Week 8 - Relative performance evaluation Flashcards
Recall the “CONTROLLABILITY PRINCIPLE”.
Should we hold employees accountable for results that are only partially controllable?
Yes, otherwise nobody cares
eg. Nokia/Kodak - you want managers to pay attention/respond to changes in customer preferences and/or new technologies
But make adjustments to pay
Controllability principle & Asymmetric compensation - Should we hold employees accountable for results that are only partially controllable?
Bertrand and Mullainathan; 2001. Are CEOs rewarded for luck? The ones without principals are. Quarterly Journal of Economics
Garvey and Milbourn; 2006. Asymmetric benchmarking in compensation. Journal of Financial Economics
- Luck = uncontrollable factors; Skill = factors controllable by employees
- Bertrand and Mullainathan’s research finds that CEO pay is influenced by both UNCONTROLLABLE + controllable factors - If CEOs are influencing their own compensation, they would INCREASE compensation when performance is BAD, but NEVER reduce compensation when performance is good
-> this goes to show that if do things subjectively, prone to manipulation and bias
In reality, controllability is a “continuum”. The real question is how much risk/uncertainty to share with employees.
5 factors affecting the willingness of employees to ACCEPT a RISKY contract
- RISK AVERSE employees would demand a risk premium
- Can that risk be diversified and/or HEDGED?
» employees cannot hedge risk related to their own company - Co-movement between risk and overall WEALTH
» if employee is very wealthy, risk may be perceived as lower - (OVER)CONFIDENCE
» eg. CEOs overestimated performance and think that they can get a higher compensation - Incentive/performance evaluation system
3 MC ways to control for “uncontrollable” factors, ie. ways to improve risk-sharing
» not perfect, still requires assumptions and forecasting
- Flexible performance standards
eg. ex-post budgeting, scenario plans - Relative PE
- Subjective PE
» can cause many problems b/c overruling decisions made in the past
3 main circumstances to use Relative performance evaluation
eg. comparing American Airlines against other American aviation companies
Performance = f(systematic, idiosyncratic)
^2 components
- Use similar peer firms, employees in teams in same org, similar branches etc. that have a COMMON underlying uncertainty!!!
- common uncertainty puts risk on employees that is 1) undesired and 2) they CANNOT deal with themselves
- then outcomes are informative about each other and the common uncertainty - STRATEGIC SUBSTITUTES
ie. not strategic complements. Firms should not induce unwanted competition - Limited ability for SABOTAGE and/or COLLUSION (main side effects of using RPE)
- sabotage = actions purely targeted at decreasing competitors’ performance
- collusion -> unhealthy competition between firms, eg. some firms collude on pricing so that other competitors cannot beat them
Another downside: employees may not want to share info with other employees due to competition; don’t want to be worse than others
Upside & downside of RPE and targets
+ don’t need to set targets anymore b/c competitors become your “targets”
- no more targets creates some uncertainty
3 forms of RPE - FORCED RANKING SYSTEMS (vs free systems)
(One form is Tournaments eg. sports, politics, promotions, employee of the month, etc.)
- FORCE evaluator to use a PREDEFINED DISTRIBUTION
eg. max 20% high performers, certain number of 4 star employees - by definition this has SUBJECTIVITY
» leads to managers tending to give everyone high ratings
» and/or compressing ratings (= give similar ratings to not hurt employees’ feelings)
-> so these ratings become useless since everyone know they’re BIASED
Performance appraisals and the impact of forced distribution - an experimental investigation (Berger, Harbring and Sliwka, 2013)
- What is the impact of forced ratings on productivity?
- When switching, does the “old” system matter?
- What if employees can harm others’ performance?
- Forced distribution system can IMPROVE PRODUCTIVITY
- employees tend to put more effort b/c comparing with each other
- there is a learning effect after playing multiple rounds of the game - However, if implement once employees are used to not having this system, ie. FREE RANKING first -> effort effects are reduced
- having experience with “forced” first leads to CONTINUOUS effort effects - If allow SABOTAGE opportunities -> all benefits of forced ranking system disappear, huge negative effect
Cardinaels and Feichter; 2021. Forced rating systems from employee and supervisor perspectives. Journal of Accounting Research
Forced ranking systems can…
1. INCREASE PRODUCTIVITY on an “effort task”
- but detrimental when sabotage is possible
- but can lead to STRESS and REDUCE “CREATIVE” performance
eg. long-term innovation
3 forms of RPE - CLUSTERING (or peer groups)
ie. explicit evaluations on others’ performance - within or between-firm
Within firms
- RPI vs RPE
» Relative performance Information won’t SHIELD competitors from uncontrollable factors - Think about SOCIAL PREFERENCES/EXTERNALITIES, that explain why RPE may not work
ie. employees’ actions have a -ve externality on their colleagues’ compensation and they don’t want to hurt them. b/c their good performance means others are performing relatively poorly
3 forms of RPE - CLUSTERING (or peer groups)
ie. explicit evaluations on others’ performance - within or between-firm
3 Consequences of Between-firm RPE
eg. comparing relative Total shareholder returns, relative pre-tax income margin improvement
- Greater idiosyncratic risk-taking
ie. innovation; but depends on payout structure, peer group difficulty and incentives - Some gamesmanship
eg. bias in peer selection, disclosure tactics - COMPETITIVE AGGRESSIVENESS
- firms don’t necessarily need to pick each other for RPE (only about 20% do)
- but when they do, in this overlapping situation SABOTAGE becomes more prevalent
- and only in this situation, firms become more aggressive towards each other
*unlike within-firm RPE, between-firm RPE is not always a tournament; depends on other firms’ choices -> prisoner’s dilemma
(but when it is, side effects are stronger)
Fine Harvest Restaurant Group case
RPE and Clustering within firm - What are the new tailored targets set?
Instead of the old targets (restaurant margin), the new targets are Restaurant profit “OPPORTUNITY’…
… so the “average cluster performance’ becomes the (relative) target
- To implement this, you need
1. performance dimensions
2. clusters
Fine Harvest Restaurant Group case
How do we evaluate whether the Performance measures are good?
Use REGRESSION analyses.
1. Check if p-value < 0.05 to know if a variable is statistically significant or not.
2. A higher adjusted R^2 tells us that more of the VARIATION in the outcome variables is explained by the measures
Fine Harvest Restaurant Group case
How do we evaluate whether the Clusters are good?
*Generally filter on firm/branch size, types of clients, types of products to explain Variation in outcomes/profit measures
Examine Descriptive statistics. Ideally,
1. the differences in mean BETWEEN clusters are relatively large
ie. diff. groups are really diff. from each other (heterogeneous between clusters)
2. the std dev WITHIN a cluster is relatively low
ie. so that everyone within a given group is quite similar (homogeneous within clusters)
Doing REGRESSION analysis also shows whether adjusted R^2 is higher with clustering,
eg. clusters explain 2-3 times more variance in new measures than old measure
Fine Harvest Restaurant Group case
Comparison of old vs new system which uses Clusters - 3 strengths and 3 weaknesses
Strengths
1. increased controllability
2. MORE GUIDANCE for poor performers {with the average cluster performance}
3. AVERAGE performance will likely move up, ie. overall performance of firm
Weaknesses
1. more complex
2. are the best measures chosen?
3. too many clusters?
*also want to use a PIVOT TABLE to check whether branches are still classified the same (Excellent, Good, Below Average, Poor) under the new performance system
» in the case, many branches SWITCHED groups after implementing new system, thus new system identifies branches VERY differently
Possibilities of implementation process
1. Start with just the 3 new performance measures (add clustering later)
2. Start with less clusters
3. Roll it out over time, eg. use a 50/50 weight on old vs new system for bonus for 1-2 years
4. Base some pay on corporate performance, to reduce potential sabotage concerns
