5. Human Science and Incentives

The topic of incentives and motivation is one in which the new understanding of human science has substantial significance. If we are not just profit-maximizing cyborgs, how exactly are we to be motivated? Evidence indicates that the way we have been compensating people can do more harm than good. Consider, for example, the financial crisis. All one had to do was look at the way in which compensation was structured at financial institutions to see that a financial crisis was eminent.² Incentive contracts for investment bankers provided an incentive to take extreme risks.³

I do not think it is an overstatement to say that one of the most critical issues facing organizations today (and many economies) is the establishment of a robust connection between rewards and performance. Take, for example, the issue of executive compensation. Much evidence supports a disconnection between executive rewards and performance.⁴ Executive compensation is often set as a function of the power of the executive rather than performance of the firm.⁵ The disconnection between rewards and performance extends well beyond those who occupy the executive suite. Research has shown substantial pay differences based on gender alone.⁶ And gender pay bias is only one form of wage discrimination; age and race are associated with biased decision practices, but not with job performance. Again, if we need to be reminded why this matters, it matters because it matters to human capital. Treating people with bias will cause disengagement (lower morale, lower productivity, higher turnover), ultimately resulting in suboptimal organizational performance.

In addition, the wages of those in the middle class have been mostly flat or declining over the past 30 years.⁷ This occurred during the same period that company profits were soaring. The flat real income growth applies to not only manufacturing and service workers, but also in varying degrees to professionals including teachers, airline pilots, nurses, and so on. Wage inequality undermines organizational effectiveness, and, more broadly, economic prosperity. This is no small thing. In the United States, consumer spending drives 70% of the economic activity. If we want growth, we need to more accurately allocate profits to those responsible for making them. This includes making a strong line-of-sight connection between rewards and performance, but it also means making informed decision about what incentive practice to use. For instance, it may mean that high-performing individuals and teams of store check-out clerks and hotel room cleaners should be getting stock options, but top executives should not.

There are a number of roles here for human science. The use of analytics can result in the establishment of a more robust connection between performance and rewards, not only for executives but for everyone in the organization. There is also the opportunity to include more pertinent information when developing incentive schemes. No two people’s situation or preferences are identical, and yet organizations reward entire job families in exactly the same manner. Organizations have relied too heavily on benchmarking data (and inaccurate assumptions about what motivates people), virtually ignoring individual motivational profiles. Treating everyone in precisely the same manner has almost certainly led to suboptimal performance outcomes. As detailed in Chapter 4, “Human Science and Selection Decisions,” a number of companies are applying human science to questions related to employee selection. However, applying these new developments to incentive contracts and performance management decisions is much less well developed.

I believe any discussion about selection decisions has to also include incentives, and vice versa. If you are attempting to choose people who will help make your organization thrive, that means you want someone who is motivated and engaged. You will want to keep them engaged, and the only way that will be accomplished is if you determine what motivates them and their team.

When I use the term “incentives,” I use it broadly, using a total compensation approach that includes both pecuniary (that is, base salary, bonus, and so on) and non-pecuniary (that is, organizational culture and brand) incentives. Earlier I discussed the privately held analytic software company SAS. Their pecuniary components of their incentives are a small part of a much larger “total compensation” approach. In many ways, this is a much more optimal way in which to structure incentives. SAS provides day care, in-house social workers, and a health club. These components of total compensation are more in-line with how humans, as opposed to econs, are motivated.

To access tools associated with this book, visit my site at DecisionAnalyticsInc.com.

As with selection decisions, compensation decisions are prone to biased decision making. Getting the incentives wrong can lead to human capital focusing on the wrong things and to employee morale and engagement problems, employee turnover, or, as some would argue, to the collapse of an organization or an entire economy.⁸

Compensation is clearly a key business decision and impacts the success of the organization. Executive compensation is a divisive topic, and getting incentive contracts right for executives (closely aligning pay and organizational performance) is critical for the well-being of the organization. However, establishing a pay-for-performance connection is important at all levels in the organization, not only the executive level.

A challenge in compensation that advanced analytics can address is overdependence on benchmarking to establish total compensation packages. One of the more beneficial aspects of advanced analytics is that it allows for the inclusion of a broader amount of information when designing incentive contracts. Not only can market data be included, but also information about what actually motivates the specific individual and team.

Representatives from one well-known organization sit down with employees when they are hired and ask them how they would like to be paid. Of course, at that point, discussions associated with base salary and any variable pay and benefits have already been discussed, but employees have room to “tailor” their compensation to their own preferences. This sort of individual tailoring of compensation may strike you as odd; it seemed strange to me when I first heard about it. After all, organizations usually dictate the various components of compensation. The individualization of incentives is exactly what advanced analytics is well suited for. One size does not fit all; what motivates me might not motive you, and our situations certainly differ. This again goes back to the issue of asymmetric or private information. The individual has information about preferences, but the organization also has asymmetric information about what they would like to see accomplished. Combining the two in a manner that maximizes both parties’ utility is the ideal.

One approach to compensation is to compensate people based on their marginal revenue product or their bottom-line impact on organizational performance.⁹ The problem with this is that it is difficult to determine individual (team or group) contribution to organizational performance.

Piece rates are viewed as one of the more efficient forms of incentives, and they do have a strong incentive effect. Piece rates compensate based on the quantity of goods produced. The challenge with this form of incentive is when you reward based on quantity, you often see a sacrifice in quality.

Generally, incentive schemes in organizations are based on a tournament model and focus on individual contribution. We come by the tournament model honestly because most organizations are hierarchical, with much of the decision-making authority residing with those at the top. Hierarchies with consolidated decision-making authority at the top of the organization are an efficient organizational design if those in the executive suite have perfect information and perform all the tasks of the organization.

Tournament-structured compensation can produce inefficient behaviors. This incentive contract structure can promote mistrust and animosity between employees. Tournament incentive structures may provide an incentive for employees of one department to withhold information that would have been helpful to those in another. The entire organization loses out when individuals and teams are rewarded for not sharing potentially value creating information.

Much of the research associated with incentive compensation has been has been done using research from sports. As a long-time fan of the Green Bay Packers (and being one of 363,948 “owners”),¹⁰ I am certainly interested in getting the incentives right in our sports teams. However, the problem has been that the findings for sports (specifically tournament compensation) have been extrapolated to compensation in organizations, and this can be problematic. As covered in Chapter 4, advanced analytics has been applied to selection decisions in sports, but has not yet been applied in the same way to incentives for athletes and team success. Research has found benefits associated with tournament pay (the winner-takes-all philosophy) for individual athletic performance; however, there is not much research evaluating incentives and the success of the team as a whole. Yes, a Michael Jordan can make or break a team (the typical NBA team has a roster of 15 players, not the 500, 5,000, or 50,000 found in organizations), but not even Jordan was an island. Without Dennis Rodman pulling down all those rebounds, Jordan does not get the ball and the shot. As in organizations, sporting teams may benefit substantially from spending more time getting not only individual, but also group and team incentive compensation, right.

Contrast Armstrong’s behavior’s with the behaviors of one of the most decorated soldiers in World War II: Audie Murphy. He was awarded the Medal of Honor at age 19 and was also awarded military honors in both France and Belgium. His entire life, Murphy insisted that the awards should have been given not to him individually, but rather to his entire military unit.¹¹

If tournament compensation does not work, or has too many potential downsides, what kind of incentive schemes do work and when? If you see the value in collaboration and cooperation within your organization, including team and group incentives are necessary.

Back when I was graduating from college, there were those who thought IBM was the top place to go work. One of the reasons everyone wanted to work for IBM was because the pay was higher there than pretty much anyone else. It was thought this attracted the top people. This pay practice is referred to as efficiency wage.

The notion of efficiency wage states that if you pay above market rate, you attract the best talent. You will work extra hard if you have been “gifted” a higher salary.¹² Of course, all the gifting was almost not enough to save Big Blue, which nearly crashed and burned. Yes, they hired from the top schools those with the highest GPAs, but it might not have been the most optimal way in which to attract and incent people. The notion of efficiency wages, with its focus on financial rewards alone, may not motivate maximum value creation from human capital.

In addition to the larger macro-level issues, a number of significant micro-influences are equally problematic. One of the biggest challenges with compensation is the problem of internal and external equity. For example, merit pay has long had a credibility problem because of the perceived, and frequently real, subjectivity that is often a part of the review process. Done correctly, the use of advanced analytics can substantially reduce subjectivity associated with merit-based compensation decisions.

Clearly, from an organizational performance perspective, rewards make a significant impact on the success of an organization—starting at the executive level and moving through the ranks of the organization. Generally, the closer one can get to making a strong connection between pay and performance, the more likely he or she is to see the compensation approach as being equitable and reasonable.

In the introduction to this section, I mentioned a talk that Daniel Ariely gave to a group of software engineers in Seattle. In the research he presented to the software engineers that day, Ariely was interested in the relationship between financial rewards, meaning and effort. In order to test these notions, he and his collaborators developed a test in which participants would, for a financial reward, assemble LEGO Bionicles (I didn’t know what those were either; sort of like miniature robots). They received $3.00 for the first Bionicle, decreasing by 30 cents each time they built one—for example, $2.70 for the second one, $2.40 for the third, and so on. In the first round of this experiment (referred to as the “meaningful condition”), Ariely and his colleagues would watch the participant make the Bionicle, place it under the table, give them the materials to make another, and so on. The average participant constructed 11. In the second experiment (they referred to it as the “Sisyphus condition,” referring to the king who was punished by God to roll a rock up a hill for eternity only to watch it roll back down right before it got to the top), they took the Bionicle that the participant had just assembled and disassembled it right in front of them. In the Sisyphus condition, they only built 7 on average.

They then described the experiment to a new group of participants, without actually having them build Bionicles, and asked them to speculate on what they thought would be the outcome. The average participant speculated that 8 would be built in the meaningful condition and 7 during the Sisyphus condition. They got the direction correct, but not the magnitude.

They also factored in a participant’s love for LEGOs; that is, some people just love building things with LEGOs, so presumably the greater the love for LEGOs, the more they would build. This is exactly what they found in the meaningful condition: a very strong relationship between intrinsic satisfaction and the number of Bionicles built. In the Sisyphus condition, there was no correlation between the number of Bionicles built and intrinsic satisfaction. Apparently, Ariely concludes, any joy or satisfaction associated with building LEGOs was crushed when the Bionicles were taken apart right in front of them.

What this is telling us is that when we are designing incentive contracts, the financial element is only part of the story and may well not even be the most interesting part. This will of course vary by person and by industry and job and location; however, contributing to something meaningful and being recognized are powerful motivators.

When it comes to designing incentive contracts that include these motivational components, expert intuition can play a very important role. It is hard to imagine that the executive at the large software company in Seattle that delivered the bad news to the software engineers about their project would leave everyone hanging. Here you have 200 software engineers demotivated and underutilized. As Ariely points out, this had an impact on productivity and could have easily led to employee turnover. Their financial compensation was probably not impacted, but the significance of their work was completely undermined.

What does this aspect of human science have to do with advanced analytics? If you are attempting to accurately predict how people are going to respond to a new policy, practice, or some other action, it is important to understand an impacting effort and engagement is about more than just money.

The means through which managers ensure that employee’s activities and outputs are congruent with the organization’s goals.¹³

The focus of much of strategic HCM is the alignment of HCM practices and policies with business objectives. For policies and practices to have an impact on performance, they need to ultimately influence value-creating employee behaviors and activities.

First, you want to determine the right set of policies and practices in your specific situation. Unfortunately, performance management practices are often handed down because this is what was always done, or it is what the firm next door is doing. Keep in mind that the overall objective is employee engagement—getting everyone on the same page, excited, and motivated to carry out the objectives of the organization.

Performance management provides the following functions:

• Help determine which performance management and incentive system to put into place

• Develop a line-of-sight connection between individual and team performance and organizational objectives

• Evaluate the effectiveness of the policies and practices

• Skills, abilities, knowledge, and individual characteristics

• Organizational characteristics and business objectives

• External and internal conditions and constraints¹⁴

Getting performance management right matters; employee engagement and satisfaction is associated with superior business outcomes.¹⁵ However, numerous factors conspire against the accurate evaluation of performance. Of all the HCM-related activities, one that certainly has much downside potential for biased decisions is performance management. Some argue that evaluators are inconsistent and that there is a high degree of subjectivity in such evaluation.¹⁶

• Appraisal politics: This is more common and more problematic than most other forms of biases. In the case of appraisal politics, evaluators consciously distort ratings to achieve individual or other division or company goals.

• Similar to me bias: We generally view ourselves as being competent and effective; so if we find someone who is very much like us, we are more prone to positively evaluate and reward that person.

• Halo and horn effect: This bias occurs when a rater ascribes one positive quality to all qualities. The inverse is the horn effect, when a negative quality is ascribed across all other characteristics.

Few topics have the potential to be more contentious within an organization than decisions related to compensation rewards. It can be a process that is rife with many of the biases outlined in Chapter 1, “Challenges and Opportunities with Optimal Decision Making and How Advanced Analytics Can Help.” For instance, merit pay has long been considered problematic because it was thought to be prone to subjectivity bias. (That is, those who were liked by the boss were given higher bonuses than those who were not.) It is safe to say that there are many, many ways to get compensation wrong, and numerous factors often go into the compensation decision.

Both Gartner and Forrester Research¹⁷ evaluate a variety of different tools to assist with total compensation decisions. There are tools that can assist with the administration of compensation; however, the application of more advanced tools to these problems is less well developed.

I have already spent some time discussing the Balance Scorecard, and this is certainly a tool that can be used to better reward executives. There are a number of different automated scorecards available that go a long way toward establishing a line sight connection between performance and individual and team behaviors and activities.

Certainly in the case of executives, all of the data is readily available to link their decisions on capital expenditures, strategic initiatives, and execution to operational and financial outcomes. Furthermore, there are excellent performance management tools such as IBM CFO Performance Dashboard or SAS Strategy Management that can provide all the necessary data.

Strategy maps can provide a variety of metrics that represent operational outcomes. It is possible to see how these metrics co-vary relative to one another, and assuming the system maintains historical metrics, it is possible to do what we discussed in Chapter 3; we are able to establish cause and effect through the use of panel data.

In addition, dashboards and scorecards can provide tools to help identify relationships between metrics and performance, including human capital metrics such as labor costs, employee turnover, and employee morale. As I mentioned earlier, Forrester Research and Gartner Inc. provide an excellent review of all these and other products.

It is also possible to build mechanisms for factoring in expert judgment into the tools themselves, as in Dr. Virginia Apgar’s test to evaluate the health of a newborn. She came up with the important variables, and it is an expert that is assessing the newborn. The same kind of “weighting” used in the Apgar test can be incorporated into data-driven decisions related to selection and incentive decisions.

It is possible to obtain predictions on how teams, departments, and divisions will respond to prospective changes in incentives. Agent-based modeling provides an ideal mechanism for attributing for “non-rationale” responses by individual agents. Most models assume “rational” responses, which we know is not how humans actually act.

• Multiple Regression Techniques: Multiple regression techniques are the workhorse of much of analytics. They can be used to determine the impact on performance associated with the introduction of a new incentive scheme.

• Decision Trees: Essentially a graph or model depicting steps to a decision. This can be used to provide evidence-based recommendations on, for example, who should receive restricted stock and how much.

• Monte Claro Simulation: This consists of using computation algorithms to arrive at probability distributions or optimizations or calculate a solution to a differential equation. This technique could be used assess the probability of someone resigning.

• Neural Nets: Are used to model complex relationships between variables. Can be used to model probable outcomes in groups.

• Linear and Non-Linear Programming: Techniques using mathematical optimization to determine most efficient outcome. Used to determine most cost-effective payment package.

In the book Pay Without Performance, Lucian Bebchuk and Jesse Fried provide a convincing argument that the determination of executive compensation is largely a function of executive power.¹⁹ Though the corporate board sets the compensation level, the CEO may still have considerable influence (depending on how much power he or she has) over the compensation package. This by definition means that these executives are receiving a disproportionate amount of the profits, which means, quite frankly, they are literally taking someone else’s earnings. The closer we can link pay to actual performance, the more fair and accurate the dispersion of profits. The use of analytics in the decision-making process can assist in eliminating any of the influences not directly associated with actual performance.

By applying advanced analytics to executive compensation, we can accurately evaluate historical performance to determine the maximum effectiveness of the compensation package. One of the most promising techniques is the use of indexing compensation against peer companies to determine the level of impact associated with a particular payment package for executives.

A significant amount of research provides information on what works when. This research could be included in an expert system and provide evidence-based research on what incentive approach to use when and where. Having a “Watson-like” expert system would allow for making evidence-based recommendations on components of incentive and motivational components appropriate for the executive and the organization.

With the use of predictive modeling, based on insights from the new human science, we can go a long way toward predicting the desired outcomes. This would entail including organizational and individual characteristics, as well as organizational and individual objectives. There are numerous data sources available that would enable one to develop and experiment across a broad range of different incentive schemes at the executive level. The Edgar database is publically available data maintained by the Security and Exchange Commission that contains detailed compensation data on every publicly traded company in the United States. This free data source provides compensation data on the top five highest paid individuals in every publicly traded company. In addition to the data on executive compensation, there is also data associated with the financial performance of these firms. Here is all the data needed to conduct a very rigorous analysis of what works in executive compensation and what does not. The same data is commercially available through S&P Execucomp and through the executive compensation information provider Equilar.

One of the more informed means of compensating executives is through the use of indexation. This comprises evaluating the performance of the executives based on how well they do in comparison to an index comprised of peers or even the S&P 500. Though this is often considered relative to the top executives, this indexation should comprise not only the CEO compensation, but also the entire executive team.

In addition, most firms will also have detailed strategic and operational outcomes. All of these data sources can be organized in order to develop. The ERP systems will almost always include the data necessary in order to evaluate the performance of the top executives.

The new human science would suggest that this group may be under-rewarded financially for their contribution, and would be excellent candidates for team or company-wide incentives in order to provide incentives to share information on customer preferences.

Here too the focus on pecuniary rewards may be at least partially misdirected. Developing a motivation profile and a reward system that motivates continued superior performance is critical.

The World Health Organization ranked 191 countries on the quality of their healthcare, including costs. The United States spends the most per capita, and was 38^th overall (immediately above Slovenia and one below Costa Rica).²⁰ It may well be that a connection exists between the extremely high costs of healthcare and the way in which physicians are compensated.

An optimal incentive approach for physicians may be the approach taken by the Mayo Clinic and the Cincinnati Clinic. They are considered two of the best healthcare facilities in the United States (and the world), and both pay their physicians a salary, rather than a piece rate system based on the number of patients seen and procedures or tests ordered.