Spero Consulting

We encour­age inter­ested read­ers to skim a very short essay on Our Con­trol Frame­work to gain famil­iar­ity with the way that we use cer­tain terms in this essay. As time per­mits, we plan to post addi­tional intro­duc­tory and ref­er­ence material.

A Com­mon Fallacy:

″My sub­or­di­nates should be rewarded the same way that I am. That will align their incen­tives with mine, and they will do what they are sup­posed to do.”

We often observe senior man­agers reward­ing their sub­or­di­nates on the same basis for which they, themselves, are held account­able. Our response in those sit­u­a­tion is always the same. We ask why?

Included in our crit­i­cism is the com­mon pol­icy of reward­ing rel­a­tively low-​level employ­ees based upon firm-​wide prof­its or share appre­ci­a­tion, includ­ing cases of grant­ing employ­ees restricted stock in the hope that the grants moti­vates the employee to act appro­pri­ately. The prob­lem is that for very good rea­sons most employ­ees can’t deter­mine which actions and deci­sions will max­i­mize such gen­eral and noisy vari­ables, and the small effect they may have can be eas­ily over­whelmed and hid­den by events out­side their control.

In this essay, we explain how such a fal­lacy can be expensive.

Risk Aver­sion:

Before pro­ceed­ing, note our work­ing premise: most indi­vid­u­als are risk-​averse with respect to compensation.

That means that if every­thing else is equal, such an employee would pre­fer a cer­tain salary to the same level of expected risky pay, e.g., some com­bi­na­tion of lower salary and unknown bonus that, on aver­age, adds to the same amount.

Alter­na­tively, that means for a given level of expected (but risky) pay, the firm could pay a lower salary and keep the per­son just as sat­is­fied (on that dimension).

That’s impor­tant because if the firm can sat­isfy (and retain) the employee with less, fixed pay, it can lower its expenses.

Moral Haz­ard:

The prob­lem is that when the employee’s effort is unob­serv­able or the rea­sons for his deci­sions are unknown, it may be dif­fi­cult or impos­si­ble to moti­vate him with­out pro­vid­ing incen­tive pay that varies with respect to one or sev­eral vari­ables. Huck­le­berry Finn said it best: “Well, then, says I, what’s the use you learn­ing to do right when it’s trou­ble­some to do right and ain’t no trou­ble to do wrong, and the wages is just the same?”

Now if effort is unob­serv­able that means that any sig­nal of effort (worth con­sid­er­ing) must be noisy; it doesn’t per­fectly explain effort. (But, of course, some sig­nals are “bet­ter” than others.)

And, if the vari­able does not per­fectly sig­nal effort, there is a change the sig­nal is wrong, and that imposes risk on the per­son, i.e., he did the right thing but may not be rewarded for it. That impo­si­tion of risk hurts the per­son – increases their dis­sat­is­fac­tion – so, they must be com­pen­sated for it. That addi­tional com­pen­sa­tion is the risk pre­mium or the aver­age, addi­tional pay required.

So vary­ing com­pen­sa­tion based upon some sig­nal of effort, e.g., profit, cost con­trol, sales vol­ume, etc. makes sense only if the firm is attempt­ing to moti­vate the employee to behave in a par­tic­u­lar way. It’s costly, but it is worth doing if it moti­vates the employee to take actions that, on aver­age, increase expected prof­its more than the expected increase in pay.

So, if mon­i­tor­ing the employee is too expen­sive or infea­si­ble, the issue becomes the best way to struc­ture incen­tive pay to moti­vate the per­son to coöper­ate, and the “best way” we mean the cheap­est way to moti­vate the desired behav­ior, and that depends upon the envi­ron­ment, the avail­able sig­nals of the person’s efforts, and cer­tain char­ac­ter­is­tics of those sig­nals. It would be truly remark­able if the solu­tions to all incen­tive prob­lems within an orga­ni­za­tion were the same, i.e., pay every­one based upon the appre­ci­a­tion in share price.

Below, we dis­cuss a few sta­tis­ti­cal char­ac­ter­is­tics, but there are oth­ers not men­tioned here.

Desir­able Char­ac­ter­is­tics of Per­for­mance Measures:

To find the best per­for­mance mea­sures for a par­tic­u­lar task, we advise using a rel­a­tively sim­ple, five-​step process. ¹ Usu­ally, those mea­sures are vari­ables or com­bi­na­tions of vari­ables that (some­what) inform about the person’s deci­sions and efforts. Usu­ally, they are some­thing that the per­son can influ­ence, but they need not be.

When we ana­lyze a ran­dom variable’s suit­abil­ity as a per­for­mance mea­sure, two of the char­ac­ter­is­tics that we con­sider are its sen­si­tiv­ity and pre­ci­sion to the employee’s effort. (If they exist.)

• Sen­si­tiv­ity con­sid­ers how the employee’s effort influ­ences the per­for­mance mea­sure; on aver­age how will the vari­able change as the employee’s effort changes? Visu­ally, one can think of sen­si­tiv­ity as the slope of line relat­ing the expected value of the mea­sure to the employee′s effort. So, in a graph, a vari­able whose expected value changes faster with respect to effort (has steeper slope) rep­re­sents a more sen­si­tive mea­sure than another with a flat­ter slope.
• Pre­ci­sion is the inverse of sta­tis­ti­cal vari­ance and mea­sures the accu­racy of the vari­able as an indi­ca­tion of effort. (Pre­sum­ing those sta­tis­ti­cal val­ues exist.) Pre­ci­sion can be thought of as the nar­row­ness of the (prob­a­bilis­tic) range of val­ues that the vari­able can take for a given level of effort. The higher the pre­ci­sion, the bet­ter the infer­ence about the employee’s actual effort. So, a per­fectly pre­cise mea­sure is the same as observ­ing the employee’s action.

All things equal, when the employee is risk-​averse, increases in sen­si­tiv­ity and/​or pre­ci­sion (of a per­for­mance mea­sure) allow for more accu­rate infer­ences. More accu­rate infer­ences per­mit the pay­ment of a smaller risk pre­mium while still moti­vat­ing the desired behavior.

So, in most sit­u­a­tions, it would be highly unlikely for one mea­sure to be opti­mal for all employ­ees. That means that the use of a sin­gle gross (and less personally-​informative mea­sure) for all employ­ees will (1) increase com­pen­sa­tion costs by mak­ing it more risky (than it need be) and (2) destroy or reduce employees’s moti­va­tion to act cooperatively.

Cor­po­rate Profit or Share Price as Per­for­mance Measures:

So let′s ana­lyze cor­po­rate profit and share price change in terms of sen­si­tiv­ity and precision.

Now, exactly how sen­si­tive is cor­po­rate profit or share price to any individual′s effort? We′ll answer that one: not very.

How pre­cisely does cor­po­rate profit or share price reflect an individual′s effort? Again, not very, espe­cially for lower-​level employees.

Share prices for large cor­po­ra­tions change almost con­tin­u­ously dur­ing a trad­ing day (and are influ­enced by many exoge­nous fac­tors). Can those changes be attrib­uted to any sin­gle employee′s effort or lack of effort through­out the day? Over a month, a quar­ter, a year? No, usu­ally not.

So, is there a com­pelling rea­son to reward employ­ees on such gen­eral measures? Anything is pos­si­ble, but one would have to tell a very spe­cial story about a very spe­cial environment.

Now, before we describe the moti­va­tional prob­lems asso­ci­ated with using the wrong mea­sures, let′s con­sider a sim­pler anal­ogy that avoids those messy human behav­ior and moti­va­tion issues.

Opti­mal Per­for­mance Mea­sures for a Car:

Con­sider an auto­mo­bile. Most of us are famil­iar with these devices used for swift trans­porta­tion; pre­sum­ably their pur­pose is to travel either a cer­tain dis­tance or at a cer­tain rate of speed. Either pur­pose pro­vides a nat­ural per­for­mance mea­sure: an odome­ter for dis­tance and a speedome­ter for speed.

Sup­pose one′s pur­pose for oper­at­ing an auto­mo­bile is speed. From the con­trol per­spec­tive of the dri­ver — say, a race car dri­ver — would one be sat­is­fied with the speedome­ter as the sole mea­sure of per­for­mance? “Don′t need a gas gauge, we have a speedome­ter; don′t need ther­mome­ter, we have a speedome­ter; don′t need tire pres­sure mon­i­tors, we have a speedome­ter… ad nau­seum.”

Unless the reader is par­tic­u­larly ornery, he or she should read­ily agree that the use of only a speedome­ter to mea­sure the effi­cacy of each sub-​function of an auto­mo­bile would not pro­vide enough spe­cific feed­back to allow the dri­ver to max­i­mize his speed or aver­age speed. “Hey, we′re going very fast. There′s no rea­son to stop for gas!”

Such sen­tences seem extremely silly, yet they are anal­o­gous to a pol­icy of ignor­ing indi­vid­ual sig­nals of effort and impos­ing one’s own par­tic­u­lar per­for­mance mea­sure on sub­or­di­nates and no dif­fer­ent than reward­ing all sub­or­di­nates on, say, divi­sional eco­nomic profit when they may have respon­si­bil­ity for only costs or only rev­enues or for neither?[4. Yeah, yeah, yeah, any infor­ma­tion econ­o­mist worth his salt should be able to eas­ily con­struct counter-​examples, but that is clearly not the point of the essay.]

So, if you want to get an indi­ca­tion of whether some­one did what they were sup­posed to do, mea­sure and account for (at least) some vari­ables in their par­tic­u­lar (decen­tral­ized) envi­ron­ment, includ­ing ones that they con­trol or influence.

Using overly-​general ones impose unnec­es­sary risk on sub­or­di­nates, and if those sub­or­di­nates are risk averse — and who′s not — that impo­si­tion leads to either higher expected com­pen­sa­tion costs and pos­si­bly lower effort and the sub­se­quent reduc­tion in out­put or higher employee turnover. ²

In addi­tion, we’ve tried to explain, in a non-​technical way, why arti­fi­cially con­strain­ing the num­ber of mea­sures reduces trans­parency and the abil­ity to con­trol. (Again, think of a car with only a speedometer.)

Beha­vo­r­ial Implications:

As yet we haven′t described the more dam­ag­ing qual­i­ta­tive aspects of such poli­cies. Here are a few behav­iors that we have observed:

• Dis­tracted employ­ees with an unnat­ural fix­a­tion on share price rather than the task at hand.
• The sense of alien­ation that fol­lows from these types of realizations.
  • “My job’s not impor­tant enough for any­one to care about (or under­stand or mea­sure) what I do; so, I get paid for what my boss does.”
  • “I’ll get paid when he (or she) gets paid; oth­er­wise, it doesn’t mat­ter.” This might include the follow-​up: “So, I’ll spend my time suck­ing up. That will increase my chances of get­ting paid more than actual work will.”
• The sense of for­saken­ness that fol­lows from the con­clu­sion, “It doesn’t mat­ter what I do, I can’t influ­ence those things; so, why bother?”
• The sense of the orga­ni­za­tion is arbi­trary as in the fol­low­ing sen­ti­ment: “They only pick those things because they can manip­u­late them. That way, if they don’t want to pay us they don’t have to. They have eleven ways to win any­way. So, why bother?”
• Cases where employ­ees will take greater risks because they think that no one is watching.

Through the years, we′ve heard many other sim­i­lar com­ments and sentiments.

So, where should one begin?

By ask­ing: What do you want the employee do? And, how can you tell if they′re doing it? (Not per­fectly, but statistically.)

To do that, one needs to under­stand their jobs; how their jobs cre­ate value. If you need more help than that, please give us a call at 1.412.779.9028. That is why we are here.

By the way, if after read­ing our essay you are still think­ing: “We’re a team, we have a com­mon goal and should a com­mon mea­sure,” then we rec­om­mend that you try coach­ing a sports team with that sin­gu­lar con­trol mech­a­nism. If that is too onerous, please reread the above essay.

Finally, if you do agree with our analy­sis, then you might also won­der why many con­sult­ing firms sell a sin­gle per­for­mance mea­sure as the unqual­i­fied best in every case. If one doesn’t (gen­er­ally) exist within a firm, how could it exist across the entire econ­omy? Our hunch is that it is best for them, but nec­es­sar­ily for you or your sub­or­di­nates or shareholders.

Copy­right © 2008 2010, Spero Consulting.

---

Foot­notes: