CHAPTER 5 RELEVANT DATA AND SUGGESTED TESTS This chapter presents data showing that the Drive-Effort Measure influences economic behavior. The data make especially clear that poorer people will expend more time and effort to increase their economic position than will richer people. The chapter also discusses types of data which could falsify the hypothesis if collected in future research. The data that are presented do not "prove" the Drive-Effect hypothesis. But taken together, they would seem to accord the hypothesis sufficient credibility so that it should be considered seriously. Many of the citations in this chapter are derived from my earlier work on quite different topics. This suggests that there is likely to be a large volume of other relevant studies and data which I have not fallen upon in the course of other sorts of excursions, and which the research for this volume has not yet turned up. I will be grateful for suggestions about other relevant materials. INDUSTRIAL PAYMENT SYSTEMS Schemes of payments to industrial workers vary in the extent to which they are directly tied to the individual's output. Therefore, the systemsdiffer in the opportunities available for the individual to increase his/her purchasing power. Economic theory seems to be silent about the relationship between the effort exerted and the scheme by which a person is paid -- flat wage, individual piecework, group piecework, etc. (For the recent state of thought and some evidence, see Hamermesh and Rees, 1984, pp. 51 and 52.) The Drive-Effort hypothesis, however, implies that because the relationship between effort exerted and pay received is closer in a piecework scheme than in a flat wage scheme, the former should elicit more effort than the latter -- a proposition which seems quite commonsensical but has not hitherto been proven rigorously, to my knowledge. Pencavel (1977) focuses upon a different mechanism than property- rights incentive in his analysis of the modes of payment and their results. He sensibly suggests that employers are willing to pay workers more with a piece-rate systen than with a time-rate system because less supervision is needed with a piece-rate system, with a consequent saving to the employer; he calls the latter an "on-the-job screening system." One might add that the greater consistency from week to week of worker output under a piece- rate system (as documented in studies referred to below) also could be a benefit to employers. Seiler (1984, Table 2) found a bit more supervision for time workers than for incentive workers in the footwear industry. (0.7 versus 0.6), but much less supervision for time workers in the men's and boys'-suits-and-coats industry (0.02 versus 0.07); which runs against Pencavel's bypothesis. The much larger standard deviation for time workers suggests that this latter comparison may be erroneous, however. If Pencavel is correct, one would expect the payments to workers to be proportionally greater under a piece-rate system than the proportional increase in productivity; this would be consistent with lower costs of supervision under a piece-rate system. I do not know of any data testing this, however, so Pencavel's hypothesis must be considered as consistent with the data, but no more consistent with it than other explanations. It should be noted that even if Pencavel's explanation is fully supported, it would not be inconsistent with a property-rights incentive explanation of worker behavior. Rather, it simply suggests that a piece- rate system motivates with money whereas a time-rate system motivates with fear of supervisory action. Hamermesh and Rees offer another mechanism which is consistent with a property-rights incentive explanation. They suggest that workers are self- selected to work in a piece-rate system. If so, there is no difference in effort by the same sort of person to be explained by the mode of payment. By the same token, the self-selection explanation is not inconsistent with Pencavel's on-the-job screening system. Before beginning on the systematic evidence, here are a few anecdotal accounts: 1. Southern Utah Fuel Company's work force is largely Mormon, as is its management. Partly because Sufco's workers are so devoted to the company, they dig an astounding 2 1/2 times more coal per man per day than is dug in the average underground coal mine in the U.S. That is good for the mine, of course, and the workers enlarge their salaries by cashing in on production incentives. (Wall Street Journal, "Great Productivity at a Utah Coal Mine Is Called `God's Plan'," April 12, 1984, p. 1.) 2. Lots of companies these days are huffing and puffing to find some good way to motivate workers. Lincoln Electric Co. found the way it wanted in 1907 and says it has liked the results ever since. The company relies on incentives. It pays most of its 2,500 employees on a piecework basis. In 1933, it added an annual bonus system. Based on performance, bonuses may exceed regular pay, and they apply far more extensively than in most companies. A secretary's mistakes, for example, can cut her bonus. Some employees complain of pressure. But they stay around. Turnover is only 0.3% a month. The Cleveland company has no unions, and it avoids strikes. It says employees have averaged as much as $45,000 a year in good times. Sales and earnings have risen at a respectable pace. (Wall Street Journal, "Ohio Firm Relies on Incentive-Pay System to Motivate Workers and Maintain Profits," August 12, 1983, p. 15.) 3. [W]here pay-for-performance has been universally and rigorously applied, it has often had dramatic results. Frank Schultz, a BankAmerica senior vice -resident, runs the company's credit card division. He has put all of his 3,5000 employees under a pay-for- performance system, which ranks them against their peers, using 200 specific criteria. "I measure everything that moves", Mr. Schultz says. ("Back to Piecework: More Companies Want to Base Pay Increases on the Output of Their Employees", The Wall Street Journal, Nov. 15, 1985, p. 15) The following studies show that the expected effect of incentive plans is indeed observed: 1. In eight auto-repair occupations Workers in auto dealer repair shops who were paid under incentive plans nearly always averaged more per hour than those paid time rates, according to a Bureau of Labor Statistics survey of 36 metropolitan areas in June 1973. Typically, the incentive-paid workers earned 20 to 50 percent more per hour than their time-rated counterparts. With only one exception (lubrication workers in San Francisco), incentive-paid workers averaged more than time-rated workers in the same occupation wherever both systems were used. (See Table 5-1) The earnings differentials varied widely, however, among occupations within an area and for an occupation compared across areas (King, 1975, pp. 45-46). --------- Table 5-1 --------- 2. Rothe (1951, p.96, and earlier publications cited there) suggested that consistency of performance is a sensible measure of the "effectiveness of incentives." It seems reasonable that great fluctuation in a worker's performance from period to period indicates a relatively low level of effort, given the concept of effort that is under discussion here. Rothe (p.94) compared 16 weeks of output rates for a) chocolate dippers in a candy factory who were "paid a one-to-one ratio for performance over standard," against b) the output records of butter wrappers and machine operators, neither of whom were paid on the basis of performance. The correlation from week to week for the chocolate dippers (r = .85) was much higher than for the other two types of workers -- an "enormous range" for butter wrappers, and r = .57 and .68 from fortnight to fortnight for the machine operators. Rothe also cited a study by Tiffin (1942) which showed a very high week-to-week correlation coefficient (r = .96) for hosiery loopers, who were paid on an "incentive system." 3. Kilbridge (1960) went beyond Rothe's idea to suggest that effective incentives not only reduce variation from period to period for the individual worker, but also tend to reduce variation among workers. And he displayed the distributions of output relative to the mean output for two incentive schemes, one he considered effective and one he considered ineffective, data which he takes as confirming his hypothesis (his tables 1 and 2). The "ineffective" plan was used in machining departments, whereas the "effective" plan was used in a clerical department; in principle, this difference in locus could be responsible for some or all of the difference in distributions, but Kilbridge apparently did not think it was so. On the other hand, Hamermesh and Rees (1984, p. 52) expect that "the distribution of total earnings is much wider within a narrow industry among workers who receive incentive pay than among those paid by the hour." Seiler (1984) analysed data on over 100,000 workers in 500 firms making "footwear" and "men's and boys' suits and coats," and found more dispersion among earning of piece-rate workers than among time workers. This seems to be a very careful piece of work using a large sample and effective statistical controls, and therefore probably deserves more confidence than does Kilbridge's earlier study. 4. Seiler (1984) offers solid evidence that incentive workers earn more--14% more--than do time workers. And after allowing for "risk" (in the form of dispersion of earnings) he finds that the "effort earnings hypothesis" explains most of the difference in earning under the two systems (p. 375). Perhaps the most careful study of time versus piece-rate systems is Pencavel's analysis (1977) of the data collected by Schultz and Rees (1970) in Chicago. He held constant such human-capital factors as schooling, and found that workers paid by the piece earned 7% more per hour. He did not, however, show that piece-rated workers produced more per hour, and if his analysis of the economics of payment modes is correct, productivity could be less different (or no different) than earnings. But taken together with other studies, it seems reasonable to conclude that productivity as well as pay is higher in a piece-rate system. 5. Weinstein and Holzbach (1973) compared, in a laboratory setting using college students as subjects, the effects of two payment plans: a) an "individual" incentive plan -- actually, different proportions of their three-person group's total payments, depending on their relative performance within the group, the group's earnings being a direct function of output; and b) a group incentive plan -- all three members of the group receiving equal shares of the group's earnings. Production was higher with individual incentives. Weinstein and Holzbach also split the groups into what they call high and low "task-flow interdependence," a measure of how an individual work rate depends upon how fast the other persons are processing their work. (This is actually the case only for the second and third workers in the group, though the first worker in the flow sees his/her effect upon the others, presumably). Low interdependence yields higher productivity than does high interdependence. The order of the groups from best to worst performance (aside from a possible reversal with respect to the most highly qualified workers, in the individual reward condition) is: individual incentive-low interdependence; group-low, individual-high; and group-high. These results are consistent with expectations concerning the effect of payment system upon effort. 6. Farr (1976) added two conditions to the Weinstein-Holzbach design: no incentive at all, and pure individual incentive. The main result is that all incentive plans do better than plans with no incentive. The group incentive does slightly better than the combined incentive, which in turn does slightly better than the individual incentive. It is interesting that psychologists have found it necessary to conduct laboratory experiments (and on-the-job studies, as reviewed above) to establish the existence of what would seem to be a commonsensical idea. This is in part because various theories concerning "equity" and job satisfaction have suggested that the commonsensical view is inadequate. And, according to Farr (p.159), "The theoretical base for much of this research has been operant conditioning." Whatever the merit of these contrary speculations, the pure incentive effect seems to dominate empirically. --------------- INSERT V-76 Terborg and Miller (1978) compared piece-rate versus hourly pay in a laboratory experiment with students. The quantity produced was higher with piece rates. Quality may have been lower, but the result was not statistically significant. If quality was indeed lower, over- all value of production may not have been greater with piece rates. _______________ 7. In the Weinstein-Holzbach and Farr experiments, in the group- incentive treatments the rest of the group presumably knows the performance of each individual. This knowledge could affect performance, and indeed, could account for the group incentive plan doing as well as the individual incentive plan. It is a fact of life that the regard of one's fellows is an important reward, and the desire for it is an important incentive. Therefore, we must consider those experiments as having two jointly-working incentive systems. Examining the two systems together may yield appropriate information for decision-makers studying which alternative plan to adopt. But that jointness muddies the waters as we seek to learn about the effects of opportunity upon effort. 8. Another group of experiments compares (implicitly) the effect of incentives that are more dependent versus those that are less dependent upon one's own productivity when the rest of the group presumably has little or no knowledge of each individual's output. In group rope-pulling, and in group shouting and clapping, the size of the cooperating group has an inverse relationship upon the per-person effort expended (Maital, 1982, pp. 126, 127). One may wonder why a person will exert any effort if one's output is unknown. Game-theoretic notions come into play here. But it may also be that in these experiments, individuals did not believe that their output would be completely unknown to others, and therefore they did not try to be completely-free riders. It would be interesting to push this line of research even further by designing experiments wherein the individual's output is known to a varying degree by others, ranging from full knowledge down to no knowledge at all. 9. More generally, a wide-ranging review by Locke et.al. (1980) of a large number of studies found money payment schemes to be "the most successful techniques for motivating employees", compared to "job enrichment", "goal setting", "participation", and the like. These results must be seen as something of a blow to social theorists who have hoped that restructuring society could induce high productivity without the "capitalist" methods that depend upon what is often called "greed". This section should not be taken as an endorsement of particular piecework incentive schemes in industry, because group pressures can operate to subvert the scheme. As noted by Lupton and Cunnison after a study of three piecework plans in Great Britain: If an operative, or a group of operatives, observes that the relationship between the effort they exert, and the cash reward they receive, is affected by factors which are beyond their control, it seems natural that they will attempt to redress the balance by operating such controls as are open to them. Depending on circumstances one may find attempts to "fiddle" on booking, on quality, or on work allocation. The extent to which these efforts are successful will depend upon factors peculiar to each workshop; such as the strength of Trade Union organisation, the degree of machine pacing of the process, and the like. (1957, p. 269) One should notice, however, the underlying assumption made by those authors that effort can be influenced by the payment scheme. HOURS WORKED Wealth and Hours Worked1 *** Though the amount of data pertaining to work time that are reviewed here is relatively large compared to the amount of data on work intensity, one should not be misled into thinking that the DrEf hypothesis refers mostly to the time aspect of effort. Rather, intensity of work effort is probably the more important and certainly the more interesting aspect of the phenomenon, both with respect to the magnitude of effort as a variable and also in the power of the DrEf hypothesis to explain and predict economic behavior. The data connecting wealth and hours worked may be explained by Becker's time theory. But they also are consistent with the Drive-Effort hypothesis, which is why they are presented here. (The data presented on hours worked with respect to variables other than wealth pertain _o_n_l_y to the Drive-Effort hypothesis, and not to Becker's theory, as is the case with the data on work intensity.) l. Kindleberger (l965) plotted the logarithm of hours of work in manufacturing in l96l against wealth (actually income2) and found *** a very clearcut relationship, as seen in Figure 5-1. And from similar data Winston (l966) estimated an elasticity of -.l. ------------ Figure 5-1 ------------ 2. Data for non-agricultural weekly work hours in the U.S. starting in l870 show a strong inverse wealth-hours relationship until well after World War II (Simon,1977, p.60), though work-week hours apparently stabilize sometime after that. Becker suggests that the apparent stabilization is due to changes in the time-intensity of the baskets of goods that consumers have purchased over the years, and if this is true, the absence of relationship in the recent period does not falsify the hypothesized wealth-work relationship. However, Rees (l979) suggests that the apparent stabilization of "average" work-week hours may be an artifact, because people may be reducing the number of weeks worked per year, and thereby reducing total hours worked over the year, even though reported "representative" work-week hours remain unchanged. The data in Table 5-1 showing the trend in the U.S. in lost workdays due to ailments from 1973 to 1980 support that supposition. (Though data in Hamermesh, 1984, goes the other way.) And the data in Table 5-2 show that during the period 1957 to 1971 for which data are available, effective working time in Germany declined much more than did paid working time. ---------- Tables 5-1 and 5-2 ---------- Table 5-3 shows the effective working time in various European countries for 1970 to 1981, for completeness of presentation as well as for general interest. --------- Table 5-3 ----------- [[NOT IN DISKS SENT TO BLACKWELL: At a micro-level, it has not been easy to establish the relationship of wealth to work hours net of the effects of wages and standardized by income, age, and other relevant variables. Smith (l977) argues, for example, that studies relating hours of work to assets in a one-period framework will produce misleading results (he refers to several such studies which he says are flawed in this respect), and he suggests that the subject needs to be studied in a life-cycle framework; Smith also shows that, contrary to expectations, the coefficient for net worth's effect on hours of work is _p_o_s_i_t_i_v_e for both men and women in the l967 Survey of Economic Opportunity (SEO) data. But this finding does not vitiate the findings of an inverse relationship in national time series or cross-sections, because most of the bias to which Smith alerts us should wash out in aggregate analysis. And such difficulties have not made itimpossible to draw some conclusions from the many studies that have been made. Killingsworth (1983) has seemingly-carefullyreviewed and summarized the large body of micro-level studies (as well as many aggregate time-series studies) of the relationship of wealth to quantity of work, both static and dynamic labor-supply models. He concludes as follows: "There is now a great deal of empirical evidence about labor supply behavior. First, and of crucial importance, pecuniary variables -- wage rates and property income [the latter being the wealth variable of interest here] -- do generally seem to have something to do with labor supply." (p. 432). That is, greater wealth leads to a less work being supplied. ]] ------------ Mark R. Killingsworth, Labor Supply (New York: Cambridge U. P., 1983) -------------- 3. Men with working wives moonlight less than men whose wives do not work (Guthrie, l966). A working wife is an important economic asset; the man whose wife works is effectively richer than the man whose wife does not, and hence it is consistent with the hypothesis that the former works less than the latter. 4. Men with "disorderly work histories" moonlight more than do other men (Guthrie, l966). A man with little job security has poorer prospects in present-value terms than does the man whose job is secure, and hence it makes sense in terms of the Drive-Effort hypothesis that he works more. 5. Even _s_u_b_j_e_c_t_i_v_e poverty leads a person to work more. Wilensky (l963) found that men who feel less well off than their parents are more likely to moonlight. 6. Shisko and Rostker (l976) explored the effects of moonlighting in a thorough multivariate fashion. Though moonlighters have less income from assets than do non-moonlighters, on average, and though the wealth term is negative in their multivariate analysis, it is not statistically significant. This study, however, may also fall victim to the statistical difficulty pointed out by Smith, mentioned just above. Transfer Payments and Hours Worked One of the great social questions always has been the effect of various sorts of welfare and transfer payments upon the recipient's propensity to work; the underlying idea, of course, is that the wealth bestowed by the payments decreases Drive-Effort. The negative income tax experiments in the United States were intended to answer this question better than other sorts of evidence that had been adduced in the past. The results of those experiments have been interpreted in a variety of conflicting fashions. But there does seem to be consensus that there is a significant reduction in annual hours worked by the recipients of payments.(See Burtless and Greenberg, 1983, p. 387, for estimates I believe to be soundly based.) ----- add to reference list: Gary Burtless and David Greenbert, "Inappropriate Comparisons as a Basis for Policy: Two Recent Examples from the Social Experiments", Journal of Public Policy, Vol. 1, Aug., 1981, 381-399. Numbers_of_Children_and_Hours_Worked More children in the family imply lower wealth per person, ceteris paribus, and therefore more children imply more work by the parents and by individual children, by the Drive-Effort hypothesis. Data on the matter follow. l. In the context of agricultural work by Russian peasants around the turn of this century, Chayanov (l923/l966) showed that a larger number of children (as measured by the ratio of consumers to workers) led to a larger number of work days, a larger output in rubles, and more land farmed per farmer. From these data I estimated an elasticity of work with respect to the number of consumers as .53.3 *** 2. Yotopoulos and Lau (l974) estimated a labor-supply function from Indian data. It is relevant here that their regressions held constant such variables as demand, prices, wage rates, debts, and especially the number of workers in the household. After these factors are allowed for, the elasticity of labor supply with respect to the total number of household family members is a huge l.l2 when their entire model is estimated simultaneously. (The elasticity in a labor-supply regression estimated by itself is .67.) 3. Perhaps the strongest evidence concerning the relationship ofdependents to agricultural work per worker comes from Scully's careful study (l962) of 38 Irish farms. He held constant farm size, soil type,costs per acre, and density of livestock, and examined the effect ofnumber of children4 separately upon (a) gross output per acre (total ***money sales), and (b) family income per acre (gross output less costsother than family labor). The results are that an increase of one child(in the mean family with 4.8 children) led to an increase in poundssterling of l.l33/23.7 = 4.8% in gross output per acre, and of .996/l5.2= 6.5% in family income per acre. If one assumes that a young child ison the average equal to half an adult consumer-equivalent -- certainly onthe high side--then a two-parent family with 5.8 rather than 4.8 childrenhas 4.8 rather than 4.4 consumer-equivalents, a difference of .5/4.4 = ll%.The relevant elasticity of work supplied with respect to number of con-sumers may then be estimated as (.048/.11) = .4 or (.065/.11) =.6. 4. Smith's (1977, p. 568) previously-mentioned analysis of SEO family data showed that men's work hours are greater by ll6 hours per year with each additional child. 5. Hill (l97l, p.384) found that among the poor white head-of- household respondents to the SEO, more children are associated with considerably more hours worked. For example, having a third child is associated with 2l9 extra hours of work yearly, a fourth child with l70 extra hours, and a fifth child with l22 extra hours, or about 5|, 4, and 3 extra weeks of work a year -- large effects indeed. 6. From survey data of the National Bureau of Economic Research and the Michigan Institute of Social Research, respectively, Landsberger (l97l) estimated .49 and 2.0 extra weeks of work per child under 6. These estimates should be more reliable than Census data. 7. In Israel's l97l Survey of Manpower data on hours worked the previous week, Gayer (l974) found that for each additional child, the head of the family worked an extra l49 hours per year, i.e., more than three extra hours per week, an increase of about 7% in work for each child. (It is interesting that the increase in the father's work time is more than three times the decrease in the mother's work time, 42 hours per year per child, so the additional baby apparently increases the total work done.) And in a study of Israeli daily time-budget data, Gronau (l974) found that fathers spend an additional 4% more time work-ing per additional child age 0-5 or 6-l2. 8. Moonlighting is positively and strongly related to number of children (Perella, l970), as Table 5-4 shows. --------- Table 5-4 --------- 9. Shisko and Rostker (l976) found a statistically-significant and economically important effect of family size on propensity to moonlight (elasticity of .350) in their multivariate Tobit analysis. One might speculate that the mix of goods that is purchased shifts with family size in such a fashion that the time-intensity rather than the wealth effect explains the larger amount of work done where there are more children. The possibility is raised here only for completeness, however, and not because this is likely to be a major explanation, in my judgment. Also, in addition to the positive effect of children on father's labor- force work, there is a negative effect on mother's labor-force work. But the latter is not relevant here, because our interest is with effort and not with a cost-benefit analysis of children. Overtime and Wealth Data on this relationship would be most appropriate and welcome, but I have found none. _O_v_e_r_t_i_m_e_ _a_n_d_ _N_u_m_b_e_r_ _o_f_ _C_h_i_l_d_r_e_n Ditto. SAVING Differences in saving behavior, along with willingness to work, are frequently invoked as an important variable in discussions of differential economic development among countries, between groups within the same country,5 and in discussions of the decline of nations. Even if *** physical saving is not a key variable in an advanced society (though this is surely going too far), if the act of saving comes from the same psychological source, with the same causes, as does the exertion of effort, then it may be that a high rate of physical saving in an advanced country could be thought of as a proxy for a more important economic variable -- Drive-Effort -- rather than as a causal variable in itself. This sort of suggestion is in the same spirit as Kuznets's analyses of institutions and cultures in various countries as being responsible for other variables (or sets of variables) which might otherwise be thought primarily to influence each other. As analyzed by Modigliana and others in a life-cycle framework, lesser wealth leads in a straightforward fashion to a larger rate of physical and financial saving. Interestingly, Becker's analysis of the relationship of wealth to work time discussed earlier is quite parallel. The act of saving involves the transfer of purchasing power from one present conscious self to some future self with the same identity who will follow the present self. (The view of the person as a set of selves in the present, with other sets of selves to be active in the future, is crucial to making sense of this phenomena. The classic treatment is that of James (1890/1963, Chapter XII). Such "giving" from a set of selves in the present to other sets of selves in the future, by way of postponement of consumption, inevitably produces pain, just as does giving to another person; it is sensed as a "loss" or diminution which -- in all except unusual cases of persons who feel relieved of the burden of wealth of any kind when they give it away -- is unpleasant. Some drive is required for such an act; experiencing such a drive is one of the elements of what is here called effort. This view of saving is consistent with the observation that a larger proportion of windfalls is kept for the present selves than is given to future selves. (Transaction and calculation costs might explain why the effect is particularly strong for small windfalls, however.) This view also is consistent with a larger proportion of each year's income being saved than is saved out of windfalls, if one assumes that much saving is done on a long-term plan and that the self doing the planning at time t = 0 is mostly transferring, not from the present sets of selves to future selves, but from specified future sets of selves to other sets of selves even further into the future; analogously, it is easier to distribute wealth or goods fairly between two other persons than between "oneself" and another person. This also makes sense of Christmas Clubs and other self-forced savings systems; the present self is willing to compel future selves to experience pain, knowing that they (as is the case with the present set of selves) do not like to do so. This accounts for installment payment plans in which the first payment is deferred; without checking, I'd bet that many Christmas Clubs have the same feature. It is still necessary to square this view with income-group cross- sectional data, international cross-sections, and historical data. Constancy of the historical saving ratio (in the United States, say), and absence of relationship among countries with different income levels, would be consistent with the idea that the distributions of individuals on this dimension are similar within nations . And lower-income groups could show a lower ratio of saving to income if a lower individual saving ratio (especially with respect to human capital) causes a person to be in a lower income bracket; this would be expected within a population but not across populations. One can therefore explain all types of data consistently and parsimoniously without invoking any psychological propensities that differ by income groups, and without doing much damage to the permanent income hypothesis or to the life-cycle hypothesis. (Behavior with respect to windfalls is, after all, of minor economic significance, and of interest mostly as a diagnostic device in the study of consumption.) Saving and Wealth l. Patinkin's survey (l965, Note M) of studies done up until that time led him to conclude that wealth has a negative effect upon saving. 2. Mayer's later survey (l972) reached the same conclusion asdid Patinkin's review. 3. Though there is controversy about the size of the effect, allwriters apparently agree that wealth in the form of Social Securityentitlements has a negative effect upon saving.6 *** 4. Blacks save more than do whites with the same current income.7 *** Wealth -- measured either as assets or as lifetime income -- is likely to be lower for blacks than for whites with the same income, which suggests a negative effect of wealth upon saving. 5. Using hypothetical questions about saving behavior under a variety of stipulated conditions, Simon and Barnes (l97l) found that saving would be less where wealth was postulated to be higher, and of a magnitude roughly in accord with predictions based on a life-cycle analysis of the stipulated conditions. Shefrin and Thaler (1983) view the deviations from theoretical expectations in the Simon-Barnes study as being consistent with their self-control theory of saving, which has much in common with the view stated here. Number of Children and Saving The relationship of saving to the number of children is a matter of controversy (Laumas and Ram, 1982) .8 Theory is ambiguous about the *** direction in which the relationship may be expected to run. On the one hand, additional children imply lower wealth per person. On the other hand, the time profile of consumption is pushed more towards the present when young adult earners acquire a child, because expenditures on the child will be spread over only (say) the next l5 years, whereas expenditures on the adults will be spread over the next (say) 35 years. This subject has not been well explored analytically, and would seem to be a fruitful topic for study. Wealth_and_Aspirations The higher the economic status of college students in sororities at Berkeley, the lower their "achievement orientation," as measured by such variables as a) high grades, and b) intention of working after graduation (Selvin and Hagstrom, l963). Working for high grades in school can be interpreted as giving up leisure in order to increase later income. So this result seems a clearcut negative effect of wealth upon intended effort as measured by achievement orientation. Length of Unemployment and Reservation Price The longer a person remains unemployed, the more that assets are depleted, and the poorer the person becomes. It therefore fits the DrEf hypothesis that the asking price for one's labor declines with unemployment as Kasper ( ) finds in an American questionnaire study, and as Lancaster and Chesher (1983) conclude from two English questionaire studies. Strengthening this explanation is that a hypothetical increase in the unemployment transfer payment size leads to an increase in the reservation wage (Lancaster and Chesher.) But there are also competing explanations for the phenomenon -- downward adjustment in expectations of finding a job, and satiation with leisure (the latter suggested by Kasper). -------- add to ref list Tony Lancaster and Andrew Chesher, "An Econometric Analysis of Reservation Wages", Econometrica, 51, Nov. 1983, 1661-1676. ---------- TESTING THE DRIVE-EFFORT HYPOTHESIS A variety of data that seem consistent with the Drive-Effort hypothesis have been discussed. Such data may buttress the theoretical argument and increase belief in the hypothesis. But no set of these data constitutes a direct confrontation between theory and data; none providesa test which was constructed for the purpose of trying to falsify the theory at hand. This is not to say that only Popperian falsification matters; in my view, all evidence that is relevant to the subject at hand should reasonably affect one's belief in a theory, rather than belief hanging only on some single formal testing criterion. But one should not stop with an ad hoc selection of empirical materials, either. Two sorts of devices that might provide a rigorous test are as follows. First, one might systematically examine evidence separately on the two links of the chain, running from the Drive-Effort Measure to expended effort, and then from expended effort to economic performance. Though no tests have been set up and carried out specifically to attempt falsifying the theory with respect to these two links, there is at least one systematic study, by Filer (1981), which seems to throw direct light on the latter connection. Filer showed that people with a higher level of what he calls "drive"--by which he means general activity level--earn higher wages, ceteris paribus. And it is reasonable to assume that a higher level of wages is correlated with a higher propensity to accept economic opportunities. Among the personality "traits" that Filer (1981) studied, "By far the most important personality trait appears to be general activity, or drive." This tends to confirm that the theoretical intervening construct used in these chapters not only is observable but also has the hypothesized relationship to the dependent variable, economic output. For the Drive-Effort hypothesis to be persuasive, there must be a connection between effort and various economic conditions. That is, it must be true that "Effort is a variable," as Maital (1982, p.122) puts it. In support of this view, Maital (p.126-127) cites the evidence mentioned on page 000 showing that the size of the cooperating group has an inverse relationship on the per-person effort expended in group rope-pulling, and on group shouting and clapping. One might go further and examine systematically whether the Drive-Effort Measure has a (lagged) positive effect upon earnings and entreprenurial behavior of individuals and groups, across a sample of nations, or of industries, or both. Wealth might serve as the proxy for the Drive-Effort Measure, which usually is not directly observable. But it would be extremely difficult to construct an after-the-fact study which would deal satisfactorily with the relevant control variables. Another way to tackle the matter would be with hypothetical questions. In this technique, subjects are asked what they would do under a given set of conditions. The hypothetical conditions are then altered, and the respondent is again asked about his/her behavior under the new conditions. The difference in people's descriptions of their hypothetical behavior may then be analyzed with respect to the difference in conditions. For example, this technique has been used to estimate demand curves by asking people how much of various goods they think they, or others, would purchase at particular prices (Gilboy,1931; Schaeffer, 1959; Simon, 1965). The technique has also been used to explore the consumption function (Simon and Barnes, l970). For present purposes, we might specify certain opportunities--say, giving blood for a fee, undertaking a dangerous job, working overtime--under a variety of specified conditions of age, family size, wealth, wages and so on. Such a test has less appeal for economists than does actual market behavior. But such a test does have the great virtue of being thoroughly controllable; the comparison between a given respondent's own responses under various conditions holds constant the respondent's demographic, economic, and cultural background, and hence makes easy the sampling requirements. And it is possible to examine whether the differences themselves are influenced by the respondent's background. (They will likely not be so influenced under easy assumptions about life-cycle and consumer rationality.) Animal experiments of the sort carried out with apes and token slot machines might also test this theory well, and especially the shape of the Drive-Effort Measure with respect to wealth. For example, one could vary the ape's stock of tokens and observe the amount of effort the ape will expend to "earn" additional tokens in a given period of time. One could also vary the "purchasing power" of the tokens to see how effort is affected. I have not found any work of this sort done heretofore, however. CONCLUDING REMARKS The aforegoing data, taken together, would seem to demonstrate that there is an inverse relationship between wealth and work effort, as the Drive-Effort hypothesis suggests. Evidence has not been presented showing that the second element of the Drive-Effort Measure, a larger payoff, induces more effort, ceteris paribus, but this proposition would not seem to require any demonstration. Together, these elements support the view that more effort will be produced when the Drive-Effort Measure is higher. No evidence has been provided here on the specific shape of the function; that will have to wait for further study. But the specific shape is not crucial in the applications discussed here. D/l36A,#3 85-37 Effort5 1-17-86 Table 5-4 Percent Holding _N_u_m_b_e_r_ _o_f_ _C_h_i_l_d_r_e_n_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _T_w_o_ _o_r_ _M_o_r_e_ _J_o_b_s No children under l8 years 6.0 l child under l8 years 7.8 2 children under l8 years 8.9 3 or 4 children under l8 years l0.5 5 children under l8 years ll.3 Source: Perella, l970. FOOTNOTES 1Much of this section is drawn from Simon (l966). 2Income per person, the variable used in all the studies mentioned in this section, is an excellent proxy for wealth in a long-run groupaverage context. With respect to consumer-held wealth in housing and other physical assets, lifetime income is on average a good measure of total wealth, and holdings of consumer assets are surely adjusted to lifetime income by households. With respect to producer wealth, the relative invariance in the capital-output ratio across nations and time-periods enables us to work backwards from income (output) to wealth (though recognizing all the defects in national estimates of capital, such as farmland evaluation). 3For a summary of this material, see, Simon, Julian L., The Economics of Population Growth (Princeton: Princeton University Press, l977), pp. l89-l92. 4On only 9 of the farms were there children of ages where they contributed any work at all to the farm, and hence the number of workers "has had little influence on the results of the analysis" (Scully, l962, p. l2l. 5For further reading see: Myrdal, Gunnar, Asian Drama, Vol. 1, (New York: Pantheon, l968); and Nair, Kusum, Blossoms in the Dust (New York: Praeger, 1962). 6For example, see, Feldstein, Martin, "Social Security, Induced Retirement, and Aggregate Capital Accumulation," Journal of Political Economy, Vol. 82, No. 5, Sept.-Oct., 1974, pp. 905-926. 7For references, see Simon, R. and J., l968; and Hodges, Lloyd Curry, Clothing and Race: An Examination of the Effects of Race on Consumption, Ph.D. Thesis (University of Illinois, 1982. 8For a general earlier summary, see Simon, Julian L., The Economics of Population Growth (Princeton: Princeton University Press, l977), Chapter l0. See also the negative effect of children on saving found by Smith, James P., "Assets, Savings, and Labor Supply," Economic Inquiry, XV (4), 1977. Concerning the continuing controversy about the cross-national relationship among Leff, Nathaniel H., "Dependency Rates and Savings Rates: A New Look," in J. Simon and J. DaVanzo, (eds.), Research in Population Economics, Vol. 2, (Greenwich: JAI Press, l980); Billsborrow, Richard E., "Dependency Rates and Aggregate Savings Rates Revisited: Corrections, Further Analysis, and Recommendations for the Future," in J. Simon and J. DaVanzo, eds., Research in Population Economics, (Greenwich: JAI Press, l980; and Ram, l98l?. 3King, Sandra L., "Incentive and Time Pay in Auto Dealer Repair Shops," Monthly Labor Review, September, 1975, pp. 45-46. 4Rothe, Harold F., "Output Rates Among Chocolate Dippers," Journal Of Applied Psychology, April, 1951, p.96, and see earlier publications cited there. 5i.b.i.d., p. 94. 6i.b.i.d. 7Tiffin, J., Industrial Psychology, (New York: Prentice Hall, 1942). 8Kilbridge, Maurice D., "Statistical Indicators of the Continuing Effectiveness of Wage Incentive Applications", Journal of Industrial Economics, 1960, pp. 83-97. 9Weinstein, Alan G. and Robert L. Holzbach, Jr., "Impact of Individual Differences, Reward Distribution, and Task Structure on Productivity in a Simulated Work Environment", Journal of Applied Psychology, 58, Dec. 1973, 10Farr, James L., "Incentive Schedules, Productivity, and Satisfaction in Work Groups: A Laboratory Study", Organizational Behavior and Human Performance, 17, p. 159, 1976. 11Ringelmann, quoted by Maital, Shlomo, Minds, Markets, and Money (New York: Basic Books, 1982), p. 126. 12Latane, 1979, quoted by Maital, Shlomo, Minds, Markets, and Money (New York: Basic Books, 1982, pp. 126-127. 13Locke, E. A., D. B. Feren, V. M. McCaleb, K. N. Shaw, and A. T. Denny, "The Relative Effectivness of Four Methods of Motivating Employee Performance" , in K. D. Dunction, M. M. Gruneberg, and D. Wallis, Changes in Working Life (New York: John Wiley, 1980). 14Lupton, Tom, and Sheila Cunnison, "The Cash Reward for an Hour's Work Under Three Piecework Incentive Systems," Manchester School of Economics and Social Studies, 1957, p.269. 17Simon, Julian L., The Economics of Population Growth (Princeton: Princeton University Press, l977), p. 60. 20Smith, James P., "Assets, Savings, and Labor Supply," Economic Inquiry, XV (4), l977, p. 568. 21Hill, 1971, p. 384. 26Filer, 1981, quoted by Maital, Shlomo, Minds, Markets, and Money (New York: Basic Books, 1982), p. 94. 27i.b.i.d., p. l22. 28Ringelmann, quoted by Maital, Shlomo, Minds, Markets, and Money (New York: Basic Books, 1982), p. l26. 29Latane, l979, quoted by Maital, Shlomo, Minds, Markets, and Money (New York: Basic Books, 1982), pp. l26-7. 30Gilboy, Elizabeth W., "Demand Curves by Personal Estimate," Quarterly Journal of Economics, Vol. 46, 1931, pp. 376-384; Schaeffer, l959; Simon, Julian L., (article in JPE, 1965).