Accounting Question

JMARVolume Five
FaU ]993
15
Antecedents and Consequences of
Participative Budgeting: Evidence on
the Effects of Asymmetrical
Information
Michael D. Shields
San Diego State University
and
S. Mark Young
University of Southern California
Abstract: A review of empirical studies on participative budgeting indicates that
most have focused on its consequences (e.g., attitude, motivation, performance,
satisfaction) rather than its antecedents. We suggest that the mixed results of
participative budgeting research are due to incomplete models of the process.
Further, we propose that researchers develop theories that tie particular antecedents of participative budgeting to specific consequences. Next, we develop
a model that uses information asymmetry between local and central management
as the antecedent condition and link it to: (1) participation of local managers in
setting their budgets, (2) budget-based incentives and (3) firm-wide performance.
Data to test these relationships were gathered from corporate controllers of the
Standard and Poor’s 500, and the relationships were tested using path analysis. Three of the four hypotheses were supported—^the effects of information
asymmetry on participative budgeting, participative budgeting on budget-based
incentives and the effect of budget-based incentives on firm-wide performance.
The fourth hypothesis—the effects of information asymmetry on budget-based
incentives was only marginally supported. We also propose three other possible antecedents of participative budgeting that can be investigated: (1) a desire to positively influence individual attitudes, behavior and performance, (2) to
reinforce a particular culture, and (3) to provide a mechanism for organizational
learning.
A decade ago, Brownell [1982a] developed a framework for organizing
research on parUcipatlve budgeting that divided the literature into two parts.
The first specified antecedent conditions which created a demand for participation such as the environment (e.g.. uncertainty), organizational variables (e.g.. structure, technology), and national culture, while the second
focused on participation’s consequences such as Increased motivation,
commitment, satisfaction and performance. In his review, he concluded
that the [p. 146-7],
This paper has beneJUedfrom comments received from Bob Capettinl Robert Chenhall.
Chee Chow. Ken Merchant. Sarah Bonner. two anonymous reviewers and participants at the accounting workshop at Utah, the 1991 Management Accounting Research Conference in Atlanta, and the Second European Management Control Symposium, Paris, 1992. We would like to thank MUn Guofor his research assistance
and Mark Young gratefully acknowledges the financial assistance of the KPMG Peat
Manuick Faculty Fellowship Program.
266
Journal of Management Accounting Research, Fall 1993
Research effort to date has tended to focus attention on the more general question of luhether participation is effective or not. rather than specifying the conditions under which It is effective. Conflicting results permeate the literature …
The main focus of participative budgeting research since Brownell’s
[1982a] review, however, has still been on Its consequences, rather than
Its antecedents (see Blrnberg et al. [19901 for a review). The literature relating to whether consequences such as performance are affected positively by participation Is still conflicting and Indicates that addressing the
reasons for participation may lead to a better understanding of the extant
research.
This paper has two purposes. First, ln the next section, we review the
literature and propose that theories of participation Include both its antecedent conditions and consequences. Second, we develop a model of participative budgeting that uses the information asymmetry between a superior and a subordinate as an antecedent and firm-wide performance as a
consequence variable. Also it is hypothesized that the relationship between
these variables is moderated by participative budgeting and budget-based
incentives. Following this discussion, the research method and results of
the hypothesis testing are presented. The hypothesis tests are based on
data gathered from corporate controllers of Standard and Poors (S&P) 500
firms. Finally, directions for future research are discussed, with particular
emphasis on identifying three other antecedent conditions of participative
budgeting.
LITERATURE REVIEW
In order to structure the review of the accounting literature, a classification analysis was performed on the 28 empirical studies of participative
budgeting that used either laboratory experimentation or surveys published
in The Accounting Review. Journal of Accounting Research and Accounting.
Organizations and Society from 1970 through 1991. Of the 28 studies, 24
examined only the consequences of participation.’ Bruns and Waterhouse
[19751 and Seiler and Bartlett 11982] examined (cross-sectional) determinants of participative budgeting and only Merchant [1981, 1984] examined both determinants and consequences. A consistent belief rooted in
early participative budgeting studies [Argyris, 1952; Becker and Green.
1962] was that participation was used to improve outcomes such as employee morale, motivation, commitment and satisfaction. The inclination
to study consequences Is highly consistent with research on participative
decision making in contexts other than budgeting [Locke and Schweiger.
1979).
‘The 24 studies are Brownell 11981, 1982b.c, 1983a,b. 1985). Brownell and Dunk (19911,
Brownell and Hirst [19861, Brownell and Mclnnes I1986I. Chenhall [1986|, Chenhall and
Brownell I1988|, Cherrlngton and Cherrlngton [1973], Collins |1978|, Dunk 11989|, Foran
and DeCoster (19741, Frucot and Shearon (1991], Kenis 11979). Merchant (1985|. Mia (1988.
19891. Milan! (19751. Onsl (1973|, Searfoss [1976( and Tiller (1983(. Field studies were
excluded because a review by Fenrelra and Merchant (in press] Indicated that no Oeld studies
on participative budgeting were published between 1984 and 1992.
shields and Young
267
Since survey research methods are used in this paper, a classification
analysis of the 23 surveys (five of the 28 cited above were laboratory experiments) which examined cross-sectional correlations between participative budgeting and its consequences was performed. One survey was
excluded because the level of disclosure did not permit a classification analysis. Two important findings emerged. First, of the 22 remaining studies,
many different theoretical frameworks were used to motivate the analyses
of the outcomes of participative budgeting (e.g.. contingency, balance, role,
leadership style, achievement motivation, locus of control, goal setting,
cognitive dissonance, level of aspiration, expectancy, personality and evaluation style). Diversity as to theories exists both across studies which used
the same variables and within a study. Our conjecture is that this diversity
stems, in part, from the lack of understanding regarding why participative
budgeting is used in organizations.
Second, a count of the number of times that a particular consequence
variable was significant (with an alpha probability of less than 0.05) was
performed. Overall, seven consequence variables were tested (satisfaction.
performance, budget attitude, motivation, job tension, slack and job involvement), and there were 24 tests of hypotheses in which the inferential
statistic was significant in < 0.05). However, there were 35 hypothesis tests in which the inferential statistic was not significant and/or the sign was not as predicted.^ These two findings are consistent with Brownell's (1982a. p. 124) earlier assessment that "...the literature remains fraught with contradiction, overlap and a general lack of conclusiveness...." This diversity of frameworks can be viewed as both a strength and a weakness. The strength is that researchers have examined the effects of participative budgeting using a variety of theories and variables. The weakness is that research has not produced a coherent or unified explanation for the conditions under which participation is beneficial to individuals and organizations. The diversity also limits the use of statistical meta-analysis (Hunter et al. 1982]. The aforementioned results relating to the budgeting context are similar to research In participative decision making in general. Several qualitative and statistical meta-analyses of the participative decision making literature also lead to the conclusion that the effects of participation on consequence variables are weak (see Locke and Latham [1990. pp. 169-170]). Locke and Schweiger [19791 classified laboratory experiments, field experiments and correlational field surveys as providing evidence of a positive, negative or no-difference effect of participative decision making on performance and satisfaction. The evidence was ambiguous regarding performance as approximately an equal number of studies found a positive or a negative effect, with the modal study finding no effect. However for satisfaction, 60 percent of the studies found a positive effect, with 30 percent finding no effect. In another classification study. Leana et al. (19901 found ten studies in which there was a positive relationship between participa*rhese 59 tests break down as follows: satisfaction (5 significant, 3 not significant), performance (12 signiflcant, 13 not significant), attitude (1 significant. 10 not significant), motivation (3 significant, 7 not signiflcant). Job tension (1 significant, 1 not significant), slack (1 significant, 1 not significant) and job Involvement (1 signiflcant. 0 not significant). 265 Journal of Management Accounting ResearcK Fall 1993 tive decision making and performance. 20 studies in which there was no relationship, two studies in which there was a negative relationship, and 15 studies in which the relationship was Interactive. Another classification of studies which examined the participative decision making-satisfaction relationship also found a great deal of diversity in results with many studies finding a positive association and many others finding either a negative or no association (Cotton et al.. 19881. Wagner and Gooding (1987bl performed a statistical meta-analysis of 70 published studies on participation and its consequences (performance, decision performance, motivation, satisfaction, and acceptance of consequences given participation) to reconcile the conflicting meta-analyses of Miller and Monge (19861 and Wagner and Gooding (1987al. Their main findings [p. 524] were that the correlation between participation and its various outcomes is 'Modest," and that "many of the noteworthy positive findings" reported were due to methodological artifacts. Further, they found that only a small correlation is due to a direct relationship between participation and consequence variables. Determinants of Participative Budgeting Four streams of research have discussed some antecedent conditions for participation. The first stream is based on contingency theory (see Birnberg et al. (19901). This research has found that contextual variables, such as environmental uncertainty and technology, are correlational antecedents of participative budgeting [Birnberg et al. 1990). The second stream stems from the work of Locke and colleagues and is focused on individual level participation and its motivational consequences (e.g.. motivation, commitment, performance, satisfaction) [Locke and Schweiger, 1979; Leana et al., 1990]. Locke proposes that the benefits of participative decision making are work motivation and/or cognitive understanding. The cognitive perspective has not been pursued nearly as much as the motivational perspective; however, this direction for research is gaining momentum as Locke et al. [1981. p. 139] state that, ... the single most successful field experiment on participation to date stressed the cognitive benefits: participation was used to get good Ideas from workers as to how to Improve performance efficiency. Recently, some participative decision making research has either tested hypotheses based on the cognitive perspective or tested the relative strength of motivational and cognitive explanations [Locke and Latham. 1990, pp. 170-1]. The third stream of research stems from the Japanese management literature. Many Japanese firms employ concepts such as participative decision making and quality circles with the goal of obtaining accurate infomiation about local conditions in the various parts of a firm [Ouchi, 1979; Young. 19921. This information is used to improve firm-wide performance directly through more efficient resource allocation. Whfie motivational consequences like improved morale and job satisfaction may occur, they are not the primary reason for participation in Japanese firms. Finally, ln contrast to many of the behavioral studies, agency theorists have analytically modeled how differences in information between a superior and a subordinate, or information asymmetries, affect the demand for Shields and Young 269 participative budgeting In organizations IMagee. 1980; Chrlstensen. 1982; Balman and Evans. 1983;Penno. 1984, 1990; Kirby etal.. 19911. The agency view is that participation Is used by a superior to learn about the local environment so that decisions such as resource allocation can be improved and optimal Incentive contracts designed for subordinates.^ Hypotheses Agency theorists ai-gue that the demand for participative budgeting arises because various parties engaged In the budgeting process possess differential information about uncertainty (e.g.. central and local management) [Magee. 1980; Baiman. 1982; Christensen, 1982; Baiman and EVans. 1983; Penno. 1984. 1990; Klrby et al.. 19911. Except for a few studies (e.g. Pope [19841 and Young [19851) the behavioral literature has not exploited this explanation for participative budgeting."* Instead, when there has been an explicit focus on antecedents, the behavioral literature usually discusses absolute levels of uncertainty (present or absent as in Young [1985]) but not differences in levels of uncertainty. The agency perspective assumes that a signiflcant reason for the existence of participation is the transfer of Information from a subordinate to a superior and that there are potential gains for both parties (e.g.. better information, resource allocation, incentive plans, performance, compensation). Information asymmetry generally would be most severe in flrms which are extremely large, and geographically dispersed with diverse products and technologies. In these firms, central management does not know relatively more about local conditions than do local managers. Thus, central management can use participative budgeting to leam about local environments as well as to provide motivation. This allows central management to better allocate resources to operating units and to offer operating managers better incentive contracts. In general, as the difference in information Increases, there is more potential gain from participation. This argument is summarized in the first hypothesis. ^As an example of how an understanding of antecedents can affect particular consequences of participation the following example is offered. Consider a situation in which a researcher is trying to understand the effect that partieifjation is having on lower level managers. Many studies have assumed that senior management uses participative budgeting to increase job satisfaction. However, in some situations senior management might decide to use participation to obtain a subordinate manager's private Information, rather than attempting to Increase his or her job satisfaction. In this situation, participation might consist of a meeting in which the superior aggressively attempts to obtain the manager's Information. A researcher surveying the firm In an attempt to determine whether participation Increased job satisfaction would probably find few positive results. Contrast this with a situation In which a superior truly desires to inerease job satisfaction and performance by allowing subordinates the opportunity for participation involving Joint decision-making |Locke and Schweiger, 19791. Structuring research (especially survey studies) on participative budgeting by including questions relating to why subordinate managers think participation is used and the antecedent conditions that are currently in place could provide more consistency for the results in the literature. ^However, behavioral research has empirically examined the association between Information asymmetry and information misrepresentation. Young [19851 examined how the effect of Information asymmetry on slack creation was moderated by felt social pressure not to misrepresent performance capability. Jaworski and Young [19921 examined how the effect of Information asymmetry on information misrepresentation was moderated by Job tension and jjerson role conflict. 270 JoamaX of Management Accounting Research, Fall 1993 HI: There is a positive association between the extent of Information asymmetry and the use of participative budgeting. Central management uses local Information gained from participative budgeting for at least two purposes. First, the Information is used to Improve the ex ante efficiency of resource allocation among the operating units and, hence, expected firm-wide performance. Second, this information can be used to design more effective incentive systems that can be used to increase motivation. It Is conjectured that when participative budgeting is used more extensively, management makes greater use of Incentives that reward performance based on meeting or exceeding the budget. Such Incentives are known as budget-based schemes (Demski and Feltham. 1978: Balman, 1982: Chow. 1983]. H2: There is a positive association between the use of participative budgeting and the use of budget-based Incentives. Since subordinates can use local Information to their personal advantage, some agency models use budget-based incentives to encourage subordinates to use their private Information to increase the level of goal congruent behavior IDemski and Feltham. 19781. When an information asymmetry exists between local and central management, in addition to using participative budgeting, central management also uses budget-based incentives. Since senior management Is unaware of how a manager might use his or her information, budget-based Incentives are employed to motivate managers to use the information to improve performance. H3: There is a positive association between the extent of Information asymmetry and the use of budget-based incentives. A critical question being asked in this research is, does the firm gain from the use of participative budgeting? One positive outcome relates to improvements in firm-wide performance. Hopwood 11976] argues that there is not a direct link between participative budgeting and performance. His position is consistent with the empirical results ln several of the studies on participative budgeting (e.g.. Brownell (1982b]. Merchant [1984]). One variable that may moderate the relationship between participation and performance Is budget-based incentives. We believe that an important benefit derived from participative budgeting and budget-based incentives is that expected firm-wide performance improves as the superior allocates resources to subordinates with the best reported opportunity sets and those who are motivated to maximize goal congruent performance. Two types of empirical research show a positive relationship between budget-based incentives and performance. Laboratory research by Chow [19831 and Waller and Chow 11985] report that performance increased as the incentives offered were more budget-based, although controlling for performance capability mediated the effect. Murphy's [1985] econometric analysis of executive compensation for the period 1964-1981 using 72 lai:ge U.S. manufacturing firms found a strong positive relationship between executive compensation and organization performance (retum realized by the common stockholders). The next hypothesis is consistent with these findings. H4: There is a positive association between the use of budget-based incentives and firm-wide performance. Shields and Young 271 Figure 1 Model Participative Budgeting (X2) Information Asymmetry (XI) ^~ H3 ^ Budget-based *" Incentives tX3) ^""^ ^j|4^ ^ Firm-wide Performance (X4) The four hypotheses ln the model are shown in Figure 1. This model is the basis for testing the hypotheses. The model expresses the four relationships in the four hypotheses as paths. (P^ J. among the four variables. The model Indicates that in order to reduce the potential for realizing adverse effects from information asymmetries, central management uses participative budgeting (P, g) and budget-based Incentives (Pj 3). ParticipaUve budgeting also has an effect on budget-based incentives (Pj 3). which in turn affects flrm-wide performance (P34). Research Method THE SURVEY The sample of firms consists of those belonging to the S&P 500. These firms were surveyed as they represent the entire U.S. economy. Corporate controllers were selected as Individuals to survey. Controllers were chosen because they: (1) play a key role in designing the information and control system of a firm and are thus likely to appreciate the overall picture of the budgeting system (probably more so than the average manager). (2) have direct and frequent access to top management to discuss Issues relating to control system design and operation, and (3) provide a perspective on the role of participative budgeting that has been missing from the literature. Thirteen firms were deleted initially from the S&P 500 population because they were either companies which were new or headquartered outside of the U.S. Of the 487 remaining firms, a total of 98 usable responses were received. The effective response rate of 20 percent is at the low end of what is considered to be the expected response rate (20-40 percent) for mail surveys [Kerllnger, 1986]. The expected response rate was low. in part, because some of the information requested was proprietary (e.g., compensation). Because the respondents were guaranteed anonymity, it also was not possible to do an additional mailing in order to Increase the sample size. While acknowledging the relatively low response rate, each sample firm represents an important business. Another way to evaluate the adequacy of this sample is to compare it to the samples used in the other survey studies on participative budgeting 272 Journal of Management Accounting Research, Fall 1993 which examined antecedents and/or consequences. Inspection of these 23 studies published ln The Accounting Review, Journal ofAccounting Research and Accounting. Organizations and Society during the 1970s and 1980s indicates that the unit of analysis ln 22 studies was a manager (typically first-line). These 23 studies are based on a total of 17 data sets that were obtained from an average of 14 firms, with questionnaires being sent to an average of 120 respondents with the average number of usable responses being 74. In most studies, a convenience sample of firms, typically fairly small ln size and located relatively close to the researchers home university, were used. Given these data, we feel that our sample compares favorably with samples used ln prior research. Measurement of Variables Information asymmetry, participative budgeting, budget-based Incentives and firm-wide performance were each measured by subscales which were subsequently aggregated to form an overall measure of each variable. Information asymmetry was measured in relative terms (e.g.. subordinate knows twice as much) as it is difflcult to assess In absolute terms. Since no established measures of lntra-organizational vertical information asymmetries exist for single respondents, one was developed. The information asymmetry examined was between central management (top management at headquarters) and the investment and profit center managers Immediately below central management. As these information asymmetries can occur In relation to Input, process and/or output, a question was developed for each. Information asymmetries were directly measured by asking. "How much more does each manager know relative to central management about the following Items?" The five items were labor input, materials input, capital input, market for outputs, and technology (transformation of Inputs to outputs). The response scales were anchored by (1) "Manager Knows the Same as Central Management" and (7) "Manager Knows a Lot More than Central Management." The extent of participative budgeting was measured using five questions adapted from prior research (Brownell. 1982b. 1985: Merchant. 1981; MUanl, 19751. The first three were: (1) "How important is the manager's contribution to the setting of the budgets?". (2) "How important Is it that budgets Include changes that were suggested by the managers?", and (3) "How important is it that a budget is not finalized until a manager is satisfied with it?" These questions were anchored: (1) "E:xtremely Unimportant" and (7) "EMremely Important." The fourth question. "How infiuentlal do you feel that the managers are ln setting the budgets?", was anchored by: (1) "Not at All Influential" and (7) "Extremely Infiuential." The fifth question was "How frequently does central management Initiate budget-related discussions with the managers?", anchored by: (1) "Extremely Infrequently" and (7) "Extremely Frequently." Budget-based Incentives were measured by responses to ilve questions. Four of the questions had seven-point response scales: (1) "The compensation system for managers is very clearly specified In terms of how compensation is related to budgeted performance." (2) "Managers' financial rewards Increase as actual performance increasingly exceeds budgeted performance." (3) "How Infiuential is the actual performance relative to the Shields and Young 273 budgeted performance of a center In affecting the probability that the manager will be promoted?" and (4) "Consider the managers whose last year's performance was In the top 25 percent of the managers' performance In your firm. How likely is it that they received larger budget performancebased bonuses than did the managers whose performance was below the top 25 percent?" The first two questions were anchored by (i) "Strongly Disagree" and (7) "Strongly Agree."The third question was anchored by (1) "Not at All Influential" and (7) "Extremely Influential." and the fourth question by (1) "Extremely Unlikely" and (7) "Extremely Likely." The fifth question asked the percentage of managers who received a bonus based on their centers' actual performance relative to budgeted performance. Firm-wide performance was measured through answers to four questions. These were: (1) percentage change In net income. (2) percentage change in common stock price. (3) percentage change In ROI and (4) a subjective rating of the overall performance of the firm, anchored by (1) "Worst Possible Performance" and (7) "Best Possible Performance." Path Analysis Path analysis was used for hypothesis testing. The effectiveness of the method lies In Its ability to construct a set of statistical (correlational) relationships among a set of variables. In path analysis, the relationships between variables are specified by path coefficients (P^ y) that are Identical to standardized partial regression coefBcients [Li. 1970; Duncan. 1966). The results of the path analysis also provide information on the relative Importance of the independent variables in terms of their effect on the dependent variable{s). RESULTS Descriptive Statistics Descriptive statistics for the four variables are shown in Tables 1 and Pearson correlations in Table 2. For budget-based Incentives and firmwide performance, the descriptive statistics for the scales are based on standardized Z values because the underlying subscales had heterogeneous scale properties. Thus, these variables' subscales were standardized and the standardized values were summed across the subscales. The convergent validity of the four variables was assessed using Cronbach's alpha. All four alphas were greater than 0.60 (Table 1). the lowest value at which a scale is considered reliable. Tests of Hypotheses In testing the four hypotheses, the expected sign of each correlation or path coefficient is positive and the expected magnitude is significantly greater than zero. In path analysis, the blvariate correlation between any pair of variables can be estimated from the paths leading from common antecedent variables. In our model, only one variable is estimated by more than one variable, budget-based incentives. Xg. Each of the other three paths. P, 3. Pj 3. and P3 4. has only one antecedent variable. The path coefficients are; 274 Journal of Management Accounting Research. Fall 1993 Table 1 Descriptive Statistics: Scales and Subscales Cronbach's Alpha SCALE Subscale' INFX^RMATION ASYMMETRY Labor Material Capital Output Technology 95 95 95 95 95 95 23.61 5.44 5.36 3.76 4.55 4.51 7.26 1.50 1.56 1.77 1.88 1.81 0.90 PARTICIPATTVE BUDGETING (1) Contribution* (2) Changes (3) Satisfied (4) Influence (5) Discussions 95 95 95 95 95 95 28.43 6.06 5.99 5.66 5.84 4.87 4.79 1.18 1.16 1.45 1.14 1.25 0.83 BUDGET-BASED INCENTIVES Z (1) Specificity (2) Rewards (3) Influence (4) Performance (5) Bonus 68 68 68 68 68 68 -0.09 4.57 5.07 4.78 4.93 56.31% 4.69 1.92 1.66 1.32 2.07 44.66% 0.93 -0.09 13.4% 24.6% 6.1% Total Performance 77 4.7 2.68 57.50% 24.10% 5.35% 1.38 0.68 ROl 77 77 77 77 FIRM WIDE PEPy^RMANCE Z Net Income Stock Price _S_ 'See Measurement of Variables subsection for descriptions of the questions and response scciles. Table 2 Pearson Correlations Participative! Information Asymmetry Budgeting Pariidpative Budgeting Budget Based Incentives Firm-Wide Performance • p