Вы находитесь на странице: 1из 34

How Changes in Compensation Plans Affect ' '' Employee Performance, Recruitment, and Retention: An Empirical Study of a Car

Dealership*
JOANNA L. Y. HO, University of California. In>ine LING-CHU LEE, National Pingtung Institute of Commerce ANNE WU, National Chengchi University 1. Introduction Economic theory argues that performance-based compensation contracts increase employees' incentives to exert effort, resulting in improved performance (Baker. Jensen, and Murphy 1988; Milgrom and Roberts 1992; Prendergast 1999). Previous empirical and laboratory studies on this topic have covered various compensation schemes and examined how changes to a more performance-sensitive incentive scheme influence employees' compensation and performance (Waller and Chow 1985; Lazear 2000; Banker, Lee, Potter, and Srinivasan 2001). However, companies may also switch from more to less performance-sensitive incentive schemes. For example, prior to 1992, Sears adopted a commission-based compensation system for auto repair salespersons. This compensation scheme enticed salespersons to falsely diagnose brake and alignment problems, which cost Sears $15 million in refunds and other settlement costs. Consequently, Seais discontinued tlie conmiission system (DriscoU 1994). Furthermore, during the 1990s, most shoe manufacturers adopted u survival strategy by shifting from a piece-rate to an hourly-rate compensation system (Freeman and Kleiner 1998). In 2001 Fujitsu also changed its compensation system from performance-based to nonperformance-based (Tanikawa 2001 ) because under the perform anee-based system, workers tended to set goals as low as possible in order to receive raises and promotions. Surprisingly, no research has addressed the impact of changes to less performance-sensitive plans on employee performance. Our study provides evidence on how a change to a less performance-sensitive incentive scheme affects individual employee productivity and compensation.' in
* Accepted by Alan Webb. We gratefully acknowledge the thoughtful and helpful cumnicnts ut twu anonymous reviewers and especially Ahm Webb (associate editor). We also acknowlcdjji: ihc helpful comments of Rajiv Banker. Chce Chow. Mahenclra Gupla. Yuhchang Hwang. ChHsln Karutia. Thomas Lin, Taylor Randall, Jae Yong Shin, atid workshop participants at the Paul Merage School of Business, University of California, Irvine, the International Association for Accounting Education and ReseaR'h (lAAER) World Congress oF Accounting Educators, the American Accounting Association Annual Meeting, and ihe Management Accounting Section Midyear Conference. We are indebted lo ihe research company for ils generous suppon in prt)viding us wiih ihc needed research data.

'

Contemporary Accounting Research Vol. 26 No. 1 (Spring 2009) pp. 167-99 CAAA doi:10.1506/car.26.1.6

168

Contemporary Accounting Research

addition, we examine whether employee ability affects productivity in light of the plan change and which employee group is affected the most by such a change. In addition to influencing employees, the compensation plan can affect company performance by influencing recruitment and retention (Stiglitz 1975; Salop and Salop 1976; Demski and Feltham 1978; Milgrom and Roberts 1992). For example, performance-based compensation contracts can attract and retain high performers and differentiate high from low performers (e.g.. Baron and Kreps 1999; Banker et al. 2001). A company benefits when lower-performing employees leave, but can suffer a setback when higher-performing employees depart. Thus, it is important to consider who will join or leave a company when the performance sensitivity of the compensation contract is changed. This study examines the types of employees who are attracted to the firm, or who leave, when the compensation plan becomes less performance sensitive. In addition, we investigate causes (e.g., compensation loss) that may account for employee turnover. Our study focuses on a car dealership in Taiwan that changed its compensation scheme from totally commission-based to a mix of fixed salary and lower commission rates. This change was in response to the requirement of the 1998 Taiwanese Labor Law amendment. Our database includes 4,392 individual-level observations (e.g., employees' compensation, sales quantities, performance ratings, and demographic information) as well as firm-level data (e.g., turnover rates, new hires, revenues, income) for a 56-month period. We find that the change in compensation plan lowered individual productivity and compensation, especially for higher-performing employees. As predicted by theory, we also observe that the less performancesensitive plan attracted low performers and retained fewer high-ability performers. Interestingly, we do not find that the decrease in individual productivity translates into poor overall company performance. Our interviews with managers and additional analyses suggest that in light of the decreased individual sales productivity, the company hired more employees and also changed its sales mix by giving employees more incentives to sell cars that had higher margins and faced more competition. This study adds to the extant hterature by being the first to examine the impact of a change to a less performance-sensitive compensation scheme on low-level employee productivity. We directly measure the impact of a plan change (i.e., from more to less performance-sensitive) on individual productivity and the adverse selection effects by using data both prior to and after the plan change. Prior empirical work in this area tends to be experimental in nature and tests for only one of the two effects (Young and Lewis 1995). However, in this study we are able to examine both incentive and adverse selection effects simultaneously. Furthermore, Banker et al. (2001) examine how implementing a more performance-sensitive incentive scheme affects employee performance. Yet they only had after-the-plan data on divisions implementing or not implementing the new plan. Banker et al. (2001, 322) were explicit about this limitation: "This lack of objective, individuallevel performance data has limited our knowledge of how an incentive plan affects employees" selection of employment and effort level." Also, Banker et al.'s inference regarding employee retention and turnover was based on comparing the sales productivity of those who stayed with those who left only after the plan implementation. CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

169

In our study, the availability of employee-level data before and after the plan allows more robust conclusions about effects on employee recruiting and turnover. Furthermore, this study goes beyond related prior studies by investigating how employee ability affects performance when a company changes to a less performancesensitive cotnpensation plan. We measure employee ability by using a factor analysis of three proxies for ability: employee annual performance rating, nutnber of cars sold, and tbe reciprocal of time to the first promotion since the etnployee joined the dealership. Our employee ability measure not only helps us identify specific perfortnance groups that are most affected by tbe compensation plan change, but also provides additional insights into causes of employee turnover after tbe plan cbange. Tbese findings also have practical implications because they tnay help top management anticipate tbe effects of a cbange to a less performance-sensitive compensation plan on incentives and efforts of different etnployee groups as well as the effects of a change on employee recruitment and turnover. Tbe remainder of tbis paper is organized as follows. Section 2 describes the field site and tbe nature of the changes made in the etnployee compensation plan. We develop our hypotheses in section 3. Data and empirical models are discussed in section 4. Section 5 presents the empirical results, and section 6 summarizes our results witb concluding remarks. 2. Field site The field site we studied is the largest car dealership in Taiwan, where tbe domestic autotnobile market has been nearly 400,000 car sales per year.^ In 2001, tbe car dealership had 1,248 sales representatives and 87 sales outlets located throughout Taiwan and mainland China. The details concerning tbe field site reported in this section are based on both interviews and archival data from the company. Our relationship witb tbe field site allowed u.s to interview tbe chief executive officer (CEO), senior and branch managers, and salespersons to get information about the compensation scheme, employee reactions, tbe legal environment, competition, and various management-related issues. Tn total, we spent roughly 220 hours in site visits, data collection, and conducting more than 20 in-depth interviews, ranging from 50 minutes to two hours. Eor eacb interview, we sent questions to tbe interviewees at least one week before the meeting. The car dealership carries various models of sedans and trucks and has a heterogeneous group of customers. For sedans, annual salary is a major factor in determining which model (economy versus luxury) a customer buys. However, customers who buy trucks are either owners of small businesses who deliver merchandise to their clients or street vendors who sell merchandise in night markets. During the sale, an employee first identifies a customer's wants and preferences (e.g., transporting children to school, family vacation, larger luggage space). Then, he or she introduces a car model tbat best matches tbe customer's needs and preferences and answers tbe customer's questions concerning tbe quality and prices of cars sold by tbe dealership and its competitors. Eurtbermore, employees need to understand potential customers' financial needs and be able to present different financial CAR Vol. 26 No. 1 (Spring 2009)

170

Contemporary Accounting Research

alternatives based on those needs. The time spent by an employee to close a deal ranges from one week to one month, depending on experience, ability, and the type of car being sold. In general, junior employees spend twice the amount of time closing a deal than senior employees, and it takes more time to sell sedans than trucks. Furthermore, an employee handles each sale entirely by himself or herself and does not share sales skills or customer information because he or she does not want to risk having customers "stolen" by colleagues. In early 1998, the Taiwanese Labor Law amendment required that, by the end of 1998, companies pay a minimum wage (around U.S.$466 per month) for all forms of employee-employer relationships.-^ At that time, the CEO of the field site learned that some competitors would superficially comply with the law's requirement by providing a base salary while maintaining the same commission rates. However, they would potentially impose a penalty by taking away the base salary from employees not meeting the minimum sales requirement. Family competitors (i.e., those selling the same brand of cars) and two nonfamily competitors (i.e., those selling different car brands but competing for the same customer groups) did not change their compensation contracts after the implementation of the new regulation."^ The CEOs of these companies were concerned about the possible negative impacts of a switch to a less performance-sensitive scheme on individual productivity as well as a potential loss of good employees.^ The Taiwanese government has not strictly enforced all of the new regulations, and the penalty associated with a failure to comply with the regulations is not severe. As such, most companies would rather risk paying the noncompliance penalty than suffer a short-term loss due to the change of compensation plan. However, the situation is different for our field site because it is the largest dealership in Taiwan, which makes it the most likely target for government investigation to see whether the industry is complying with the law. After considering all factors (e.g., potential government investigation and penalty, high political costs, etc.), the CEO decided not to follow competitors' actions, but instead chose to comply with the law's requirement by adding a base salary and reducing commission rates for all levels of employees. The reason for lowering commission rates was to keep total compensation expenses at the pre-plan change level.^ In May 1998, the field site announced the new plan, which became effective on July 1, 1998. This plan was implemented throughout the company rather than on a regional or phased-in basis. As such, we found a unique setting to study employee recruitment and turnover because the field site and some of its competitors used different compensation plans in response to the government regulation. Furthermore, the dealership introduced an Improvement of Low Performance Plan with a minimum sales requirement of nine cars in three months or a monthly average of three cars during a three-month period. This minimum requirement may make the new plan riskier for poorly performing employees. The CEO indicated that the minimum requirement was based on a detailed analysis of employees' past productivity and expected additional compensation expenses due to the imposed bar. ' Because of the seasonal nature of the car industry, the dealership determines whether an employee will be terminated using the minimum requirement of nine CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

171

cars every three months instead of three cars each month. This average was instituted because in Taiwan, the car industry normally performs best in January when individuals receive their annual bonus and celebrate the Chinese new year holiday. Conversely, the industry's worst month is August, perhaps because some Taiwanese believe it is unlucky to conduct any major activities, including making personal decisions such as getting married, during this period. So, for example, if an employee sells zero, zero, and nine cars in November, December, and January, respectively, he or she will not be terminated in February or Marcb because he or sbe sold nine cars in January. If an employee does not attain the minimum requirement after three months, he or she receives an initial warning and will be terminated if bis or her performance does not improve in the following two months. We chose this field site for three reasons: (a) tbe change in its compensation scheme from totally commission-based to a less performance-sensitive plan (base salary plus lower commission rates) was not voluntary, but in response to the 1998 law's requirements;^ (b) it is the largest dealership in tbe Taiwanese automobile market, with nearly 18 percent of tbe market share over the past five years: and (c) tbe company agreed to provide us with an extensive arcbival database containing 4.392 sales representatives and 87 branches over a 56-month period. Top management also allowed us to review most documents and to interview managers and sales force personnel. 3. Hypotheses development The premise of agency theory is that a principal designs a contract to align an agent's actions witb the principal's best interests (e.g., Holmstrom 1979; Feltham and Xie 1994; Prendergast 1999; Lambert 2001). Classic agency models bave focused primarily on the design of performance-based contracts to motivate employee effort (i.e.. incentive effect) (e.g.. Baker et al. 1988; Baiman 1990; Milgrom and Roberts 1992). Also, prior studies bave suggested selection effects, where performance-based compensation contracts can attract and retain bigh performers (e.g., Milgrom and Roberts 1992; Baron and Kreps 1999; Lazear 2000; Banker et al. 2001). Below, we discuss both tbe incentive and adverse selection effects and the related bypotbeses to be tested. Incentive effect Agency theory identifies the circumstances under wbicb fixed (base salary) and variable (commissions and other forms such as bonuses or options) components can better align shareholder and worker interests. In tbe field site's old commissionbased compensation plan, the linear individual contracts take tbe form y = bx, wbere b is commission rate and x is productivity or output. Under tbis old contract, C, the employee receives compensation for each unit sold and gets zero compensation if no cars are sold. Therefore, compensation is sensitive to botb x and b, and employees bave an incentive to work hard to increase pay. The field site tben changed tbe compensation plan C to a new one, C , witb base salary and lower commission rates. Linear individual contracts under the new pian are of the form CARVol 26 No. 1 (Spring 2009)

172

Contemporary Accounting Research

y' = a + b'x, where a is base salary, b' is the new commission rate, and b' < b. The graphical presentations of the old and new compensation contracts, C and C . are summarized in Figure 1. Since the company implemented the Improvement of Low Performance Plan, an employee must sell al least nine cars every three months (or an average of three cars each month) to remain on the job. The fixed pay under the new plan is technically not a guaranteed salary. In essence, the new plan can be thought of as consisting of two components: "a" is paid when employees sell a minimum of an average of three cars per month, and "Z?" is paid per unit after the employees achieve the minimum requirement and are paid for all the units sold. As pointed out by Demski and Feltham 1978. the requirement of achieving a minimum number of car sales imposes some risk on the employees as an inducement to expend some agreed level of effort. Empirical studies have supported the effects of performance-based incentives on employee behavior (e.g.. Banker. Lee, and Potter 1996; Bailey, Brown, and Coceo 1998; Lazear 2000). Banker et al. (1996) report that implementitig a performancebased compensation plan increases sales and that the effect persists and grows over time. Bailey et al. (1998) find that improvement rates in individuals' initial and overall performances of an assembly task are higher when an incentive plan is in place. Similarly, Lazear (2(X)0) examines the impact of piece rates on the performance of workers who install auto windshields. He finds that worker output increased after the company switched its compensation scheme from hourly wages to a piece rate plan. Conversely, these findings suggest that if a company switches from contract C (performance-sensitive plan) to contract C (less performance-sensitive plan), a lower h' may induce less effort from employees to work hard, resulting in.' < jc. In light of the arguments and empirical findings that performance-based incentives affect employee behavior, our first hypothesis is as follows:

Figure 1 Compensation and sales productivity under the commission-based and base salary plus lower commission plans Compensation (y) Commission-based plan ( Q Base salary plus commission plan (C)

Number of cars
sold (jc)

Minimum requirement
CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention


HYPOTHESIS

173

1. Employees ' average sales productivity (cars sold) will decrease after the company changes its compensation plan to a less performancesensitive scheme.

Figure 1 shows the relationship between compensation and cars sold under both the C and C compensation schemes. Clearly, when sales productivity increases, the differences in compensation {y y') under the two schemes increase, which suggests that switching from compensation plan C to C will cause higher-performing employees to lose more compensation (due to lower commission rates) than lower-performing employees. In light of this greater compensation loss, higher-performing employees will reallocate their efforts, expend less elfort, and enjoy more leisure time than lower-performing employees, which will result in comparatively lower sales productivity, leading to the following hypothesis: HYPOTHESIS 2. The decrease in sales productivity (cars sold) of higher-performing employees after a company changes to a less performance-sensitive compensation scheme will be greater than that of lower-performing employees. Adverse selection effects Companies design compensation schemes not only to induce more employee effort but also to attract and retain high-potential employees (Baker et al. 1988). Adverse selection effects include recruiting and tumover effects. The former relates to tbe type of employees who join the company and the latter to the type of employees who leave. For example, contracts with higher piece rates attract high-ability employees (Lazear 2000), and adding a performance-based bonus attracts higherability employees (Banker et al. 2001). Prendergast (1999) argues that compensation contracts are an important means for a company to recruit higher-ability workers. Employing a perfonTiance-sen.sitve contract will benefit high-ability workers more than low-ability workers, resulting in an increase in the percentage of high-ability employees who are attracted to the company. Conversely, when a company changes compensation schemes from C to C , it is likely that the guarantee of a base salary will attract lower-ability employees. Therefore, the recruiting effect suggests that the sales productivity of employees hired after the change to contract C will be lower than that of employees hired before the plan change. These aiguments lead to the following hypothesis; HYPOTHESIS 3. The average sales productivity (cars sold) of employees hired under the less performance-sensitive compensation scheme will be less than that of employees hired under the performance-sensitive compensation scheme. Hollenbeck and Williams (1986) point out an important distinction between the frequency and functionality of turnover. While the former refers to tbe number of turnovers, the latter refers to the implication of those turnovers for an organization. Using a meta-analysis of 55 studies, Williams and Livingstone (1994) report CARVol 26 No. ! (Spring 2009)

174

Contemporary Accounting Research

that poorer performers tended to leave when pay was ba.sed on performance and to stay wben it was not. The implication of their findings is that performance-based compensation plans can lead to functional turnover. Lazear (2000) reports that about one-third of improved performance can be attributed to selection effects (i.e., less productive workers leave the company and are replaced by more productive workers). In Banker et al. 2001, results sbow that a performance-based scheme in a retail firm attracts and retains more productive employees, while the performance of the less productive sales staff declines before they leave. When the company changes from a linear system to a less performance-based system scheme with a kink that is, a nonlinear segment in the pay-for-performance function it is likely to create turnover among employees (Jensen 2003).** Feedback on compensation could help employees learn how they are affected by a less performance-sensitive compensation plan. Higher-performing employees who continually compare tbeir compensation with past levels and outside opportunities are likely to be dissatisfied with tbeir compensation. As discussed earlier, some of tbe field site's competitors did not lower tbeir commission rates. It is likely that some unsatisfied higb-performing employees cbose to join tbese competitors to increase their compensation. Tbis is consistent witb Jensen's 2003 argument tbat if the commission rate does not sufficiently reward employees for tbeir extra effort, the company will likely lose high performers to competing firms. Therefore, we expect the sales productivity of employees wbo were hired under contract C and left under contract C to be higher than that of employees wbo left before the plan change. Regarding employees wbo were hired and left under contract C. their departure may be attributed to either disappointment witb tbe compensation or being forced out by the Improvement of Low Performance Plan. Because these two groups' sales productivity may offset each other, we expect no difference in sales productivity between those who were hired and left under contract C and tbose who left under contract C. Taken together, we expect sales productivity of etnployees who left after the change in compensation scheme from C to C to be higher than that of employees who left before the change. We propose the foiiowing hypothesis concerning turnover effects: i HYPOTHESIS 4. The average sales productivity (cars sold) of employees who leave under the ess performance-sensitive compensation scheme will be higher than that of employees who leave under the performance-sensitive compensation scheme. 4. Model development In this section, we discuss tbe regression models developed to test our hypotheses. Because employee productivity (i.e.. number of cars sold per person-month) is a non-negative, integer-valued, random variable, the use of ordinary least squares regression may result in biased, inefficient, and inconsistent estimates (Cameron and Trivedi 1998; Long 1997). Therefore, we use nonlinear models (i.e.. Poisson regression or negative binomial regression). Poisson regression models assume an equality of mean and variance (Greene 2(X)0). Because ouj analysis shows an overly CAJ? Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

175

diverse distribution (i.e., the true variance is greater than the mean), we develop a negative binominal (NB) regression model to examine the impact of the compensation plan change on employee productivity. The sampling period covers 97.541 person-months and 5,121 branch-months of data from January 1996 to April 2001 (56 months).^ We partition the sample into two subsamples: (a) before the change to the new plan (a total of 28 months from January 1996 to April 1998),'** and (b) after the change to the new pian (a total of 28 months from January 1999 to April 2001 ). The models' parameters are estimated using pooled time-series data over a 56-month period and crosssectional data for the 4,392 individuals and 87 branches. Below is the NB regression model:

-h \^^B_EMPLOY, + ^M^,+
m = I

s,

(1),

where / = {1, ... , 4,392}. a subscript to denote employee i, = {1.... , 56}. a subscript to denote the time period in which the employee works, and ei is a random error term. We measure employee sales productivity iIND_PROD) as the number of cars sold per person-month. PLAN is equal to 1 if the employee was on the new plan during the given month. Hypothesis 1 (hypothesized incentive effect on sales productivity) is tested by examining whether the coefficient of PIAN (aj) is negative in (1). Also, to rule out the possibility that the effect on sales productivity was due to employees' being hired at different times, we include EXPER (number of months employed) as a control variable. In the model, we also include two types of competition control variables (competition type and competition intensity) to capture the general economic conditions. There are two types of competitors in our field site: family competitors {FAM_COMP) and nonfamily competitors {NON_FAM_COMP). Regarding competition intensity, we use CITYJSALE (i.e., total monthly sales of competitors in the same city) as a proxy. The log forms of monthly car sales for FAM^COMP, NON_FAM_COMP, and CITY_SALE are used. We expect positive coefficients for both FAM_COMP and NON_FAM_COMP because good economic conditions help employees sell cars more easily, thereby increasing their productivity. However, we cannot predict the sign of CITY_SALE because it not only captures the competition intensity among car dealers in a specific city, but also reflects the purchasing power of that city. Although high competition intensity negatively affects individual employee productivity, high purchasing power has a positive effect. We also include month dummies, M^,, in the models to capture both seasonality effects and "new car model effects" (e.g., when a car dealership launches a new car model, it attracts consumers' attention, resulting in increased sales productivity that nonnally lasts for several months). Because of the seasonal nature of car sales discussed earlier, we expect M^ to display different signs in different months. All CARWol 26 No. 1 (Spring 2(X)9)

176

Contemporary Accounting Researcb

four control variables (FAM^COMP, NON_FAM_COMP, CITY^SALE, Mj) are at the market level. Finally, to control for tbe impact of hiring additional salespeople on productivity, we include a brancb-level control variable, ^B_EMPLOY (i.e., tbe log form of number of employees in eacb month). We expect a negative sign for #B_EMPLOY because a bigher number of employees implies more intense competition among tbe employees, resulting in lower individual productivity. A list of variable definitions is included in tbe Appendix. Because ( 1 ) explores how the compensation plan change affects all employees, we do not include employees' ability. To examine whether the bigher-ability employees are more likely to be affected by the plan change, we include employees' ability (ABILITY) in (2). = OQ + a^PLANi, + yEXPERi, + Yj, X PLANi, + ^FAM^COMP,
11

+ ^
m= 1

M^, + Si,

(2).

We take three steps to measure employee ability. First, we collect different measures that can capture employee ability: the annual performance rating, tbe number of cars sold prior to tbe month of interest, and tbe reciprocal of time to tbe first promotion since they joined the dealership. Gibbs (1995) argues that tbe employee performance rating is a logical proxy for expected ability and that expected ability declines monotonically with tbe number of promotions employees receive over time. In tbe dealership, each employee is evaluated by the branch manager annually on a four-point scale, with "4" being the best performance. Tbe average number of cars sold reflects changes in an employee's productivity from the first month in whicb the employee joined the dealership to the preceding month of interest. Similar to Gibbs 1995, we also use tbe reciprocal of time to the first promotion since tbe employees joined the dealership, as the tbird measure. Because of the high turnover rate, we do not use tbe number of promotions employees received over time. Second, to preserve a degree of freedom, we use factor analysis to collapse our three measures into one variable ABILITY, which does not vary between tbe pre- and post-new-plan periods." Finally, we use the rank order o ABILITY to classify employees into the following three groups: HIGH^ABIUTY (in the first quartile), MEDIUM_ABILITY (in tbe second and tbird quartiles). and LOW_ABILITY {\n the fourth quartile). Hypothesis 2 predicts a negative sign of the coefficient oABIUTY X PLAN {^). Also, we classify the sample on the basis of the type of compensation plan and on the timing of employee recruitment and turnover as follows: OldPlan {D^^ = 1 , 0 otherwise): employees who joined tbe dealership before the change to the new plan (July 1. 1998).

CA/? Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

177

OldPlanQuitBefore {D^Q^ = 1 , 0 otherwise): employees who joined the dealership before the change to the new plan and left before July 1, 1998. OldPlanQuitAfter (D^Q^ = 1 , 0 otherwise): employees who joined the dealership before the change to the new plan and left after July 1, 1998 but before the end of our sample period. NewPlan {D^^^ = 1 . 0 otherwise): employees who joined the dealership after the change to the new plan.

NewPlanQuitAfter {D^Q^ = 1,0 otherwise): employees who joined the dealership after the change to the new plan and left after July I, 1998. but before the end of our sample period. To avoid potentially large variation in individual monthly data, we use the moving average of sales productivity to capture trend and direction. Figure 2 displays the three-month moving average of sales productivity for different employee types from January 1996 to April 2001. As seen in panel A of Figure 2. the average sales productivity of the old plan is better than that of the new plan, suggesting that the less performance-sensitive plan attracts more low performers. Furthermore, panel B of Figure 2 shows that the average sales productivity of OldPlanQuitAfter is higher than that of OldPlanQuitBefore, implying that more high performers left the firm after the plan change. To test hypothesized adverse selection effects, we specify the following NB regressions (3) and (4): IND_PROD, = 0
11

^
m = 1

^M^, + e,,

(3),

IND_PRODi, = 00 + iEXPERj, + r2D^Q'^ + r^D^QA -\- \yFAM_COMP,


It

^iM^^+Bi,
m =1

(4).

Model (3) tests the recruiting effect (i.e., comparing the sales productivity of old-plan and new-plan groups). We test Hypothesis 3 by examining the coefficient of D^^^ (ri) of (3) and we predict r^ to be negative. Model (4) tests turnover effects. Therefore, only employees who left the dealership (i.e., D^Q^, D^Q^, and D^QA) are included. Hypothesis 4 predicts that the average sales productivity of OldPlanQuitBefore will be less than that of OldPlanQuitAfier, while there will be

CAR\ol

26 No. I (Spring 2009)

178 Figure 2

Contemporary Accounting Research Sales productivity by timing of employee employment and tumover

Panel A: Employees by different time of employment OldPlan NewPlan

S a
u

6-

o
S 5-

4-

9 _

M
~T 1 r 1 1 1 1 -30-26-22-18-14-10 -6 -2 r 3 1 1 1 1 1 1 r7 11 15 t9 23 27 31

Month before (-30 t o - 1 )/after PLAN Panel B: Employees by different time of tumover
7-1

65

- OldPlanQuitBefore - OldPlanQuitAfter " NewPlanQuitAfler

21n [ I I I I 1 1 1 1 1 \ \ 1 1 r

|Vv'\

-30-26-22-18-14-10 - 6 - 2

II 1 1 23 27 31 5 9

Month before (-30 to -1) / after PLAN


Notes:

Data covers the 64-month period from January 1996 to April 2001, month - 3 0 to - 1 before PLAN and month 1 to 34 after PLAN. As indicated in endnote 11, we excluded eight months (May to December) of 1998 from our sample in order to control both the transition-period and seasonal-month-sales effects. The variables are as defined in the Appendix. CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

179

no difference in sales productivity between NewPlanQuitAfter and OldPlanQuitBefore. Thus, we expect that the coefficient of r^ will be greater than zero and r2 will be equal to zero. In botb (3) and (4), we do not include ABILITY, PLAN, or tbeir interaction variable, because in these models we test only for the differences between the groups (i.e.. new plan versus old plan. D^Q^ versus D^Q^, and D^^^ versus D^Q^). Because employees may have different times of employment and departure, we also include EXPER as a control variable. 5. Results Descriptive statistics Table 1 summarizes the descriptive statistics of tbe key variables. Means (standard deviations) for tbe entire period and the two subperiods are presented in panel A of Table 1. The average number of cars sold per month per employee {IND_PROD) is 3.30, wbicb is slightly higher than 3. the minimum requirement under the Improvement of Low Performance Plan. The turnover rate per month (MTHJTURNOVER) is 2.31 percent (or 27.45 percent per year), which is the lowest rate in the car sales industry.'2 This high turnover rate suggests that we are able to examine adverse selection effects. Further analysis shows that the turnover rate for employees without dismissal threat is only 0.34 percent, which is similar to other industries. On the other band, the turnover rate for employees with dismissal threat is 1.97 percent. Furthermore, the turnover rate increases slightly for both employee subgroups after tbe plan change; however, the difference is statistically insignificant. Also, the average age of employees hired under the new plan (32.88) is older than tbat under tbe old plan (31.46). As seen in panel A of Table 1, the average of IND_PROD under tbe new plan (3.15) is significantly (p < 0.0001) lower tban that under the old plan (3.49), suggesting tbat the change to a less performance-sensitive compensation plan decreased individual sales productivity. Under the new plan, the fixed pay is $466 and tbe average commission rate is $225 (s.d. = $250). which is significantly lower than tbat under tbe old plan (mean = $495; s.d. = $287;p < O.OOOl).'^ This shows tbat the plan change causes a significant decline in the average monthly compensation from $1,612 to $1,255 ($357; p < 0.0001), which can bave a significant impact on employees' living standards.''* Furthermore, these results suggest tbat the Improvement of Low Performance Plan was effective. Eor example, as seen in panel A of Table 1, the standard deviation of IND_PROD significantly decreased (from 4.09 to 3.45; p < 0.0001) after the change in compensation plans. Also, additional data sbow that, in general, the average sales productivity of NewPlanQititAfter is below tbree cars over the 34-month period. Eurtbermore, tbe monthly employee turnover rates rose from 2.19 to 2.41 percent (a 10 percent increase), but the difference is not statistically significant (f = LOI, p = 0.3152). Note that there is also a decrease in tbe company's market share after tbe plan switch (from 18.2 percent to 17.0 percent); however, this decrease is not statistically significant. Panel A of Table 1 sbows that sales of family competitors increased significantly (from 4.68 to 7.88), but sales decreased in the same city (C1TY_SALE; from

CA Vol. 26 No. 1 (Spring 2009)

180

Contemporary Accounting Research

1 a
a

Xi

a. O

11

.1
o -s

o
BU

3
JO p B
I

"a

a
U
es

"o
c
>~

._ 73
e
iS

._ 'S
es
h3

.5 " O
c
H

i . " i
"5 c
n

B
p

cnp

CO

a tr

ion

1
o p

a o a. C
<

03 a
< C O

P Q CAR Vol. 26 No. I (Spring 2(X)9)

i! i i 8 s'
o

I i

Employee Performance, Recruitment, and Retention

181

i
Q

o o o o -' -'

r - o o t o t s m " - ' l i l i l

a. o O

r--r'i(Noor-.(SOv'ooopN\oo\
o o r - ^ o ^ ' t ^

C S -

- o O^^ ^ N r #o? '^i ^ " ' t ' ^ J - " t^ o o & ' O ^ ^ o / io C N o

I .
^ .2 o t

o
[I.

u
o Z

CQ

iS

CA Vol. 26 No. 1 (Spring 2009)

182

Contemporary Accounting Research

^ r n c s O
a

Q.

I
13

a. <
I
ON

Os

a. n "
S
T3

u
C

1 c
Q.

o O

T3

p p n

i
m r i *

2 2
A

CA Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

183

\^

00

" V -I: <^

TT

o :

00

NI

(/5 de

' _r rn rj, a -

E
CO

t^

;orl are as deti ned i arenth eses.


0

97,;541 personperiod

s: c

D.

here: 56-

>

a.

%
o

i
o

++

. 26 No. 1 (Spring 2009)

tatisti

tatisti

statis

ariab

o E

.3
o

"3 u

ignific ant al:the

i'

iI

I i

1^

s s

'S
o

o
tn

S. a.

a
un
to

ignific

184

Contemporary Accounting Research

7.50 to 7.39) and also for nonfamily competitors (NON_FAM_COMP; from 10.01 to 9.96). Panel B of Table 1 presents tbe distribution of individual sales productivity for both the old-plan and new-plan periods. These are the actual numbers of cars sold by the employees, including those wbo sell more than 10 cars. As shown in panel B of Table 1. the change to a less performance-sensitive plan caused a decrease in sales productivity. Eor example, there are more employee-montbs with no cars sold under the new plan (24.8 percent) than under the old plan (20.1 percent). However, there is a slightly higher percentage of employee-months where three cars were sold under tbe new plan than under the old plan ( 15.7 percent versus 13.7 percent). This suggests that lower-ability employees exerted greater effort to meet tbe minimum sales burdle to avoid job loss. Regarding the high performers, after the plan change there was a decrease (from 20.6 percent to 17.3 percent) in the percentage of employee-months wbere six or more cars were sold and also a slight decrease in the employee-months where 10 or more cars were sold (from 4.5 percent to 3.4 percent). It is possible tbat lower commission rates discouraged top employees from exerting effort. To examine tbe impact of the plan cbange on different groups, we present means (standard deviations) of cars sold and compensation by ability group in panel C of Table 1. These results sbow that the new plan has a negative impact on sales productivity for the HIGH_ABILTTY group (from 6.2 to 5.9; p = 0.051). and a positive impact on the MEDIUM_ABILnYemployees (from 3.3 to 3.4;;; < 0.00! ). MEDIUM_ABILITY emp\oyeea may have worked harder to compensate for tbeir loss in income due to lower commission rates. Although the average sales productivity for the LOW_ABILITY group is still less tban tbe minimum requirement of three cars after the plan change, ii increases from 1.3 to 1.7 (/? < O.OOOI), Nonetheless, all three groups have significant decreases In compensation ($518 for HIGH_ABILITY; $303 for MEDIUM_ABIUTY; $92 for LOW_ABILITY). Tests of incentive effects (Hypothesis i and Hypothesis 2) Hypothesis 1 predicts tbat employees' average sales productivity will decrease after the plan cbange. Panels A and B of Table 2 show NB regression results with individual sales productivity (ND_PROD) as the dependent variable. As discussed above, we expect a negative coefficient of PLAN (a,). Panel A of Table 2 (column 1) shows that aj is negative and significant (-0.153;/) < O.OOOI).'-'^ Recall that the dealership also introduced an Improvement of Low Performance Plan. Therefore, the overall impact of contract C on sales productivity may not necessarily be negative because the Improvement of Low Performance Plan forces lowest-performing employees facing a threat of dismissal to work harder (Baker et al. 1988). To further examine the possible effects of threat of dismissal on sales productivity, we classify our sample into two groups according to wbetber tbe individual's average sales in the last three months working in the dealership was above or below tbree cars (i.e., a requirement under the Improvement of Low Performance Plan). As shown in column 2, the coefficient of PLAN for tbe subsample without dismissal threat is -0.132 (p < 0.0001). Interestingly, column 3 shows the coefficient of PLAN for tbe subsample with dismissal threat ( -0.027) to be negative, CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

185

but it is not significantly less than zero (p = 0.0923). This suggests that the threat of dismissal under the Improvement of Low Performance Plan does not improve the productivity of employees with poor performance. One explanation is that individual productivity is jointly determined by effort and skills (Demski and Feltham 1978; Gibbs 1995), and incentives do not aiways work when the tasks require skills TABLE 2 Negative binomial regression results of hypotheses testing (dependent variable: ND^PROD') Panel A: Test of incentive effects {Hypothesis 1 ) The plan change decreases average sales productivity Model 1 Subsample without dismissal threat (2) 40,082 -5.149+ -0.132t 0.017+ 0.02+ 0.568+ 0.057+ -0.033* 21,961 0.0001 Subsample with dismissal threat (3) 36,768 -4.615+ -0.027 0.008* 0.032+ 0.446+ -0.035+ 0.061* 10,177 0.0001

Full sample (1) n Intercept PLAN EXPER EAMJCOMP NON_FAM_COMP CITY_SALE m_EMPLOY Goodness-of-fit (LR statistic) p-value 76.850 -5.290f -0.153+ 0.062t 0.023t 0.524+ 0.046+ -0.005 47,135 0.0001

Panel B: Test of incentive effects (Hypothesis 1) Short-term versus long-term effects of the plan change Model I Sample n Intercept PLAN EXPER FAM_COMP NON_FAMj:OMP CTY_SALE m_EMPLOY Goodness-of-fit (LR statistic) p-value Adjustment period (from - 2 to 16 months) 62.988 -7.3951 + -0.0794+ 0.0701 + 0.025 It 0.7449+ 0.0396+ 0.0106 9,854 0.0001 Status quo period (from 17 to 34 months) 60,915 -5.2658+ -0.1514+ 0.0614+ 0.0242+ 0.5159+ 0.0566+ -0.0144 11,511 0.0001

(The table i.s continued on the next page.) CARWoX. 26 No. 1 (Spring 2009)

186

Contemporary Accounting Research

TABLE 2 (Continued) Panel C: Test of incentive effects Hypothesis 2 (model 2 for the effects of the plan change on employee groups with different levels of ability) Model 2 n Intercept PLAN ABILITY ABIUTY*PLAN EXPER FAM_COMP NON_FAMjCOMP CITY^SALE m_EMPLOY Goodness-of-fit (LR statistic)
/?-value 73,239

-5.749t
0.225t

1.987t
-0.427t -0.013t O.O35t

0.495t 0.05 It -0.028

26,000
0.0001

Panel D: Tests of Hypothesis 3 (model 3 for recruiting effect) and Hypothesis 4 {model 4 for turnover effects) Model 3 Model 4
Turnover sample without retirement possibility

Sample n Intercept
[)NEW
QNQA

Full sample (1) 76,850 -6.086t -0.226t

T\imover sample

(2)
37,670 -I2.360t 0.013 0.484t

(3)
37,568 -12.299t 0.014 0.484t 0.046t -0.018t

Turnover sample without dismissal threat (4) 9.534 -10.078t 0.064 0.231 + -O.O34t -0.0011 1.108t 0.014 -0.114t 1,660 0.0001

OOQA

EXPER EAMJOOMP NON_FAM_COMP CITY_SALE #B_EMPLOY Goodness-of-fit (LR statistic)


/j-value

0.053t 0.018t 0.6 i Ot 0.046t

0.045t -O.OI8t 1.225^ 0.0135 -0.0305 6.566 0.0001

-o.oio
14,255 O.OOOi

1.2I9 0.0138 -0.033 6,546 0.0001

(The table is continued on the next page.)

CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention TABLE 2 (Continued) Notes: * Variables are as defined in the Appendix. t Statistically significant at or less than the 0.0001 level (two-tailed). t Statistically significant at the 0.01 level (two-tailed). Statistically significant at the 0.05 level (two-tailed).

187

or experience (Ashton 1990; Merchant, Van der Stede, and Zheng 2003). Skills are normally developed over time and cannot be changed in the short run. As such, while the new plan may encourage employees with a threat of dismissal to exert more effort, the increased effort may not always translate into an increase in sales productivity if the required skills are not possessed. Furthermore, one branch manager made a comment that the new compensation contract is more risky for junior employees than for senior employees. This is especially the case for employees who recently joined the car sales industry. Therefore, the minimum requirement for car sales may impose more risk for poorly performing employees but it did not improve productivity. Collectively, our results suggest that the new plan had a negative effect on employees and therefore the results support Hypothesis 1. Some studies in the organizational behavior literature indicate initial reactions to changes in compensation plans that tend to stabilize over time (Lazear 2000). Therefore, we also analyze the short-term (or adjustment period) versus long-term (status quo period) effects of the change in compensation plan on individual productivity. Specifically, we separate the sample period into two stages: initial adjustment period ( - 2 through 16 months) and status quo period (17 to 34 months). As shown in panel B of Table 2, in both subperiods we observe significant negative coefficients of PLAN (-0.0794 in the initial adjustment period and -0.1514in the status quo period).'^ These results suggest that the negative effects of the plan change are stronger in the status quo period, which is consistent with Banker et al. 1996, who report that the compensation plan change effect grew over time. Hypothesis 2 predicts that the magnitude of the negative impact of the compensation plan change on sales productivity is greater for higher-performing employees than it is for lower-performing employees. Therefore, we expect the coefficient of ABILITY X PLAN, j, to be negative. As discussed earlier, ABILITY is derived through a factor analysis and is a continuous variable. Panel C of Table 2 shows that the coefficient o ABILITY X PLAN has a significant (p < 0.0001 ) negative sign (-0.427), consistent with the prediction of Hypothesis 2.'"' Therefore, our results support Hypothesis 2 in that the plan change has a more negative impact on higher-ability employees than on lower-ability employees. Tests of the adverse selection effect (Hypothesis 3 and Hypothesis 4) To test for the recruiting effect (Hypothesis 3), we examine whether sales productivity of employees hired under the less performance-sensitive new plan is lower CARVdi. 26 No. 1 (Spring 2009)

188

Contemporary Accounting Research

than that of employees hired under the old plan.'8 in (3), we include only avoid potential multicollinearity problems. The average sales productivity is o, and the average sales productivity of D^^^ is Q -I- y,. Thus, Hypothesis 3 is tested by examining the following inequality: (a^ + Ti) - o = Ti < 0. As seen in column 1 of Table 2, panel D, yi is -0.226, which is significantly less than zero (at the 0.0001 level), which suggests that sales productivity was higher for employees hired under the old plan than for those hired under the new plan, thereby supporting Hypothesis 3. To test turnover effects (Hypothesis 4), we first compare the differences in the sales productivity of those who left before the change in compensation plans (OldPlanQuitBefore, D^QB) with those of employees who were hired under the old plan and left after the change {OldPlanQuitAfter, D^Q^). We expect that the coefficient of average sales productivity oD<^Q^ (Q) will be less than that of 0^0-4 (^Q -I- ^3). In other words, jy will be greater than zero. As shown in column 2 of Table 2, panel D, 73 is 0.484 (significantly greater than 0 at the 0.0001 level).'^ Recall that the higher 73, the higher the chance that the dealership will lose higher-performing employees by implementing the new plan. We then compare the sales productivity of DOQB with that of NewPlanQuitAfter (D^Q^ ) (i.e., those who were hired under the new plan). Recall that we expect the coefficient of average sales productivities of D^^ (Q + 72) will not be different from that of D^ (OQ) (i.e., 73 w' not be different from zero). As shown in column 2 of panel D, 73 (0.013) is not significantly greater than zero. Employee tumover may not be caused by disappointment with the compensation plan change, but instead by retirement or threat of dismissal. While dismissal is related to poor performance, retirement is not. Furthermore, after working for 25 years employees can decide when they want to retire, but dismissed employees are forced to leave immediately after they receive the dismissal notice. Thus, to rule out the possibility that our tumover results may be confounded by employee retirement and threat of dismissal, we conduct two additional analyses to measure these potential effects on tumover. Our interviews with managers indicate that employees can receive all retirement benefits if they serve tbe company for 25 years. Tbey can receive partial retirement benefits if they serve the company at least 10 years. Because the company was established in 1975, fewer than five employees were eligible to receive full retirement benefits in 2001. Therefore, we use employee tenure of more than 10 years as a proxy for possible retirement to further separate tbe sample. As seen in column 3 of panel D, the results for the tumover sample without retirement possibility are similar to those of the entire tumover sample. Specifically, 73 is positive and significant (0.484;/^ < O.OOOi) and72 (0.014) is not significantly greater than zero. We also exclude employees with dismissal threat from the sample. Column 4 of panel D (Table 2) shows that tbe tumover sample without dismissal threat has results that are qualitatively similar to tbe entire tumover sample. 73 is positive and significant ( 0 . 2 3 l ; p < O.OOOI) and 72 (0.064) is not significantly greater than zero. Taken together, these results support Hypothesis 4. As discussed earlier, the new compensafion plan is likely to cause greater loss to higher-performing employees, which may lead to their departure from the company. CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

189

Also, the new compensation plan may force lower-performing employees to leave due to tbeir lack of ability and experience. To examine the effects of compensation loss (COMP_LOSS) and an employee's experience (EXPER) and ability (ABILITY) on employee departures, with and witbout the threat of dismissal, we ran logit regression models separately for tbese two employee groups. COMP_LOSS is the difference between the average new compensation minus the average old compensation, divided by the average old compensation. Therefore, we only include employees who joined the dealership in the old plan and separate them into two groups: employees witb and without dismissal threat. Also, we include 0B_EMPLOY, a brancb-level variable, because a higher number of employees enhances the intensity of competition among the employees, which may increase tbe likelihood of employee turnover. As such, we expect a positive sign of m_EMPLOY. Table 3 summarizes tbe logit regression results. Model A shows that wben employees face a threat of dismissal, their departure decisions are not influenced by compensation loss but their level of experience does matter (-0.392; p < 0.0001). Conversely, as seen in model B, departure decisions by employees who are not facing the threat of dismissal are affected by botb compensation loss (-2.329; p < 0.0(X)1 ) and experience (-0.252; p < 0.0001). In sum, our additional analyses provide deeper insights into Hypothesis 4. The higher average sales productivity of employees who leave under tbe new plan versus the old plan is attributable more to the departure of high-ability employees because of greater compensation loss than to the departure of employees witb less experience and lower ability.

TABLE 3 Results of logit regression of employee turnover (dependent variable: LEFT) Model A
with dismissal threat

Model B
without dismissal threat

291 (LEFT = 1) 80 (LEFT = 0) Intercept


COMF_LOSS*

149(L/T = 1) 472 (LEFT = 0) -1.082 -2.329' -0.252" 0.22 0.305 0.1901

EXPER ABIUTY m_EMFLOY Max-rescaied R^


Notes:

1.232 -0.159 -0.392* 1.762 0.441 0.1687

Variables are as defined in the Appendix. * Statistically significant at or less than the 0.0001 levei (two-tailed). CARVoX. 26 No. I (Spring 2009)

190

Contemporary Accounting Research

Additional analysis of the impact of the change in compensation plan on company overall performance An optimal compensation contract needs to balance the benefits of increased firm performance resulting from additional employee efforts and the compensation costs incurred (Abowd 1990, 55). As discussed earlier, the change to a less performancesensitive compensation plan hurts individual sales productivity. Nonetheless, the decreased individual sales productivity may not result in lower overall company performance. This change occurs because the new compensation plan may bring cost savings, and management may take actions to mitigate the negative impact of decreased individual sales productivity (e.g., hiring more workers). Due to the lack of data, prior research has focused on individual effort and pay perfonnance (e.g., Banker et al. 2001) and also assumed a positive relationship between individual productivity and company performance (e.g., Lazear 2000). Because we were able to gain access to data at both individual and division levels, we also investigated the relationship between individua! productivity and overall company performance. To study that relationship, we collected additional company-level performance (PERFORM) data as follows: revenues (REVENUE), gross profits (GS_PROFIT), income before headquarter expenses allocation (INCOME), and number of cars sold (CAR_SOLD). To control for possible infiation over the sample period, all financial performance variables are deflated by the monthly consumer price index in Taiwan. Also, we include two competition-type control variables (FAMjCOMP and NON_FAM_COMP), the number of employees at the field site (i^F_EMPLOY), and sales mix of sedans versus trucks (SALES_MIX) to capture the general economic condition. Our model is: PERFORM, = 0 + (XiPLAN, + \^FAM^COMP, +
II

+ iSALES_MIX, + X^#F_EMPLOY, + V ^Ai^, + e,


HI = I

(5).

Table 4 summarizes the regression results. None of the coefficients for PLAN is significantly less than zero (except for CAR_SOLD), suggesting that the change in compensation plan does not negatively affect overall company performance. To shed light on why the negative effects of the plan change on individual productivity did not translate into poor company performance, we interviewed two branch managers. One manager attributed the lack of negative impact on the company's overall performance to hiring more employees and lower compensation costs. His explanation is consistent with our finding that on average there is a 7.81 percent increase (from 16.52 to 17.81) in ihc number of employees per branch after the plan change (panel A of Table 1). Furthermore, our additional analysis shows that the plan change lowered the ratios of compensation costs to both revenues (from 2.74 percent to 2.23 percent), and gross profit (from 31.58 percent to 28.25 percent). These results also support the manager's argument that lower compensation expenses mitigate the negative impact of decreased individual productixity un firm performance. CAR Vol. 26 No. I (Spring 2009)

Employee Performance, Recruitment, and Retention

191

Other possible explanations for the lack of negative impact on company performance include change in sales mix, more effort by etnployees leading to higher customer satisfaction ratings, and more repeated sales. Branch managers suggested that sales mix can be measured either by type of car (i.e.. high-margin sedans versus low-margin trucks) or by exclusive versus nonexclusive cars. Exclusive cars are those that can be sold only by the field site and nonexclusive cars are those that can be sold by both the field site and family competitors. In general, compared with exclusive cars, nonexclusive cars have higher profit margins and more intense competition. The managers we interviewed mentioned that in response to different levels of competition, the field site has significantly lowered the commission rates for exclusive cars but not for nonexclusive cars. However, the company only had sales data on exclusive and nonexclusive cars from September 1997 to April 2001. Results (not tabulated) based on these limited data show that the plan change caused a significant increase in sales of nonexclusive cars (from 40.71 percent to 50.24 percent; p < 0.001). This increase suggests that employees allocated more effort to selling higher-margin and more competitive nonexclusive cars under the new plan, which mitigates the loss resulting from selling fewer cars. Regarding the sales mix of sedans and trucks, our additional analyses (not tabulated) show that the percentage of sedans did not change significantly (from 73.43 percent under the old plan to 72.55 percent under the new plan;/? = 0.2547) during this sample period. TABLE 4 Regression results at the company level (n 64)*
Dependent variablet Intercept PLAN FA M COMP NON FAM COMP SALES MIX F EMPLOY Adjtisted R~ Notes: We carry out the Diirbin-Watson fDW) test for AR 1 that is, first-order serial correlation. Our results show that DW is close to 2, and there is no serious firstorder serial correlation. Also, we find that all variance inflation factors at^ less thnn 5. TTiis stiggests there are no serious multicollinearity problems (Maddala 1992: Hair, Anderson, Tatham, and Black 1998: Greene 2000, 257). The variables are as defined in the Appendix. Statistically significant at or less than the 0.0001 level (two-tailed). Statistically significant at the 0.01 level (two-tailed). Statistically significant at the 0.05 level (two-tailed). CAR SOLD (1) -30,084*
-465

REVENUE

GS PROFiT

(2)
-57.395* 2.956 2.568* 56,182* -49,607 58* 0.877

0)
-37.0485 -436 215" 3,174 1,481
5

INCOME (4J -5.611 -3.52 44 385 -2,270 2 0.354

15 3,5lli -7.701 3* 0.902

0.665

CA Vol. 26 Nu. 1 (Spring 2009)

192

Contemporary Accounting Research

6. Concluding remarks Because of a lack of access to objective, individual-level performance data, prior empirical studies of compensation schemes have been limited in their ability to further our knowledge of how an incentive plan affects employees' selection of employment and effort level (Banker et al. 2001). In this study, we have used data at both the individual and company levels to empirically test how a company's change to a less performance-sensitive compensation plan affected employee performance, recruitment, and retention. Our empirical results support predictions based on economic theory that a change to a less performance-sensitive compensation scheme reduces individual sales productivity but not overall company performance. Consistent with Banker et al. 1996, we also observe that the negative impact of the plan change on individual productivity was higher in the status quo period than in the adjustment period. Furthermore, our findings suggest that while lower commission rates reduce the incentive for employees to work hard after meeting the minimum requirement, the hurdle rate motivates some employees to exert more effort to avoid losing their jobs. In addition, we found that lower-performing employees were attracted to the company while fewer high performers were retained. This study also extends prior research by investigating the employee groups most affected by the change to a less performance-sensitive plan. As predicted, we find that higher-performing employees were affected more than lower-performing employees. These findings suggest that managers should anticipate the impact of alternative compensation plans on the efforts of different employee groups (e.g.. ones with different levels of ability). As such, top management may need to refine the compensation contracts or take various actions to mitigate potential dysfunctional impacts. Furthermore, our results support the recruiting effect; that is, the average sales productivity was higher for those hired under the old plan than for those hired under the new pian. Our results also support the turnover effect. We find that employees who were hired under the old plan and quit after the plan change had higher average sales productivity than those who left before the plan change. However, we find no significant difference in sales productivity between those who were hired and left under the less performance-sensitive plan and those who were hired and left before the change. Results from additional analyses show that turnover occurs either when employees experience higher compensation loss or have less experience and poor performance, causing them to fail to meet the minimum requirement of productivity. These results may help management anticipate that employees with certain characteristics will be more likely to leave if the compensation plan becomes less performance-sensitive and has a minimum requirement of productivity. Interestingly, while we find the change from a performance-sensitive plan to a less performance-sensitive plan hurts individual productivity, it does not have a negative effect on the company's overall perfonnance. Our interviews with managers and additional analyses reveal that in light of the decreased individual sales productivity, the field site hired more employees and provided incentives to employees

CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

193

to sell higher-margin, more competitive nonexclusive cars. However, why the increased sales mix of higher margin cars did not increase the company's overall revenue is an open question. Although we find an increase in the proportion of high-margin cars being sold, the total number of cars sold decreases significantly after the change to the new plan. As such, the increase in revenues due to a favorable sales mix change may have been offset by the decrease in the number of cars sold, resulting in no change to the company's overall revenue. Although our study has advanced the understanding of tbe effects of perfomiancebased incentive schemes, it is important to recognize some limitations. First, due to the lack of nonfinancial measures (e.g., customer surveys) at the field site, we cannot measure other goals for employees to achieve and therefore bave to limit our analyses to sales productivity and compensation. Second, like prior studies on changes of incentive schemes (e.g.. Banker et al. 1996; Lazear 20(X); Banker et al. 2001; Brickley and Zimmerman 2001), our analysis was restricted to a large data set from one organization. Therefore, our results may have low generalizability to otber organizations and contexts. Finally, although we have included both competitor performance and local competition intensity in our models, we do not have information about the competitors' strategy, which may have also affected the performance of our field site. To shed more light on the change to a less performance-sensitive plan, future studies could examine optimal contracts for employees with different levels of ability. Our results show that the change to a less performance-sensitive compensation scheme has a greater negative impact on tbe productivity of high-ability employees and that tbe threat of dismissal does not improve the productivity of the poorest performers. Future studies could examine tbe best incentive schemes when companies are under different competitive circum.stances. For example, what are the benefits and costs associated with different types of contracts (e.g., higher fixed salary associated with a higher hurdle; same fixed salary with higher commission rates)? Finally, worker heterogeneity (i.e., workers with different skill levels) commonly exists, and companies may use team-based compensation plans to motivate employees to share their knowledge (Hamilton, Nickerson, and Owan 2003). As such, future studies could investigate bow team-based compensation plans can reap the benefits of worker heterogeneity by motivating higher-ability workers to teach lower-ability workers how to deal with customers more effectively, thereby improving their performance.

. 26 No. I (Spring 2009)

194

Contemporary Accounting Research

APPENDIX Variable definitions Variables ABIUTY AGE CARJSOLD CTY_SALE Definition = employee's ability = average age of employee = number of cars sold per month = log of monthly market sales of local competitors. The hranch and its local competitors are located in the same city = the dollar commission per car = compensation per person-month = (the average compensation after the new plan - the average compensation before the new pian)/the average compensation before the new plan = an employee who joined the dealership after the new plan = an employee who joined the dealership after the new plan and who left the dealership after July 1, 1998 and before the end of otir sample period = an employee who had joined the dealership before the change to the new plan (July 1. 1998)
= an employee who joined the dealership before the change =

COMMISSIONJiATE COMPENSATION COMP_L0SS

zyVEW (NewPlan) D^Q^ (NewPlanQuitAfter)

DOLD (OldPlan)

DOQA {OldFlanQuitAfter)

to the new plan and who left the dealership after July 1, 1998 and before the end of our sample period i
DOQB (OldPlanQuitBefore) ^- an employee who had joined the dealership before the

change to the new plan and left before July I. 1998 EXPER FAM COMP FIXED_PAY GS_PROFIT HIGHJiBIUTY INCOME IND_PROD LEFT = number of years given employee had been employed = log of monthly market sales of family competitors that sell the same brand of cars = fixed salary per person-month = gross profit per month = a dummy set equal to 1 if employees' ability is in tbe first quartile = income before allocating the operating expenses of headquarters and income tax per month = cars sold per person-month i = a dummy set equal to 1 if the employee left the dealership, otherwise 0 (The appendix is continued on the next page.)

CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention


APPENDIX (Continued)
Variables LOW_ABIUTY

195

Definition - a dummy set equal to 1 if employees' ability is in the fourth quartile = market share in Taiwan car market = a dummy set equal to 1 if employees' ability is in the second and third quartiles ^ turnover rate per month = log of monthly market sales of nonfamily competitors that sell different car brands but compete for the same customer groups = company-level performance = a dummy set equal to 1 if the employee is on the new plan during given month = revenue per month = sales mix of sedans versus trucks = log of monthly head count of employees in a branch, not including the administrative staff = log of monthly head count of employees in the company

MARKET_SHARE MEDWMJiBlLlTY
MTH_TURNOVER NON_FAM_COMP

PERFORM PLAN REVENUE


SALES_MIX

m_EMPLOY
#F EMPLOY

Endnotes
1. We use the term "less performance-sensitive plan" because the commission under the new plan is lower for each car sold, which is generally the case for employees whose productivity is far greater than the required minimum number of cars sold each month and who have a low risk of falling below the minimum threshold. 2. Quantities of sale are 361,615, 374,955, 394,160, 357,634. and 350,461 from 1996 to 2000, respectively (source: Taiwanese Directorate General of Highways). 3. For presentation purposes, throughout the paper we have converted Taiwanese dollars to U.S. dollars with an exchange rate of 34 to 1. 4. According to the CEO, family competitors make up the major competition. In Taiwan, the top five car brands (i.e., Toyota, Mitsubishi, Nissan, Ford/Mazda, and Honda) have more than 80 percent of the market share, and each car manufacturer has at least two dealerships. 5. Our interviews with managers and employees suggest that employees leave the company for one of three likely employment opportunities. The first and best alternative is to work for family competitors. Because they sell the same brand of car, there is no need to learn about new products, and employees may be able to keep customers who have a loyalty to the car brand. The second-best alternative is to join nonfamily competitors. In this case, employees can still apply their sales skills and industry-specific knowledge to their new jobs. However, they have to invest time

C-4 Vol. 26 No. 1 (Spring 2009)

196

Contemporary Accounting Research


learning about new cars and recruiting new customers and also incur an opportunity cost of lo.sing existing customers who are loyal to the car brand of the field site. The third employment alternative is to work for noncar industries such as real estate and life/product insurance. Although employees can use their general sales skills, they need to invest a tremendous amount of time learning about new products and new industries.

6. At that time (before the implementation of the new compensation plan), fewer than three cars were sold in 45.2 percent of the employee-months (i.e., no cars were sold in 20.1 percent of the employee-months; one car in II .3 percent of employee-months; and two cars in 13.8 percent of employee-months). Because fewer than three cars were sold in almost half of the employee-months, the total compensation costs will increase dramatically if the company were to pay a base salary on top of the old (higher) commission rates. Therefore, to control total compensation costs, the CEO decided to lower the commission rates. 7. The field site was forced to include fixed pay by governmental regulation. As a result, this plan was implemented throughout the company rather than on a regional or phased-in basis, which can avoid potential selection bias of employees with specific risk preferences choosing to work for different regions or branches. In addition, as discussed earlier, some of the field site's competitors did not lower their commission rates, which also allows us to examine how the change in compensation plan affects the retention or turnover of employees with different ability levels. 8. As shown in Figure 1, there is a kink at the minimum requirement in the new contract (C). Specifically, employees receive a base salary before they reach the minimum requirement but are paid additional commission after the minimum requirement is met. 9. The company did not have a complete data set available until January 1996. To make the monthly data covered in the two subsamples equal, we only include data up to April 2001. 10. We exclude eight months (May to December) in 1998 from our sample for two reasons. First, we want to avoid a possible manipulation of the sales timing. There were two months between the announcement date of the new plan (May I ) and the effective date (July 1 ). During these two months, in light of the future decrease in commission rates, it is likely that some employees boosted sales to earn higher compensation. Second, we controlled for possible seasonal-month-sales effects. 11. The eigenvalues of these three factors are 1.77, 0.85, and 0.38. Following Kaiser's 1960 rule (i,e., eigenvalue greater than unity), we use factor 1 as the ability proxy. The factor loadings for the number of cars sold, employee annual performance rating, and reciprocal of time to the first promotion since he or she joined the dealership are 0.8719. 0.8416, and 0.5516. Including the number of cars sold prior to the month of interest may confound our measure of ability because the new plan results in lower productivity. Therefore, we ran a robustness check excluding the number of cars sold prior to the month of interest from our ability proxy. Because of the qualitatively similar results, we use three factors as the proxy for employee ability in the text and tables. 12. The unemployment rates are 2.69 percent, 2.93 percent, 2.99 percent, and 4.58 percent from 1998 to 2001. respectively (source: the Directorate General of Budget, Accounting and Statistics of ExectJtive Yuan in Taiwan). 13. Furthermore, as discussed in the hypothesis development section, we assume that the average compensation per ear under the old plan is greater than that under the new plan
I

CAR Vol. 26 No. I (Spring 2009)

Employee Performance, Recruitment, and Retention

197

that is. b> b' + alx for x > 3 cars. Recall that b {b') is the average commission rate under the old (new) plan, x is the number of cars sold, and a is fixed salary. As seen in Table 1, 6 is $495, b' is $225, and a is $466. Therefore, an employee who sells three cars receives an average compensation rate of $495 under the old plan, but reeeives a lower compensation of $380 ($225 + $466/3) under the new plan. Similarly, an employee who sells 10 cars receives an average compensation of $495 per car under the old plan, but receives only $271.6 ($225 -\- $466/10) under the new plan. As such, our assumption that all employees receive lower compensation rates under the new plan is supported. 14. According to the Directorate General of Budget, Accounting and Statistics of Executive Yuan in Taiwan, the average national income per capita in Taiwan from 1996 to 2001 is $12,292. The national income per capita from 1996 to 2001 is $12,418, $12,707, $11,522, $12,324, $13,090, and SI 1,692, respectively. The decline in the average annual income per capita at our field site during this period is 35 percent ($357*12/$12,292). 15. Because of repeated observations on the same set of cross-section units (Johnston and Dinardo 1997, 388), Hypothesis 1 can also be tested by examining whether a, is negative in panel data with the employee fixed effects model. We observe qualitatively similar results by using panel data with tixed effects. 16. We cannot directly compare the coeffieients for PLAN betvi-een the adjustment and the status quo periods because they are estimated in two different regressions. Therefore, we include the two proxies (PLAN_Adjustment and PLAN_ StatusQuo) in one regression and then use an F-test to compare the two coefficients. Our results show that the coefficient on the status quo period (-0.1237,;? < 0.0001) significantly (p = 0.0036) differs from that on the initial adjustment period (-0.0736./) < 0.0001). 17. To test the robustness of (2), we also use dummy variables for ability (i.e., classifying employees into HIGHjKBlUTY, MEDIUM_ABILITY, or LOW_ARIUTY group). Compared with the LOW_ABlLny group, the plan change had a significantly higher negative impact on the MEDlUM_ABILnY group (the coefficient of MEDWM_ABIUTY X PLAN is -0.543;/; < 0.0001) and on the HIGH_ABILITY group (the coefficieni of HIGH_ABILITY X PL\N is -0.736;/) < 0.0001). 18. The fixed-effects model using panel data is not feasible for testing adverse selection effects because there was a high employee turnover rate, with some employees leaving the dealership and new entrants joining the tirm at the same time. 19. We also include the coefficient of variation of sales to control for a possible effect of unstable sales over time on employee turnover. We find similar results tbat y^ is 0.486 (significantly greater than Oat the 0.0001 level) and 72 is 0.012 (not significant at conventional levels).

References
Abowd, J. 1990. Does performance-based managerial compensation affect corporate perfonnance? Industrial and Luibor Relations Review 43 (3): 52S-73S. Ashton, R. H. 1990. Pressure and performance in accounting decision settings: Paradoxical effects of incentives, feedback, and justification. Journal of Accounting Research 28: 148-80. CAff Vol. 26 No. 1 (Spring 2009)

198

Contemporary Accounting Research

Bailey C , L. Brown, atid A. Coceo. 1998. The effects of monetary incentives on worker learning and performance in an assembly task. Journal of Management Accounting Research 10: 119-31. Baiman, S. 1990. Agency theory in managerial accounting: A second look. Accounting, Organizations and Society 15 (4): 341-71. Baker, G. P., M. C. Jensen, and K. J. Murphy. 1988. Compensation and incentives: Practice vs. theory. The Journal of Finance 43 (3): 593-616. Banker, R. D., S. Lee, and G. Potter. 1996. A field study of the impact of a performancebased incentive plan. Journal of Accounting and Economics 21 (2): 195-226. Banker, R. D., S. Lee, G. Potter, and D. Srinivasan. 2001. An empirical analysis of continuing improvements following the implementation of a performance-based compensation plan. Journal of Accounting and Economics 30 (3): 315-50. Baron, J. N., and D. M. Kreps. 1999. Strategic human resources Frameworks for general managers. New York: John Wiley & Sons. Brickley, J. A., and J. L. Zimmerman. 2001. Changing incentives in a multitask environment: Evidence from a top-tier business school. Journal of Corporate Finance 1 (4): 367-96. Cameron. A. C . and P. K. Trivedi. 1998. Regression analysis of count data. New York: Cambridge University Press. Demski, J., and G. Feltham. 1978. Economic incentives in budgetary control systems. The Accounting Review 53 (2): 336-59. Driscoll, P. A. 1994. Sears to link incentives for auto service sales to customer satisfaction. Marketing News 28 (8): 8. Feltham, G., and J. Xie. 1994. Performance measure congruity and diversity in multi-task principal/agent relations. The Accounting Review 69 (3): 429-53. Freeman, R. B., and M. M. Kleiner. 1998. The last American shoe manufacturers: Changing the method of pay to survive foreign competition. Working paper. National Bureau of Economic Research. Gibbs, M. 1995. Incentive compensation in a corporate hierarchy. Journal of Accounting and Economics 19 (2-3): 247-77. Greene, W. H. 2000. Econometric analysis. Upper Saddle River, NJ: Prentice Hall. Hair, J., R. Anderson, R. Tatham, and W. Black. 1998. Multivariate data analysis, 6th ed. Upper Saddle River, NJ: Prentice Hall. Hamilton, B. H., J. A. Nickerson. and H. Owan. 2003. Team incentives and worker heterogeneity: An empirical analysis of the impact of teams on productivity and participation. Journal of Political Economy 111 (3): 465-97. Hoilenbeck, J. R., and C R. Williams. 1986. Tumover functionality versus tumover frequency: A note on work attitudes and organizational effectiveness. Journal of Applied Psychology 71 (4): 606-11. Holmstrom, B, 1979. Moral hazard and observability. W/yourna/o/Economici 10 (t>: 74-91. Jensen, M. C. 2003. Paying people to lie: The truth about the budgeting process. European Financial Management 9 (3): 379-406. Johnston, J., and J. Dinardo. 1997. Econometric methods. Singapore: McGraw-Hill.

CAR Vol. 26 No. 1 (Spring 2009)

Employee Performance, Recruitment, and Retention

199

Kaiser. H. F. 1960. The application of electronic computers to factor analysis. Educational and Psychological Measure 20 (1): 141-51. Lambert, R. A. 2001. Contracting theory and accounting. Joumal of Accounting and Economics 32 (1-3): 3-87. Lazear, E. P. 2000. Performance and productivity. American Economic Review 90 (5): 1346-61. Long, J. S, 1997. Regression models for categorical and limited dependent variables. Advanced quantitative techniques in the social sciences. Thousand Oaks, CA: Sage. Maddala, G. S. 1992. Introductions to econometrics. Englewood Cliffs, NJ: Prentice Hall Intemational. Merchant. K. A.. W. A. Van der Stede. and L. Zheng. 2003. Disciplinary constraints on the advancement of knowledge: The case of organizational incentive systems. Accounting. Organizations and Society 28 (2-3): 251-86. Milgrom, R, and J. Roberts. 1992. Economics, organization and management. Englewood Cliffs, NJ: Prentice Hall. Rrendergast, C. 1999. The provision of incentives in firms. Joumal of Economic Literature 37(1): 7-63. Salop, J., and S. Salop. 1976. Self-selection and turnover in tbe labor market. Quarterly Journal of Economics 90 (4): 619-27. Stiglitz, J. 1975. Incentives, risk, and information: Notes toward a theory of hierarchy. Bell Joumal of Economics 6 (2): 552-79. Tanikawa, M. 2001. Fujitsu decides to backtrack on performance-based pay. New York Times, late ed., East coast, Marcb 22, W.I. Waller, W. S., and C. W. Chow. 1985. The self-selection and effort effects of standard-based employment contracts: A framework and some empirical evidence. The Accounting Review 60 (3): 458-76. Williams. C. R., and L. P. Livingstone. 1994. Another look at the relationship between [jerformance and voluntary turnover. Academy of Management Journal 37 (2): 269-98. Young, S. M.. and B. Lewis. 1995. Experimental incentive-contracting research in management accounting. In Judgment and Decision-Making Research in Accounting and Auditing, eds. R. H. Ashton and A. H, Ashton, 55-75. New York: Cambridge University Press.

CAR Vol. 26 No. 1 (Spring 2009)

Вам также может понравиться