have knowledge of these relations in natural
  settings with unexpected relations (sign, form,
  magnitude) (i.e., combinations of IVs).

27
C. CONTEXT

• What is the context of MA (figure-ground)?
• What are the distinguishing features of MA
  contexts relative to other contexts?
• How much MA context should be included in an
  experiment?
• How operationalize MA context?
• Four context effects (Haynes and Kachelmeier
  1998)

28
1. Information

• Adding new context (task)-relevant information
• Adding nonfinancial to financial performance
  measures can affect judgments and decisions.
• Adding information can adversely affect judgments
  and decisions due to information overload,
  outcome information, and irrelevant or
  nondiagnostic information.
• Clarifying meaning of a context (task)
• Using terms like audit and tax can cause people
  to think about the setting differently (ethics,
  feelings about government, public
  responsibility).
• People are more likely to violate logic in
  abstract contexts. When context is revealed
  (concrete), people with knowledge about the
  context are less likely to make logical errors.

29
2. Salience Influences Attention

• Focal
• order of presentation, location in report
• Unexpected, unusual, surprising, vivid
• indicating an expense has future-period profit
  effects
• Context can influence how people allocate their
  limited attention among information
• planning vs. control, prediction vs. diagnosis,
  explaining your own vs. others performance
• Salience induced by context of information
• performance-report design focuses attention on
  particular parts of a report
• presenting budget variances as well as budget and
  actual focuses attention on the variance even
  though it is not informative

30
3. Context Affects Motivation

• Context can highlight consequences of behavior.
• Framing economically-equivalent outcomes as gains
  or losses affects preferences.
• Choice of economically-equivalent incentive
  contract depends on whether contract outcome is
  framed as a bonus or penalty.
• Accountability makes people more vigilant,
  effort-intensive in justifying judgments and
  decisions.
• Including accountability increases peoples'
  attention and effort if the accountability has
  valued consequences for them.

31
4. Context Affects Memory

• Context can interact with knowledge in
  individual's memories to affect their judgments
  and decisions.
• People with context experience can activate their
  knowledge as prompted by the context.
•  
• Context provides cues as to what knowledge to
  activate.
•  
• Memory effects can be
• positive (e.g., use more relevant knowledge to
  make judgments and decisions) or
• negative if context activates the wrong knowledge
  (e.g., typical knowledge for an atypical
  situation).

32
ENHANCE CONTEXT WHEN

• generic language would be confusing or ambiguous
  (information)
• attention-directing contextual features are of
  theoretical interest (salience)
• want to study the influence of motives (ethics,
  pubic responsibility) beyond those characterizing
  the economic task (motivation)
• 4. individuals' knowledge is consistent with a
  richer context (memory)

33
 USE A NEUTRAL CONTEXT WHEN

• context cues can mislead more than clarify
  (information)
• theory assumes individuals will give equal
  attention to all information (salience)
• contextual motives could bias incentives of
  theoretical interest (motivation)
• 4. individuals' knowledge either is incapable of
  interpreting a richer context or might result in
  interpreting a context in ways that go beyond the
  concept of interest (memory).

34
CONTEXT and M A

• Include context features that are expected to
  interact with MA variables.
• Interaction of MA and non-MA context and MA and
  non-MA knowledge affects judgments and decisions.
•  
•  
• If include context, then give reason related to
  theory why it is included and how context is
  expected to affect the IV and DV.
• Don't add context just to increase mundane
  realism.
•  
• If exclude context, then give reason related to
  theory why a richer context would not be
  desirable.

35
D. PEOPLE

• Single Person
• Motivation, communication, judgment and decision,
• learning, performance, adaptation
•  
• Multiple Person
• Hierarchy superior-subordinate communications
  and
• negotiations
• Teams cooperation, performance
• Horizontal negotiations between subunit managers
• Markets Information exchange (collusion,
  transfer of
• asymmetric information), learning, market pricing
• (assets, audits), market structure

36
PEOPLE SURROGATION

• Are people within boundary conditions of theory?
• Dont need random sample of people within
  boundary conditions.
• Need random assignment of people to treatments.
• What about using students?
• Some students are investors and some are or have
  been managers or accountants.
•  
• A homogeneous sample (e.g., students) can reduce
  error variance, which increases the power of
  statistical tests.
•  

37
PEOPLE SURROGATION

• Key is whether people have the ability,
  knowledge, motivation, and preferences to be good
  surrogates for people in the target population
  you want your theory to generalize to.
•  
• Using students or other inexperienced/novice
  individuals is appropriate if the purpose of a
  study is to test how
• people learn and use information presented in a
  study
• incentives provided in a study affect behavior
  and performance
• peoples' general cognitive abilities affect their
  behavior
•  
• Using students or other inexperienced/novice
  individuals is not appropriate if the purpose of
  a study is to test how
• people use their task-related knowledge or
  motivation that they brought to the experiment.

38
E. INCENTIVES

• Provide incentives that will motivate people to
  be involved in the research study, but will not
  bias the results research realism.
•  
• If a study uses people to perform research tasks
  that are directly related to their everyday jobs,
  then using incentives can distort or interfere
  with the effects of their real-world incentives.
• Not providing them with incentives can reduce the
  magnitude of IV effects but not alter the
  direction of effects.
•  
• If a study uses people to perform research tasks
  that are not the tasks they do in their everyday
  jobs, then using incentives to motivate them is
  appropriate to increase research realism.

39
INCENTIVES

• What should be the incentive?
• Monetary or non-monetary?
• Fixed or performance contingent?
• Tournament?
•  
• If a study is testing for incentive effects, then
  using incentives consistent with those in the
  theory is appropriate.
• If using incentives is appropriate, then does the
  theory specify how sensitive they should be to
  performance?
•  
• Equal expected pay across treatment levels to
  eliminate the magnitude of incentives as a
  competing explanation for results.
•  
• In the U.S., the amount of incentive pay ranges
  from 5 to 100 per hour, with the average being
  about 8 to 20 per hour.

40
4. READING LIST

• Birnberg, J., M. Shields, and M. Young. 1990. The
  Case for Multiple Methods in Empirical Management
  Accounting Research. Journal of Management
  Accounting Research 33-66.
• Chow, C. 1982. The Effects of Job Standard
  Tightness and Compensation Scheme on Performance
  An Exploration of Linkages. The Accounting
  Review 667-685.
• Frederickson, J., S. Peffer, and J. Pratt. 1999.
  Performance Evaluation Judgments Effects of
  Prior Experience Under Different Performance
  Evaluation Schemes and Feedback Frequencies.
  Journal of Accounting Research 151-165.
• Haynes, C. and S. Kachelmeier. 1998. Effects of
  Accounting Contexts on Accounting Decisions A
  Synthesis of Cognitive and Economic Perspectives
  in Accounting Experimentation. Journal of
  Accounting Literature 97-136.

41

• Libby, R. R. Bloomfield, and M. Nelson. 2002.
  Experimental Research in Financial Accounting.
  Accounting, Organizations and Society 775-810.
• Locke, E. (Ed.). 1986. Generalizing from the
  Laboratory to Field Settings. Lexington, MA.
  Lexington Books.
• Luft, J. and M. Shields. 2001. Why Does Fixation
  Persist? Experimental Evidence on the Judgment
  Performance Effects of Expensing Intangibles. The
  Accounting Review 561-587.
• Luft, J. 1994. Bonus and Penalty Incentives
  Contracting Choice by Employees. Journal of
  Accounting and Economics 181-206.
• Luft, J. and M. Shields. 2003. Mapping Management
  Accounting Graphics and Guidelines for
  Theory-Consistent Empirical Research. Accounting,
  Organizations and Society169-249.
• Schepanski, A., R. Tubbs, and R. Grimland. 1992.
  Issues of Concern Regarding Within- and
  Between-Subject Designs in Behavioral Accounting
  Research. Journal of Accounting
  Literature121-150.

42

• Sprinkle, G. 2003. Perspectives on Experimental
  Research in Managerial Accounting. Accounting,
  Organizations and Society 287-318.
• Swieringa, R. and K. Weick, 1982. An Assessment
  of Laboratory Experiments in Accounting. Journal
  of Accounting Research, Supplement 56-107.
• Young, M. and B. Lewis. 1995. Experimental
  Incentive-Contracting Research in Management
  Accounting. in R. Ashton and A. Ashton (Eds.),
  Judgment and Decision Making Research in
  Accounting and Auditing (Cambridge University
  Press).
• Vera-Munoz, S. 1998. The Effects of Accounting
  Knowledge and Context on the Omission of
  Opportunity Costs in Resource Allocation
  Decisions. The Accounting Review 47-72.