Value Competition

Modern and Traditional Social Preferences—Can We Test Them Experimentally in a Modern Setting?

Reading Henrich, Heine, and Norenzayan, The weirdest people in the world?   Fascinating stuff…

What follows are some reflections on whether it’s possible to test some of the claims I made in an earlier post about the social preferences associated with modernity among a population of American university students…the new material is in bold…

1.  Now: Tied altruism (valuing to some degree the additional social value provided by work and/or leadership) and competitiveness (wanting to be ahead of the other player) lead to choice of "work" (or "lead") over "don't work" (or "follow").  Positive individual consequence: The altruistic-competitive type gets to lead instead of the egoist.   Positive social welfare consequence: Positive externalities from higher level of work effort. 

Then: Egoism.  Social welfare consequences: Lower level of work effort than with altruistic-competitive social preferences, fewer positive externalities.

How to test?   Could test to see whether work effort is higher in a game with positive externalities from leadership…scenario could be that experimenter donates some money to charity when players work to lead and more money when they coordinate on the optimal leadership-following combination.  

Also, could test to see whether altruism and competitiveness are in fact positively associated at the individual level.   

PD—pro-social altruism                     Schelling—pro-social competitiveness

(anti-social egoism and competitiveness)         (sub-optimal egoism and altruism)

 

2, 2    0,3                                                                   3,3     1,0           

3,0     1,1                                                                   0,1     2,2

Coventry—anti-social altruism            Stag Hunt--anti-social competitiveness

(pro-social egoism and competitiveness)          (sub-optimal egoism and altruism)

 

1,1    0,3                                                                   3,3      0,1

3,0    2,2                                                                   1,0      2,2

Rat Trust—anti-social competitiveness

(optimal egoism and altruism)

3, 3   1,2

2,1    0,0

Hypothesis: There is a type that is simultaneously ALT, COMP, and PRO-SOC.   Playing ALT in PD will be positively associated with playing COMP in Schelling, EGO-COMP in Coventry, and EGO-ALT in Rat trust. 

 

2.  Now: Egoism with respect to non-nuclear family.  Positive social welfare consequence: Work and leadership not inhibited by shame at surpassing (or equaling) a member of one's in-group.

Then: Shame with respect to surpassing a member of an in-group.  Social welfare consequences: Reduced level of work and leadership in leadership games. 

How to test: a) Coventry played with parents or older family members in reality or hypothetically.   Hypothesis: There will be a problem with suboptimal altruism.  b) Coventry played with other students hypothetically or for money.  Hypothesis: Compared to the real or hypothesized game with family members, there will be no problem with suboptimal altruism.  c) Coventry played with cousins or other extended family members in reality or hypothetically.   Hypothesis: There will be no major problem with suboptimal altruism.  d) Coventry played with siblings.   Not sure about this one…

3.   Now: Situational, merit-based leadership.  Shame at failing to maximize total returns to players and society by following the lead of the other when that is optimal.  Positive social welfare consequence: More social welfare optima are reached.

Then: Ascribed lead and follow roles based on factors such as age, gender, etc.--altruism, competitiveness, and shame associated with adherence to these roles.  Social welfare consequences: Coordination on lead-follow readily achieved in leadership games; poor incentives for work and enterprise.

How to test: Leadership games (BOTS, Leadership, etc.) with no optimum lead-follow combination versus leadership games with an optimum lead-follow combination.  Prediction is that people will be impelled by shame and will coordinate to a fair degree on the optimal combination regardless of gender or other cues that can serve as a focal point for one player leading and another following.  In a game with no optimal combination, prediction is that people will use a neutral focal point (such as whose name is listed first).  No clear prediction on whether contemporary American college students will use ethically charged categories such as gender and ethnicity as cues or not.

Ultimatum and Public Goods games with and without an experimenter stake

1)  In the standard Ultimatum game, the experimenter is better off economically if there is a rejected offer.  (For example, an accepted proposal of $1.50 out of a $10 pot results in payoffs of 8.5, 1.5, 0 for the offeror, offeree, and experimenter, while a rejected offer results in payoffs of 0, 0, 10.) 

A hypothesis: The fairly high level of rejection of low offers observed in affluent populations will be reduced in a revised Ultimatum game in which money not taken by the players is destroyed and that is made known to the players. 

A second hypothesis: Mean Ultimatum offers by affluent populations will go down in the revised game, and the tendency for offers to cluster around 40-50% will diminsh.

A conjecture: Reductions in mean offer and in rejected offers in the revised game will not be as marked in small-scale, non-affluent populations like the ones studied by Henrich, et al, including populations with generous offers and/or high rejections of low offers.

2)  In the standard Public Goods game, the experimenter is better off if there is lower trust by the first player.   (For example, if the first player can withdraw $5 or triple it by allowing it to remain in the pot, the payoffs are 5, 0, 10 for the first player, the other players, and the experimenter for a withdrawal; the payoffs are x, 15 - x, 0 for trust, with x being the amount the next player returns to the first player.)

A hypothesis: The mean trust level of around 50% in affluent populations will increase in a revised experiment in which money not earned by the players is destroyed. 

A conjecture: The increased level of trust in the revised Public Goods game will not be as marked in small-scale, non-affluent populations as it is in affluent populations.

The reasoning behind the hypotheses: Social preferences of players can and in some cases do apply to the experimenter as well as to the other player.   Arguably people are considerably more motivated to achieve a social surplus and to avoid waste than is apparent in a standard game format.  The revised format in which Ultimatum game rejections and non-cooperation in Public Goods games cause a loss of value through money being destroyed is a way to test that proposition.

The reasoning behind the conjectures: It is plausible though not necessarily the case that affluent subjects feel a higher level of altruistic social preferences for the experimenter than is the case for subjects in non-affluent, small-scale societies.  On the other side, one could speculate that a population of, say, hunter-gatherers in the Kalahari would be more concerned about taking the money of an anthropologist administering the game than, say, UCLA students would care about how much money a lab has to run other experiments.  If that is so, the revised game format would be expected to boost social welfare-enhancing outcomes more among the non-affluent subjects than the affluent ones.  Both stories are plausible, though my money is on the first one.

 

Social Preferences and Mixed Nash

With egoistic players, mixed Nash reasoning has a weak, often negative relationship to maximizing the collective utility of the two players.   If social preferences are put into the mix, can mixed Nash become an ally rather than a foe of utility maximization?  

To analyze MNE, what I want to try out here is a simpler model of social preferences and imperfect information rather than the fairly elaborate perfect information Bentham-Pascal model of social preferences that I usually use.   

The intuition: Both players face a choice between a calculable, standard MNE strategy that assumes the other player is an egoist and an indeterminate, incalculable, non-standard MNE strategy that makes a guess at the other player's social preferences.   Faced with that choice, both players prefer to assume egoism by the other, both because it is a focal point and because the best response by the other player to MNE based on that assumption is to play the optimal or HJV strategy.  In turn, with both players knowing that the best response to MNE assuming egoism by the other is HJV, both players play HJV. 

Why is it a best response for you to to play HJV when the other plays standard MNE and you know you have some degree of social preferences?  The assumption that is necessary to bring this home is that you act with some p according to a social preference function such as Bentham or Pascal that if known by the other would tilt the other player's MNE tilt toward a *higher* proportion of the suboptimal or non-HJV strategy being played.   The result of that for the player aware that he/she has social preferences fitting this property is that HJV becomes a best response to the other playing standard MNE.

To make the verbiage concrete, let's start with a standard Stag Hunt matrix with egoistic preferences assumed:

               Stag             Hare

Stag         3, 3              0, 1

Hare         1, 0             2, 2

MNE is 1/2 Hare in this case.

Now, assume that with some p unknown to Player 1, Player 2 has social preferences that make P1's MNE for Hare higher than .5.  Both Bentham and Pascal fit the bill.  First, Bentham (i.e., subjective payoffs equal sum of P1 and P2 payoffs):   

                  Stag              Hare

   Stag         6, 6               1,  1

   Hare         1, 1               4, 4

Here MNE is 5/7 Hare.

Now Pascal (sum of competitiveness and shame, as explained in other posts):

                   Stag                Hare

Stag             0, 0               -1, -4

Hare            -4, -1               0, 0

Here MNE is 4/5 Hare.

Now suppose all you know is that you and the other player have some unknown admixture of egoism with social preferences that have the property of making MNE for Hare higher than .5.

Now suppose you face a choice between MNE of .5 Hare (assuming egoism for the other) and some other MNE of an indeterminate higher amount that guesses at the nature and probability of the different types of social preference that may be operating in the other. 

The first claim is that assuming egoism on the other's part and therefore playing an egoistic MNE of .5 Hare is indicated under focal point logic, given that there is no other calculable MNE.

The second claim is that a .5 Hare MNE is also indicated on the basis that a best response to it by a Player 2 who knows that he/she is at least partly driven by social preferences with the properties of Bentham and Pascal is to play Stag.

The third claim is that both players will reason the same way as to the other's best response and will hence play Stag.  

Need to think about this some more...there is a bootstrapping issue here with the players' adopting MNE of .5 Hare and then abandoning it for pure Stag based on the best response being Stag, but at the moment it doesn't seem like this is a contradiction that knocks down the argument.

[Another issue to think about is whether the same conclusion is reached if instead of social preferences we assume impulsivity in the other and oneself under which with some known p one will simply play Stag or Hare rather than calculating.  I don't think so...following are some notes on that issue I wrote up earlier today that represent a blind alley but that I keep here for what they're worth...

In the B-P matrix as in the egoistic one, there are no dominant strategies.   MNE against a player with known B preferences is 5/8 hare; MNE against a player with known P preferences is 4/5 hare.   Averaging those two based on an equiprobability of B and P means playing 57/80 hare.   (That's a higher proportion of hare than the .5 that is MNE in the standard egoistic stag hunt.)    Assuming both players are purely calculating (AKA "Nash") at the time of deciding their strategies, both play 57/80 hare, leaving the other player with equal payoffs regardless of whether they pick all stag, all hare, or anything in between. 

At this point, B-P has not gotten us out of the box that MNE with egoism puts us in.  To try to do that, we can add another assumption.   Suppose both players have a certain known propensity to act on impulse in a contradictory way ("Whitman") rather than to calculate ("Nash").   Suppose this leads both to play Stag .25 of the time, Hare .25 of the time, and to calculate the remaining .50 of the time. 

MNE with known B-P preferences and all calculation and no impulse is 57/80 Hare.  With known impulsivity as described above, a player who is calculating knows that .5 of the time he will not be in a calculating state and has simply played S or H on impulse.  Knowing his/her own impulsive tendencies, the player plays 37/40 Hare in the calculating state in order to equalize the payoffs to Player 2 from any combination of S and H by Player 2.  

Now we've reached a blind alley--at least with known impulsivity, it doesn't look as though B-P social preferences are going to work to get us out of the link between MNE and suboptimal outcomes.]

Explaining inconsistency in cooperation games

In a short new paper on SSRN, McCubbin, Turner, and Weller contend that people are all over the map in regard to egoism, social preferences, and calculation.  The pattern is no pattern--subjects who play Nash by being selfish in one situation in a Trust game may be generous in a Dictator game or a Donor game, and the same inconsistency applies to subjects who are generous in the Trust game.

One way to explain this: Suppose that people are evolved so as to cooperate but also to resist cheaters.  An individual with social preferences will want to cooperate contingent upon the probable or possible cooperation of the other.   Given that, it makes sense for calculation itself to be probabilistic, with the player acting some or most of the time according to cooperative or egoistic impulse rather than calculation.   Acting according to impulse some of the time allows the players to rely on the other's character in a way that would not be possible if they were pure calculators.  Acting according to calculation some of the time allows the players not to be taken advantage of (or to be an unacceptably aggressive partner) in a way that they would be if they were always governed by impulses of one kind or another.    

 

A possible bias in game theory lab results

In a standard format lab game, the experimenters pay out more to players who achieve the highest joint value outcome than to players who do not.  So, for example, in a prisoner's dilemma game two players who both cooperate may be paid $2 each, while two players who both defect will be paid $1 each. 

Players are supposed to ignore the experimenter--"pay no attention to that man behind the curtain!"   And under an assumption that players are egoistically motivated, they will.  But is that assumption right?   It is perfectly plausible for a player to have social preferences that embrace the experimenter and the lab, not just the other player.   One might feel altruistic toward the lab--"if the other player and I win a lot, there's less money for other experiments"--or competitive--"the other player and I are a team together against the guys in the white coats."

The size of a bias and its nature can be tested.   Alongside a standard lab PD (or other game) in which the lab has a stake in the players not doing well, one can test responses to a modified PD in which the players are told that a given amount of money will either be given to them or destroyed.   My intuition is that cooperation levels will rise in a PD run that way compared to the standard way.  I'd also expect that highest joint value outcomes will more often be reached in modified assurance and leadership games in which the lab lacks an economic stake in not having the players reach the best results.

A priori, my sense is that the modified format is the way to go as a matter of proper experimental protocol.  At least when social preferences are a focus of interest, it is a mistake to use a procedure in which the lab itself has an economic interest in the outcome.  

Building Better Agents: Modern and Traditional Social Preferences

1.  Now: Tied altruism (valuing to some degree the additional social value provided by work and/or leadership) and competitiveness (wanting to be ahead of the other player) lead to choice of "work" (or "lead") over "don't work" (or "follow").  Positive individual consequence: The altruistic-competitive type gets to lead instead of the egoist.   Positive social welfare consequence: Positive externalities from higher level of work effort. 

Then: Egoism.  Social welfare consequences: Lower level of work effort than with altruistic-competitive social preferences, fewer positive externalities.

2.  Now: Egoism with respect to non-nuclear family.  Positive social welfare consequence: Work and leadership not inhibited by shame at surpassing a member of one's in-group.

Then: Shame with respect to surpassing a member of an in-group.  Social welfare consequences: Reduced level of work and leadership in leadership games. 

3.   Now: Situational, merit-based leadership.  Shame at failing to maximize total returns to players and society by following the lead of the other when that is optimal.  Positive social welfare consequence: More social welfare optima are reached.

Then: Ascribed lead and follow roles based on factors such as age, gender, etc.--altruism, competitiveness, and shame associated with adherence to these roles.  Social welfare consequences: Coordination on lead-follow readily achieved in leadership games; poor incentives for work and enterprise.

4.  Now: Forgiveness of those who fail to make optimal response of trusting in an assurance game.   Positive social welfare consequence: Egoist has confidence that forgiver will trust; egoist will then also trust.  Qualification: Forgiveness does not work well against a competitive and spiteful other who does not care about social welfare and who cares about beating the other player rather than his or her own payoff.

Then: Retribution against those who fail to trust in an assurance game when you have trusted; honor culture.  Social welfare consequences: Those punished who are in one's in-group will trust if signal is sufficiently clear.  Those punished who are in one's out-group will retaliate, resulting in a suboptimal Nash equilibrium for the assurance game. 

5.  Now: Generalized shame reduces the "temptation" payoff for, say, not leaving a tip in a restaurant one will never return to, in a one-shot PD.   Positive social welfare consequence: PD solved some or all of the time for anonymous, one-shot interactions.

Then: Shame is particularized and only applies to interactions within one's in-group.  Social welare consequences: PD solved for one's in-group but not otherwise.

6.  Now: Shirkers/defectors in PD games who are punished respond in some cases by improving their behavior.  Positive social welfare consequence: More PDs get solved.

Then: Shirkers/defectors in PD games who are punished respond in some cases by punishing those who punish them, especially when the punishers are from out-groups.  Social welfare consequences: Difficulty in enforcing norm of cooperation in repeated PDs.

7.  Egoism as governing principle in harmony games in which self-interest of players results in dominant strategies that are socially and individually optimal.   Positive social welfare consequence: Harmony games get solved.

Competitiveness as governing principle in harmony games with a member of an out-group.  Social welfare consequences: Problems in solving a harmony game with a member of an out-group.

8.  Now: Competitive and altruistic feelings based on political ideology that cut against as well as in favor of one's demographic in-groups.  Positive social welfare consequence: Better resolutions of political issues.

Then: Altruistic feelings toward members of in-group and competitive feelings toward out-groups.  Social welfare consequences: Suboptimal group-based politics.

9.  Now: Competitive and altruistic feelings based on compliance with rules and on adherence to cost-benefit analysis that cut against as well as in favor of one's in-groups.   Positive social welfare consequences: Better resolution of organizational and legal issues.

Then: Less well-developed altruistic and competitive feelings associated with cost-benefit analysis.  Social welfare consequences: More rule-oriented, less flexible, less optimal approaches to organizational and legal issues.

10.  Now: Altruistic feeling toward members of ideological in-group; competitive feeling toward members of ideological out-group.   Social welfare consequences:  More optimal solutions to games with members of ideological in-group when ideology is salient.   Conflict with members of out-group when ideology is salient and potential for suboptimal outcomes unless ideological adversaries are accountable to non-ideologues through elections or otherwise.  

Then: Weakness of ideological as opposed to demographically-based political feeling.   Social welfare consequences:  See above. 

11.  Now: Altruistic feeling toward nation and members of national in-group; competitive feeling toward other nations and nationals.   Social welfare consequences:  More optimal solutions to games with co-nationals.   Conflict with other nations and their nationals; potential for suboptimal outcomes through war and otherwise.  

Then: Weakness of national feeling.   Social welfare consequences:  See above. 

12.  Now: Altruistic feeling toward your organization and members of your organization; competitive feeling toward competing organizations and their members.   Social welfare consequences:  More optimal solutions to games with others in your organization.   Competitive feeling toward others may lead to better outcomes through  motivating higher work effort or worse outcomes; interaction here between the individual level  of social preferences and the group and societal level of institutions and whether they channel competitive and egoistic feeling related to organizational loyalties in a socially productive direction.    

Then: Weakness of organizational  feeling.   Social welfare consequences:  See above.   

13.  Now: Low or zero valuation of having higher hierarchical rank than another.   (Decline of hierarchical competitiveness)    Individual advantage for social egalitarian: Doesn’t have to pay a premium for hierarchy or impose it coercively; the social egalitarian thus earns more than the hierarch in a market.  Social welfare consequences:  Less coordination on leader-follower roles, fewer struggles over hierarchical rank (as opposed to leadership).

Then: Strength of hierarchical feeling.   Social welfare consequences: See above.

14.  Now: Low or zero concern about being characterized as dishonorable for violating social codes that do not enhance social welfare.   (Decline of non-utilitarian shame)    Individual advantage for social egalitarian: Doesn’t have to pay a premium to uphold costly social codes or to fight honor battles.    Social welfare consequences:  Optimal outcomes from abandonment of codes that do not enhance social welfare.  Negative outcomes from situations in which codes that in fact enhance social welfare are mistakenly abandoned.  

Then: High concern about being characterized as dishonorable for violating social codes that do not enhance social welfare.    Social welfare consequences: See above.

 

 

 

 

Social Preferences and the Unpleasant Math of Base Rates

Reflecting some more on the themes of the last post...Dolly Chugh's audit study of academics reminds me of my old days doing real estate tests to look for evidence of racial steering...a key assumption that audit tests like mine and Dolly's use is that to treat people from different groups with equal credentials unequally is discriminatory...

The point here is to introduce a toy model in which neutrality results in a substantial tilt in favor of members of a group with a higher avwerage level of skills, with the aim of suggesting that social preferences play an important role in achieving the social objective of treating people with the same credentials equally.

The toy model: Assume as in the previous post a stronger group S and a weaker group W.  Assume the employer needs a job candidate with an expected level of skill that is good or G.  The average member of S has an S average skill level, SA, while the average member of W has a below average skill level, BSA.  Among S members, 50% are SA, 20% are G, and 5% are very good or VG; 20% are BA and 5% are much below S average or MBSA.  Among W members, 1% are VG, 4% are G, 20% are SA, 50% are BSA, 20% are MBSA and 5% are very much below S average or VMBSA.

All candidates have credentials that allow the employ to evaluate them as VG, G, SA, BSA, MBSA, or VMBSA.  The credentials however are muddy signals.  There is a 20% chance that a candidate will appear to be one notch above his/her actual skill, a 5% chance of appearing two notches above, and a 1% chance of appearing three notches above.  The same numbers apply to the chance a candidate will appear one, two, or three notches below his/her true skill.

Now, suppose a candidate's credentials signal G.  What is the chance the candidate actually has a skill level of G or VG?   Here the base rate math has a nasty twist...

First, the math for an S candidate signalling G: The 20% of G candidates in the pool will signal G 50% of the time and the 5% of VG candidates will signal G 20% of the time.  Thus, 13.5% of S candidates  will sgnal G when they are G or VG.  Now, the weaker side: The 50% of SA candidates in the pool will signal G 20% of the time, and the 20% of BSA candidates will signal G 5% of the time; the 5% of MBSA candidates will signal G 1% of the time.  Thus, 11.05% of S candidates will signal G when they are SA or below.  The probability that a signal of G actually corresponds to a G or VG candidate equals 13.5/24.55 = 54.99%.

Now, the math for a W candidate signalling G: The 4% of G candidates in the pool will signal G 50% of the time and the 1% of VG candidates will signal G 20% of the time.  Thus, 2.2% of S candidates  will sgnal G when they are G or VG.  Now, the weaker side: The 20% of SA candidates in the pool will signal G 20% of the time, the 50% of BSA candidates will signal G 5% of the time, and the 20% of MBSA candidates will signal G 1% of the time.  Thus, 6.7% of S candidates will signal G when they are SA or below.  The probability that a signal of G actually corresponds to a G or VG candidate equals 2.2/8.9 = 24.72%.

Notice that there is a radical difference here in favor of the S candidate signalling G over the W candidate sending the same signal of G.  To achieve a social objective of treating the two candidates equally, social preferences--or a mix or social preferences and netural preferences--on the part of employers are necessary. 

This example can be used to reinforce the point in the last post about a possible efficiency-based rationale for the stigmatizing of negative social preferences toward a weaker group as racism.  In this case as in the earlier one, negative social preferences toward W do not have the utility that positive social preferences toward W do.

 

 

 

 

Social Preferences in the Context of Hiring and Groups

Missing today's Paduano seminar and an interesting presentation by Stern school prof Dolly Chugh on how fictional graduate students with female names and names associated with minority groups did better compared to those with names associated with white males when they asked for immediate appointments rather than appointments a week away.  (Here's a related paper of hers from AER.)

After reading the paper and moving away from thoughts directly related to the study (e.g., it would be interesting to see what would happen if the white male and for that matter the white female names had a less tony, subcultural sound--e.g., Billy Bob Anderson instead of Brad Anderson or Penny Sue Roberts instead of Meredith Roberts), I got to thinking about how social preference modeling could be applied to hiring and positive and negative discrimination...

Suppose employers fall into three types.   Type 1 is egoistic.  Type 2 has positive social preferences for a weaker group W, meaning that it is willing to incur some costs on behalf of members of that group that are not recaptured by the employer.   Type 3 has negative social preferences for the weaker group, meaning that it is willing to incur some costs against members of that group that are not recaptured by the employer.  

Now suppose employers are deciding whether to hire W job candidates for a job with a standard wage and benefits and a number of incumbents in the firm already performing the job.  All three types of employer share a prior that members of the stronger group S are superior in terms of their average job skills.  Information about the individual quality of S and W candidates relative to incumbent employees can be acquired, but doing so costs the employer more than the return from knowing the information.

If all employers are Type 1 egoists, no W candidates will be hired.  The result is socially suboptimal, assuming that some of the rejected W applicants are in fact more qualified than some of the accepted S ones and that the social cost of determining the identity of some or all of these W applicants is less than the corresponding social benefit of hiring tham. 

The analysis is exactly the same is true if all the employers are Type 3 negative discriminators.  That result of this simple model gets us to a perhaps key point: One reason for stigmatizing negative social preferences against a weaker group as racism is that such negative preferences may well not have any positive social value, compared to simple neutrality with respect to that group.

Finally, if all employees are Type 2 positive discriminators, W employees may be hired, depending on the strength of the social preferences for them.   This may produce a social welfare improvement, depending on whether the social benefits exceed the Type 3 employers’ costs.

 

Leading from Below (more rat ex and social preference notes)

The employer and the employees have an incentive to commit to high wages and high effort; the problem is that each employee has an incentive to defect, since lower effort is not detectable by the employer.

One potential remedy is to align employee incentives with the firm contractually.  Another potential remedy, which might go along with the first, is charismatic, pro-social leadership by an employee who is committed to high effort and whose example brings the others along in one way or another. 

To put the point in terms of hard-headed game theory: To be effective, pro-social leadership from below needs not to be deterred by another employee who engaqges in low effort.  A fixed commitment on the part of the leader to high effort, as opposed to one that is contingent upon what the other player does, can bring the other along if the underlying game is Trust or Leadership.  If the underlying game is the PD, the egoistic other will defect; the remedy for that could be ambiguity as to what the game is (which could lead the egoist to cooperate, since the game might well not be a PD); vengefulness on the part of another employee against the defector harming the saintly co-worker, which could similarly lead the egoist to cooperate; or egoism giving way to at least a measure of shame.

 

 

 

Leading from Below (modeling-rational choice-egoism notes)

[Notes for a paper that would apply the selective altruism concept to workers]

I want to get to a social preferences/beyond egoistic calculation punch line—that there are two main types of partially altruistic, charismatic leadership among “low level” employees like fast-food workers and assembly line workers, one of which helps the firm and society, and the other of which hurts the firm and may hurt society.  To get to the punch line, some set-up is necessary.  

The set-up involves basic rational choice modeling with a standard egoism assumption.  The first core idea to model is that lower level employees may have an incentive to sabotage the firm as a means to get the employer to pay above-market wages.   The second core idea to model is that the very same employees may also have a rational basis to commit to high-level effort to obtain higher than market wages, but that such a commitment is unstable because of the incentive of each individual employee to defect from high-level effort and the difficulty of detecting when that is occurring. 

First, let’s sketch out a sabotage model.  In the game tree, the employer moves first and faces a choice between paying market or above-market (or “efficiency”) wages.  The employee team moves second and faces a choice between having one or more of its members sabotage or not.    The employee then moves again, with the choice being between terminating the employee team or not doing so.  If the employer has been paying market wages, it also faces a choice as to whether to now raise wages above market.   (We will assume that firing a sabotaging team already receiving high wages is better for the employer than lowering their wages, so that choice is assumed away.)  For purposes of the analysis, it will be assumed that sabotage can be detected, but not the identity of the individual saboteur on a work team; the employer in the model thus faces a choice between firing everyone and firing no one.

Here is a back-solve of the logic of a model in which equilibrium is above market wages and no sabotage.   Starting at the end and working backward: The employer faces a choice between firing everyone or no one.  We can assume firing everyone is better for the employer if there has been sabotage and not otherwise.   Knowing that, a team of employees receiving above market will elect not to sabotage, because getting fired from an above market job is costly.   Knowing that, an employer will choose to pay above market as long as the firm’s payoff from doing so is higher than its payoff from paying market.   Assuming that the employer faced with sabotage does best to pay above market going forward to end future sabotage, the employer should simply pay above market from the beginning.

Given the assumptions just noted, the equilibrium is that the firm will choose to pay above market and the employee team will choose not to sabotage.    The outcome is not desirable for the firm compared to an outcome in which sabotage is not a possibility and market wages may be safely paid.   It is also difficult—though not impossible-- to contend that a high wage outcome based on the sabotage possibility is a socially desirable one.    One may assume—as it is reasonable though not clear-cut to do--that low-wage workers should receive more than their market wage through a “social wage” paid by tax credits, public funding of education, and the like.    But for workers to receive above market wages from their employers as a result of employer fear of sabotage would seem to be ethically dubious, and possibly outright wrong.