The University of Massachusetts Amherst Annual Fund raises money for the university by hiring students to cold call alumni, parents of students, and other friends of the school. Potential donors are separated into different lists based on individual profiles (i.e., people who have never donated, people with high incomes, and people who donated the previous year). Each student worker is randomly assigned to call different lists of people during a work session with the goal of receiving as many donations as possible. Because donations fund the cost of faculty, campus upkeep, scholarships, and other endeavors beneficial to the university,¹ the Annual Fund is a vital university organization.

The Annual Fund measures success by the amount of pledge dollars received, so understanding the pay schemes that most motivate student workers is crucial. Before the fall of 2011, the Annual Fund used a piece-rate compensation scheme. A piecerate compensation scheme is a pay schedule in which a worker is paid a base salary plus a “piece” for each unit produced or action performed. For instance, in 2011 each employee at the Annual Fund earned a base pay of $8 an hour and received additional bonuses for receiving donations on credit cards, receiving large pledges, or meeting nightly goals. The Annual Fund’s 2011 compensation scheme is reminiscent of the piecerate scheme Edward Lazear (1996) studied in the autoglass industry. Lazear found that a piece-rate scheme not only maximizes a worker’s output (the effort effect), but also attracts more productive workers to the firm (the sorting effect). Lazear’s piece-rate findings are reaffirmed by Shi’s research on the lumber industry (2010) as well as by Heywood, Wei, and Ye’s research on university professors (2011). These studies seem to confirm the significant effects of piece-rate schemes on worker productivity. However, the literature regarding piece-rate compensation is devoid of a calling center example, such as the Annual Fund.

Starting in the Spring of 2011 the Annual Fund changed their pay scheme from a piece-rate scheme to an hourly wage structure. Under the new compensation scheme, workers earned a base pay of $9.25 an hour—an increase of $1.25 an hour. However, all piece-rates and bonuses were removed. According to Sarah Sligo, the Executive Director of Annual Giving,² the compensation scheme switch aimed to create a sorting effect. In theory, offering a base pay of $9.25 an hour - the highest student wage on campus - prompts high ability workers to apply to the Annual Fund, while simultaneously discouraging unproductive workers from applying. Assuming the sorting effect holds true, it would subsequently create an effort effect in which pledge dollars are maximized because only high ability workers are present at the firm.

There is some research done on the effects of changing from a piece-rate to an hourly compensation scheme; however, the literature does not support Sligo’s assertions about creating sorting and effort effects. For instance, Freeman and Kleiner (1998) found that production declined when an unnamed shoe manufacturing company switched from piece-rate to hourly pay. However, the company saved money because both the reduction in wage costs and reduced workers’ compensation claims outweighed the costs of reduced production. Although production declined as a result of the switch to hourly pay, Freeman and Kleiner (1998) factored in additional costs of piece-rate schemes—such as supervisor costs, quality control costs, and hiring costs—in order to analyze the benefits of switching to an hourly wage scheme.

I hypothesize that a change from a piece-rate compensation scheme to an hourly compensation scheme negatively affects the output of workers. I examined twenty-two call lists, each consisting of the aggregate amount of pledge dollars, call attempts, and called party responses for Spring 2011 and Spring 2012. Also included are the month-by-month totals for both springs. The Fall of 2011 is not included because the fall and spring calling periods vastly differ due to the compositions of call lists. In order to protect the confidentiality of the student callers, personal information such as major, gender, and year of graduation were not disclosed; however, the data includes the number of student callers who worked in both springs. The information provided on the sheets allows for a comparison of the effort effect, sorting effect, and tenure effect between the two springs. Through an extensive literature review, I will examine research done on compensation schemes to support my hypothesis that a change in compensation scheme from a piece-rate scheme to an hourly compensation scheme negatively affects the output of workers. Next, I will look for differences in output across all twenty-two lists between the Spring of 2011 and the Spring of 2012 for an effort effect. Because the turnover for student callers is so high between semesters, the presence of an effort effect implies a sorting effect. Essentially, if worker effort in Spring 2012 is different from worker effort in Spring 2011, it can be inferred that different ability workers were hired. Finally, I will examine the month-by-month data to see if there is a tenure effect (i.e. if workers’ ability increased by learning the nuances of the job) as a result of the compensation scheme change.

¹ http://www.umass.edu/giving/opportunities/

² http://www.umass.edu/giving/contact/directory/

Edward Lazear’s 1996 research on Safelite Autoglass Company’s compensation scheme change remains the predominant study assessing the effects of such modifications. Safelite initially paid their workers an hourly wage; however, they gradually switched all of their workers to a piece-rate scheme. Lazear tested for significant differences in worker output before and after the compensation scheme switch. More importantly, Lazear computed the optimal piece-rate that maximizes worker output and subsequently the firm’s profit. He predicted three main effects: both productivity of workers and the variances between workers’ productivity would increase (the effort effect), low productivity workers would leave while high productivity workers would take their place (the sorting effect), and the longer a worker receives a piece-rate the more productive he or she will be (the tenure effect).

In order to predict worker output, Lazear created a model using a dummy, or indicator, variable for the type of compensation scheme used and subsequently controlled for how long the worker functioned under the new compensation scheme. Lazear found both the effort effect and the tenure effect to be positive and significant. Not only did the piece-rate scheme incentivize workers to increase productivity, but also variance in productivity increased because high ability workers improved their efforts. The sorting effect, which tested to see if low productivity workers were replaced with high productivity workers, was also proven. Thus, high productivity workers were attracted to the firm because their effort dictated their pay, and low productivity workers left because they could not compete with the higher achieving workers.

Lazear’s model also encompasses other controls —such as paid sick time—in order to measure the optimal piece-rate that dissuades workers from calling out sick. Because Safelite’s piece-rate compensation scheme incentivized workers to increase output, quality control measures became necessary. For example, if a Safelite employee failed to satisfy Safelite’s standards (i.e., the autoglass chipped soon after installation), the task would be randomly assigned to another worker at the firm for no piece-rate compensation. This created a peer pressure effect in which an employee’s fear of being ostracized from his or her work group deterred low quality workmanship. Lazear’s findings at Safelite paved the way for researchers to look for these cascading effects in other industries

A study by Lan Shi (2010) examined the effects of piecerate changes in the lumber industry. A group of workers at a tree-thinning company in Washington State were randomly assigned to receive piece-rate compensation instead of an hourly wage. The basis of the piece-rate scheme was to reward workers for removing “bad apples,” or apples of lesser quality, from trees (Shi, 2010). Following the compensation scheme change, the company had an incentive to ensure that all of the bad apples were removed. Thus, when the company switched to a piece-rate scheme, worker productivity statistically improved, as did the variance in the worker productivity. The variance in worker productivity is expected because the effort is greater for high ability workers compared to low ability workers. Similar to Lazear, Shi tested for the optimal piece-rate the business should pay workers in order to maximize productivity and profit.

Booth and Frank (1999) expand piece-rate literature by examining the wide scale effects of a compensation scheme change. Aside from supporting previous findings that piece-rates cause an increase in overall production (Lazear, 1996; Shi 2010), Booth and Frank found that companies and industries using piece-rate schemes attracted workers with higher “unobservable ability.” Higher unobservable ability means that these workers have a low cost of effort and will work more than a worker with a high cost of effort. Interestingly, Booth and Frank also found differences in gender responses to piece-rates. There was a 9% earnings increase for males compared to 6% for females. This seems to suggest that men respond to piece-rate schemes slightly more than women.

Heywood, Wei, and Ye (2011) found that even professors respond positively to piece-rate compensation, in which piecerates are given for publishing research. For example, a highly ranked research university in China greatly increased their piecerate for publishing research and tracked, with the help of Hey-word et al., the subsequent publishing rate. Heywood et al. used a similar research university in China as a control group. Different academic journals were put into tiers according to prominence and assigned appropriate piece-rates. The results were both positive and extremely significant across each tier, with increases of 55% to 76% for each tier. Most of the increases were attributed to full-time professors and professors who were already research active. Besides documenting an effort effect, Heywood et al. also found evidence of a sorting effect. For instance, professors hired after the piece-rate change were found to be more “productive” overall than the current professors.

Not all institutions find that piece-rate compensation is optimal. Freeman and Kleiner (2005) found that changing from a piece-rate compensation scheme to an hourly compensation scheme benefited a United States shoe firm. Faced with foreign competition that employed cheap labor, the US shoe firm changed its labor costs to remain profitable. Switching to an hourly compensation scheme reduced production, but firm profits improved because the reduction in wage costs offset reduced production. Freeman and Kleiner attribute this to the effects foreign competition has on domestic manufacturing industries.

Another researcher who finds support for hourly compensation schemes over piece-rate is Clemens. He focuses on the determinants of a piece-rate pay decision and argues that when output is hard to measure or there are hard to define piece-rates, industries should use an hourly compensation scheme (Clemens, 2012). Thus a piece-rate scheme is only more likely if production is dictated by endogenous factors. By using data from the Fourth European Survey on Working Conditions, Clemens controlled for such factors as workplace team pace, customer pace, target pace, automatic pace, and boss pace. He found that automated pace is not related to piece-rate schemes; however, team pace, customer pace, and target pace are. This implies that companies whose workers have more control over their production are likely to have piece-rates because the determinants of work pace are endogenous. Hence, companies whose workers have less control over their production are likely to pay with an hourly compensation scheme.

This study adds to the preexisting literature by providing a unique example of an abrupt pay scheme change in a call center environment. It may interest non-profits or other organizations that rely heavily on donations by laying the framework for an optimal compensation scheme. Additionally, most of the compensation literature focuses on the effects of switching from an hourly wage to a piece-rate, whereas my study examines the opposite change—from a piece-rate scheme to an hourly compensation. Furthermore, this study will test if the effort effects are linear for calling centers—meaning that I will determine if going from a piece-rate to an hourly wage reduces output.

The change from a piece-rate compensation scheme to an hourly compensation scheme was enacted in the fall 2011, however the fall giving period is typically not comparable to the spring period. Thus, data from Spring 2011 and Spring 2012 are used. The data was received directly from the Annual Funds records. There are 23 call lists in total: 11 from Spring 2011 and 12 from Spring 2012. The additional group from Spring 2012 is omitted from the results because there is not a corresponding control group from Spring 2011.

The list categories were as follows:

• List 2 (Second Ask): Individuals asked to be called in spring for second donation.
• List AB (Assigned Prospects): High income individuals given to the Annual Fund to call.
• List D (Donors): Individuals who give $0-$250 every year for under 5 years in a row.
• List F (Donor First): Individuals who gave for the first time last year.
• List G (Long Lapsed): Individuals who haven’t shown support for at least 4-7 years.
• List H (Holding Poll): Individuals whom research has shown to have a high income.
• List L (Lapsed): Individuals who gave 2-3 years ago.
• List N (New): Individuals who have never shown support.
• List P (Donor Plus): Donors that give $250-$1000 annually.
• List R (Recent Grads): Individuals who graduated from UMass in the last 5 years.
• List S (Spring Donor): “Call only in spring” donors.

Table 5.1: Call attempts and pledges by list

The number of call attempts decreased for groups 2, D, F, G, H, L, and R; the amount of call attempts increased for groups AB, N, R, and S. The amount of pledge dollars decreased for groups 2, D, F, G, H, L, N, and P; the amount of pledge dollars increased for groups AB, R, and S. The total amount of call attempts increased by 68,333; however, the total amount of pledge dollars decreased by $118,390.

Table 5.2: 2011 and 12 pledge per attempt by pay scheme

∗ indicates a percentage decrease in pledge per attempt

For lists 2, AB, D, F, G, N, and S the amount of pledge dollars received per call attempt decreased. For lists H, L, P, and R the amount of pledge dollars received per call attempt increased. Overall, seven out of the eleven lists decreased in pledge dollar amount per attempt, which is consistent with how much pledge dollar amount decreased overall. Additionally, the overall amount of pledge dollars per call decreased by almost twenty-five cents.

Table 5.3: Percentage changes in pledges by compensation and industry

The aggregate work productivity declined after the pay scheme change; thus there is evidence of both an effort effect and a sorting effect. I found that the change in incentive scheme from a piece-rate scheme to an hourly compensation scheme adversely affected pledge dollars received. For a majority of the lists, pledges per call attempt decreased after the pay scheme, as shown in Table 2. Table 1 shows that even though the amount of call attempts increased by nearly 10%, the amount of pledge dollars decreased by 17%. As expected, the amount of production, or pledge dollars per call attempt, decreased by about 25%. The decrease in production from switching to an hourly wage scheme is consistent with my hypothesis that the effort effect is linear.

Because an effort effect was found, and only six employees out of hundreds worked both springs, a sorting effect could be inferred. The high wage with low incentive to increase worker productivity attracted lower ability student callers to the Annual Fund. As shown in Table 3, Lazear (1996) found a 36% production increase for the autoglass industry, Shi (2010) found a 20-23% production increase in the tree-thinning industry, and Heywood, Wei, & Ye (2011) found a 50% increase in research publication for professors. According to my hypothesis, the Annual Fund should have had a decrease in production similar to the amount of increase in production for other industries. This hypothesis is supported as the Annual Fund saw a 25% decrease in production.

There is no clear tenure effect from the data, meaning workers did not become more productive after each successive month. The Spring 2011 production was not linear; however, the Spring 2012 data looks more promising. After the initial month in January Spring 2012, effort decreased by 22% and 26% in the next two months respectively. Production then shot up in April but decreased in May back down towards the level seen before April. Depending on what happened in April to make production rise for that month, the prevalence of a tenure effect is inconclusive.

The factors involved in making a successful cold call are both endogenous and exogenous. For example, the student caller could be more talkative or friendly, which may incline the potential donor to give a pledge. However, the potential donor could be irritated that he or she was called at an inconvenient time or not have the money to donate, and thus not donate regardless of the student worker’s effort. According to Clemens (2012), if the factors are more endogenous there should be a piece-rate, whereas if the factors are more exogenous they should use an hourly wage scheme. Because the student workers responded so strongly to the incentive, it can thus be inferred that the determinants of a successful call (one that ends in a pledge) are mainly endogenous. The implications of this are profound. Other donation calling centers which use an hourly pay scheme because of the perceived exogeneity of factors leading to success in a cold call could drastically increase their workers’ production by switching to a piece-rate scheme.

The limitations in the data prevented more in-depth statistical analysis. With better data from the Annual Fund on the individual callers, a model could control for both the different call lists and individual worker ability:

Y = β₀ + β₁Gender + β₂SR + β₃JR + β₄SOP H + β₅FRO + β₆Major + β₇List2 + β₈ListAB + β₉ListD + β₁₀ListF + β₁₁ListG + β₁₂ListH + β₁₃ListL + β₁₄ListN + β₁₅ListP + β₁₆ListR + β₁₇ListS + β₁₈Scheme

(5.1)

where the dependent variable, Y, would be pledge dollars per attempt. The different call lists are controlled for by a dummy variable, with a 1 indicating that list is being called and a 0 indicating that the list is not being called. The scheme type is also controlled for using a dummy variable, with 1 indicating an hourly wage and 0 indicating a piece-rate. Additionally, both gender and year of graduation are indicated by dummy variables. The majors would be broken down and grouped by which college they belong to. Dummy variables similar to the ones set up for lists would be used to control for colleges.

If the descriptive statistics hold true, there would be a statistically significant change in production between compensation schemes. Specifically, when all else is controlled for, production should go down by around 25%. Because this model incorporates gender, differences in how males and females respond to piece-rates could be examined. Booth and Frank (1999) found that males responded to piece-rates at a 9% level compared to 6% for females. This model would be able to expand the literature on gender differences in compensation schemes.

If the data were available for how much money each employee made under the piece-rate, I could have calculated the optimal piece-rate to award for each pledge dollar. Because the Annual Fund does not have a minimum standard for pledge amount to remain employed, the methods used in Lazear (1996) do not directly apply; however, the intuition remains similar. The Annual Fund would want to create a piece-rate that maximizes profit as well as induces low ability workers to maximize their effort. High ability workers will produce at a high level of output regardless, and the lost output from not maximizing the incentive for the high ability workers is more than counteracted by the savings to the firm.

Production was drastically reduced, but the effects on the Annual Fund’s overall profits are uncertain. The research done by Freeman and Kleiner (2005) found that a shoe company saved money from switching to an hourly wage scheme from a piece-rate scheme despite the decrease in production. However, without knowing how much money the Annual Fund saved by switching to an hourly wage, no conclusions can be drawn on the profitability of this change.

The Annual Fund changed its compensation scheme from a piece-rate scheme to an hourly wage scheme. The rationale was that high ability workers would be attracted to a job with a high base wage. The subsequent influx of high ability workers would increase production, measured in pledge dollars per attempt. However, this was not the case. The data suggests that production decreased by nearly 25%, which means that low ability workers were attracted to the high base wage with no piece-rates, and thus productivity went down. No linear tenure effects were found, which means workers were not consistently more productive or less productive after being assimilated with the pay scheme.

The changes in output suggest that the Annual Fund workers respond best to a piece-rate compensation scheme. According to the research done by Clemens (2012), this implies that the factors of productivity are endogenous. Call donation centers or other nonprofit organizations that rely heavily on call donations could increase their profit from switching to a piece-rate. This research provides additional support that switching compensation schemes from a piece-rate to an hourly wage scheme decreases production and it paves the way for more research on incentivizing cold callers.

[1] Booth, Alison L., and Jeff Frank. 1997. “Performance-Related Pay.” EconLit. Accessed 3 Dec. 2012.

[2] Clemens, Adam. 2012. “Pace Of Work And Piece-Rates.” Economics Letters 115.3 : 477-479. EconLit. Accessed 12 Dec. 2012.

[3] Freeman, Richard B., and Morris M. Kleiner. 2005. “The Last American Shoe Manufacturers: Decreasing Productivity And Increasing Profits In The Shift From Piece-Rates To Continuous Flow Production.” Industrial Relations 44.2 : 307-330. EconLit. Accessed 13 Dec. 2012.

[4] Heywood, John S., Xiangdong Wei, and Guangliang Ye. 2011. “Piece-Rates For Professors.” Economics Letters 113.3 : 285-287. EconLit. Accessed 12 Dec. 2012.

[5] Lazear, Edward P. 1996. “Performance Pay And Productivity.” EconLit. Accessed 3 Dec. 2012.

[6] Shi, Lan. 2010. “Incentive Effect Of Piece-Rate Contracts: Evidence From Two Small Field Experiments.” B.E. Journal Of Economic Analysis And Policy: Topics In Economic Analysis And Policy 10.1 EconLit. Accessed 3 Dec. 2012.