Truth-telling between Salespeople and their Managers (The Search for a Non-Truth-Telling Equilibrium) December 4, 2007 Presentation for: MGT 703: Experimental Economics
_In_Class_Presentation.ppt 1 Yale School of Management Motivation Experimental setup Results from in-class experiment Critique and Advice – I need your help … Agenda
_In_Class_Presentation.ppt 2 Yale School of Management What is the context? Managers rely on salespeople for this information – What are sales going to be? – Who is buying? (What? How much?) The manager represents the company – How much should the company produce? – Strong incentives to not over-/under-produce We see successful sales organizations in reality – Salespeople cooperate with their managers Salespeople represent the “front lines” of most profit-making companies – Most interaction with customers – Privy to the most information “on the ground”
_In_Class_Presentation.ppt 3 Yale School of Management Where’s the problem? Managers do not have the same incentives – Growth of company is very important – Maximize efficiencies – Minimize wasted effort/production (losses) Sales force has strong incentive to sell more – Salespeople typically receive two-part (or more complex) compensation schemes -Big bonus incentives for matching/exceeding sales targets -Some bonus part of salary expectations Managers also have to set targets – Salespeople often have better information – Revealing information can compromise the salesperson’s bonus
_In_Class_Presentation.ppt 4 Yale School of Management Motivation for Experiment… Everyone has an incentive not to cooperate Salespeople would like to keep information to themselves – Make targets lower – Earn larger bonuses Managers would like to know more information – Make accurate targets – Run the company most efficiently -Sales force bonuses are also a “cost” But, in reality, companies survive Is there equilibrium behavior? What mechanism drives this? Can we study the problem in a lab setting?
_In_Class_Presentation.ppt 5 Yale School of Management Experimental Setup: The Base Case Step 1: Salespeople see a private signal, and generate a sales “estimate” which is provided to their manager Step 2: The Manager creates a sales target based on that signal – (As well as information observed in prior periods) Step 3: Payoffs are calculated and observed by both the Salesperson and the Manager
_In_Class_Presentation.ppt 6 Yale School of Management Experimental Setup: Advanced Cases… Used to make the problem more realistic Case 1: Salespeople allowed to (privately) invest extra effort to grow sales – Potential for increasing bonus Case 2: In addition to Salesperson’s effort, Manager now allowed to provide “side-payment” – Manager makes (binding) decision up-front – Salesperson receives incentive to tell truth -“Insurance” against lost bonus -Proxy for other benefits that a manager may bestow in reality Experiment proceeded in stages – Allow subjects time to understand incentives (learning)
_In_Class_Presentation.ppt 7 Yale School of Management Results: Did people get the idea? Normalized Shading, Adjustment and Effort Exerted (Normalizations relative to Signal, Estimate or relative to Max Allowable Investment - $1k, respectively). N=8 Base CaseWith Effort Effort & Side-payment % Shading (relative to Private Signal) % Adjustment (relative to Estimate) % Effort (relative to $1000 max)
_In_Class_Presentation.ppt 8 Yale School of Management Results: Were results tending to stability? Normalized Shading, Adjustment and Effort Exerted (Normalizations relative to Signal, Estimate or relative to Max Allowable Investment - $1k, respectively). N=8 Base CaseWith Effort Effort & Side-payment % Shading (relative to Private Signal) % Adjustment (relative to Estimate) Tit-for-tat? Stability? Outlier group No s-p in last period
_In_Class_Presentation.ppt 9 Yale School of Management Results: Were results tending to stability? Normalized Shading, Adjustment and Effort Exerted (Normalizations relative to Signal, Estimate or relative to Max Allowable Investment - $1k, respectively). N=8 (N=7 in Effort & Side-Payment Case) Base CaseWith Effort Effort & Side-payment % Shading (relative to Private Signal) % Adjustment (relative to Estimate) Outlier group removed
_In_Class_Presentation.ppt 10 Yale School of Management Results: No changes in manager accuracy… Normalized Shading, Adjustment and Effort Exerted N=8 (N=7 in Effort & Side-Payment Case). Min Accuracy 73%, Max Accuracy 96%. Base CaseWith Effort Effort & Side-payment Behavioral Metrics Manager Accuracy
_In_Class_Presentation.ppt 11 Yale School of Management Results: Within-group variations… Within-group Variation: Shading & Adjustment (Normalizations relative to Signal or Estimate, respectively). N=8 Behavioral Metrics Base CaseWith Effort Effort & Side-payment +142% % Shading (relative to Private Signal) % Adjustment (relative to Estimate)
_In_Class_Presentation.ppt 12 Yale School of Management Results: Within-group variations… Within-group Variation: Effort Exerted/Side-Payments N=8. Y indicates offer of side-payment in round 1 (of 2). No side-payments were offered in round 2. With Effort Effort & Side-payment $ Effort “Invested”/$1000 YYYYY
_In_Class_Presentation.ppt 13 Yale School of Management Results: Behavioral Correlation Shading & Adjustment Between Treatments N=8. Red = Base Case; Blue = Effort; Green = Side-Payments (Some data points not on chart). % Adjustment by Manager % Shading by Salesperson Treatment or learning effects?
_In_Class_Presentation.ppt 14 Yale School of Management Results: Review Considerable within-group variation However, some reason to believe that there is a “stable” outcome – Even in base case, coordination seems to evolve – Very low shading/adjustment in side-payment case -After excluding outliers Shading and adjustment are positively correlated – Means that experiment is being understood – Unclear as to whether learning or treatment effects are taking place Most significant tests are rejected (n too small) Not enough time for “stable” behaviors to evolve (t too small)
_In_Class_Presentation.ppt 15 Yale School of Management Critiques and Advice… Experimental setup – Noise in private signal may not be necessary – Insufficient inclusion of “growth” motive -Currently “normalized” out – is this an oversimplification? – Random walk DGP -Would an extra condition (Target t >Target t-1 ) be more appropriate? In analysis, I have not (yet) discussed payoffs to either group Other items? What am I missing in the experimental setup/analysis? What mechanism might help sustain ‘equilibrium’ behavior observed in reality?