As firms adjust to changes in the economic environment, new jobs are born and old jobs die.
One of the most enduring “factoids” about the American economy is that most new jobs are created by small firms, which are often perceived to be the sole engine of economic growth.
The Small Business Administration, for example, claims that “the term, ‘Great American
47 Mark Montgomery, “On the Determinants of Employer Demand for Part-Time Workers,” Review of Economics and Statistics 70 (February 1988): 112–117; and Ronald G. Ehrenberg, Pamela Rosenberg, and Jeanne Li, “Part-Time Employment in the United States,” in Robert Hart, editor, Employment, Unemployment and Labor Utilization, Boston: Unwin Hyman, 1988, pp. 256–281.
48 Richard Disney and Erica M. Szyszczak, “Protective Legislation and Part-Time Employment in Britain,” British Journal of Industrial Relations 22 (March 1984): 78–100.
Many European countries experienced large unem- ployment rates in the 1990s. In France, the unemploy- ment rate remained above 10 percent for much of that decade. In Germany, the unemployment rate has hov- ered around 9 percent since 1994. The persistence of this unemployment gave rise to the theory that unem- ployment can be reduced by sharing the available work among the many potential workers. In other words, more jobs would be created if the government man- dated a reduction in the standard number of straight- time hours that the typical worker could work.
Several countries adopted this theory and reduced the length of the standard workweek. In 2000, for example, the French government mandated a reduction in the workweek from 39 to 35 hours. In Germany, the unions negotiated sizable reductions on an industry-by- industry basis. In the metalworking and printing sectors, for instance, the standard workweek fell from 40 to 36 hours between 1984 and 1994.
The concept of work-sharing can have an important unintended consequence—and may actually further reduce the demand for labor—because it ignores the fundamentals of economic theory. A reduction in the standard workweek imposes yet more constraints on the firm’s decision of whether to hire an additional worker.
After all, an employer who planned to use the new
workers for a 40-hour workweek at the straight-time wage will now have to pay an overtime premium after 35 hours. In effect, the reduction in the standard work- week may actually increase the average wage associated with hiring a new worker. As a result, employers who find it optimal to staff their factories with workers hired for a 40-hour shift find that the mandated reduction in the workweek increases the cost of hiring an additional worker. As a result, the employer will cut down on the number of hours demanded per worker and on the number of workers hired.
This is precisely what happened in Germany. The reduction in the standard workweek reduced the aver- age number of hours worked weekly and increased the average wage rate—but total employment declined. Put differently, work-sharing held weekly income relatively constant for those lucky workers who remained at a job, but increased the number of persons without a job.
Sources: Jennifer Hunt, “Has Work-Sharing Worked in Germany?” Quarterly Journal of Economics 114 (February 1999): 117–148; see also Bruno Crépon and Francis Kramarz,
“Employed 40 Hours or Not Employed 39: Lessons from the 1982 Mandatory Reduction of the Workweek,” Journal of Politi- cal Economy 110 (December 2002): 1355–1389; and Phillippe Askenazy, “A Primer on the 35-Hour in France, 1997–2007,”
IZA Discussion Paper No. 3402, 2008.
Theory at Work
WORK-SHARING IN GERMANY
Job Machine,’ is appropriately applied to American small business,” and President Clinton’s 1993 State of the Union Address asserted that “because small business has created such a high percentage of all the new jobs in our nation over the last 10 or 15 years, our plan includes the boldest targeted incentives for small business in history.” 49
As our analysis of adjustment costs suggests, small firms would have an advantage in creating jobs if they could respond to favorable changes in the marketplace much faster than bigger firms (that is, if small firms face lower adjustment costs when creating new jobs). It also might be that small businesses have carved out a niche in the fastest-growing areas of the economy, or that the law of diminishing returns prevents large firms from expanding and hiring more workers.
A number of studies of the U.S. manufacturing sector conclusively show that a great deal of job creation and job destruction is going on at the same time. For example, in a typical year, nearly 11.3 percent of manufacturing jobs disappear, whereas nearly 9.2 percent of manufacturing jobs are newly created. 50 The annual net loss of jobs in the manufacturing sector was on the order of 2 percent.
The research also indicates that small firms are not the engines of employment growth that they are widely believed to be (at least in the manufacturing sector). Instead, large firms account for most newly created and newly destroyed manufacturing jobs. 51 In fact, firms with at least 500 workers account for 53 percent of all new jobs created and 56 percent of all jobs destroyed. Moreover, newly created jobs tend to last longer if they are created in larger firms. In particular, the probability that a newly created job still exists after one year is 76 percent for large firms and 65 percent for small firms. This is not surprising because large firms tend to be more stable; they create jobs that have a higher probability of surviving.
Despite the popular mythology, therefore, it seems that large firms account for most of the new jobs in the U.S. manufacturing sector, and they create jobs that tend to be longer lasting.
3-12 Rosie the Riveter as an Instrumental Variable
A great deal of the state-of-the-art research done by labor economists involves trying to estimate labor demand and labor supply curves for particular groups. The findings reached by these studies are often used to predict how particular labor market shocks or policy changes will alter earnings and employment opportunities for workers and firms.
The typical effort to estimate a labor demand curve starts by observing data on employ- ment and wages in a particular labor market—for example, the employment and wages of women. Figure 3-23 shows how the observed employment and wage data can be generated
49 These quotes are drawn from Steven J. Davis, John Haltiwanger, and Scott Schuh, “Small Business and Job Creation: Dissecting the Myth and Reassessing the Facts,” Business Economics 29 (July 1994):
13–21. See also David Neumark, Brandon Wall, and Junfu Zhang, “Do Small Businesses Create More Jobs? New Evidence for the United States from the National Establishment Time Series,” Review of Economics and Statistics 93 (February 2011): 16–29.
50 Steven J. Davis, John Haltiwanger, and Scott Schuh, Job Creation and Destruction, Cambridge, MA:
MIT Press, 1996. See also Christian Belzil, “Job Creation and Job Destruction, Worker Reallocation, and Wages,” Journal of Labor Economics 18 (April 1985): 183–203; and Simon Burgess, Julia Lane, and David Stevens, “Job Flows, Worker Flows, and Churning,” Journal of Labor Economics 18 (July 2000): 473–502.
51 Davis, Haltiwanger, and Schuh, “Small Business and Job Creation: Dissecting the Myth and Reassessing the Facts.”
by our theory. Initially, the labor market is in equilibrium at point P, yielding wage w 0 and employment E 0 . Suppose that the supply curve of women shifts to the right. The new equilib- rium would be at point Q, yielding wage w 1 and employment E 1 . The data we would observe consist of the pair of wages ( w 0 and w 1 ) and the pair of employment statistics ( E 0 and E 1 ).
Figure 3-23 shows that these data can be used to essentially trace out (or identify ) the labor demand curve. In other words, if we could observe a real-world situation where the only curve that shifted was the supply curve, the resulting data on wages and employment would allow us to estimate the labor demand elasticity.
Naturally, in most real-world situations, both the supply curve and the demand curve are shifting at the same time. When both curves shift, the new equilibrium would be at a point like R, with wage w 2 and employment E 2 . The data we now observe consist of the pair of wages ( w 0 and w 2 ) and the pair of employment statistics ( E 0 and E 2 ). These data would allow us to trace out the curve ZZ in the figure, a curve that provides no information what- soever about either the labor supply elasticity or the labor demand elasticity. When the two curves are moving at the same time, therefore, the resulting data on wages and employment do not help us identify the underlying structure of the labor market. Put differently, the resulting data (that is, the line ZZ ) could not be used to predict how a particular policy shift (for example, an increase in the demand for high-tech workers by NASA) would affect wages and employment in the high-tech sector.
FIGURE 3-23 Shifts in Labor Supply and Labor Demand Curves Generate the Observed Data on Wages and Employment
The market is initially in equilibrium at point P, and we observe wage w0 and employment E0. If only the supply curve shifts, we can observe w1 and E1, and the available data would then allow us to trace out the labor demand curve.
However, if both the supply and demand curves shift, we then observe w2 and E2, and the available data trace out the curve ZZ, which does not provide any information about the shape of the underlying labor demand curve.
Dollars
Employment E0
D0
w0
w2
w1
Q R Z P
Z D1
S1
S0
E1 E2
The “trick” for estimating the labor demand elasticity, therefore, is to find a situation where some underlying factor is shifting the supply curve but is leaving the demand curve fixed.
In an econometric framework, we call a variable that shifts one of the curves and not the other an instrument or an instrumental variable. The availability of an instrument for supply lets us then use the method of instrumental variables to estimate the labor demand elasticity. 52 A recent study provides a simple (and instructive) illustration of how particular his- torical events generate instruments that can be used to estimate the labor demand curve. 53 Nearly 16 million men were mobilized to serve in the Armed Forces during World War II, and around 73 percent of them were sent overseas. This shrinking in the number of male workers drew many women into the civilian labor force for the first time, giving rise to the stereotype of Rosie the Riveter, a woman who aided the war effort by performing “men’s work.” In 1940, only 28 percent of women over the age of 15 participated in the labor force. By 1945, the female participation rate was over 34 percent. Although many of these women left the labor force after the war, nearly half of them stayed, permanently increasing the number of working women by 1950 above what it would have been. 54
To understand how the method of instrumental variables can be used in this context to estimate the labor demand curve for female labor, it is important to get a better sense of the historical circumstances. In October 1940, the Selective Service Act began a mandatory national draft registration for all men aged 21–35. By 1947, when the draft finally ended, six separate registrations had been mandated, eventually requiring all men aged 18–64 to register. After each of these registrations, the local draft boards used lotteries to determine the order in which registrants were called to active duty.
The local draft boards were authorized to grant draft deferments to particular groups of men. These deferments were typically based on a man’s marital and parental status and on whether he had skills that were essential to civilian production. Farmers, for instance, were typically deferred because food was obviously needed to support the war effort. Because of these deferments, men living in farm states were substantially less likely to be drafted than men living in more urban states like New York or Massachusetts. In addition, because most military units were segregated during the war, relatively few blacks were drafted, and the geographic distribution of the black population created even more geographic differences in mobilization rates. Table 3-4 reports the mobilization rate for the various states, defined as the proportion of registered men aged 18–44 who served in the military between 1940 and 1945. The interstate variation is substantial. The rate was 41 percent in Georgia, 50 percent in California, and 55 percent in Massachusetts.
The mobilization rate provides the instrument that shifts the supply curve of female labor differently in different states. After all, Rosie would be more likely to become a riveter in those states where draft boards sent a larger fraction of men into active duty. As Figure 3-24 a shows,
52 Analogously, an instrument that shifted only the demand curve would allow us to estimate the labor supply elasticity.
53 Daron Acemoglu, David H. Autor, and David Lyle, “Women, War and Wages: The Effect of Female Labor Supply on the Wage Structure at Midcentury,” Journal of Political Economy 112 (June 2004): 497–551.
54 Claudia Goldin, “The Role of World War II in the Rise of Women’s Work,” American Economic Review 81 (September 1991): 741–756.
there is a strong positive correlation between the 1939–49 growth in female employment and the state’s mobilization rate. The regression line (with standard errors in parentheses) is
Percent change in female employment =
-94.56 + 2.62 Mobilization rate (3-23) (31.88) (0.67)
This regression equation implies that a 1 point increase in the mobilization rate increased female labor supply by 2.62 percent.
It also turns out that the interstate differences in mobilization rates are strongly correlated with the wage growth experienced by female workers. Figure 3-24 b shows a strong negative TABLE 3-4 Mobilization Rate of Men and Changes in Female Wages and Employment, 1939–1949
Source: Daron Acemoglu, David H. Autor, and David Lyle, “Women, War and Wages: The Effect of Female Labor Supply on the Wage Structure at Midcentury,”
Journal of Political Economy 112 (June 2004): 497–551. The mobilization rate gives the proportion of men aged 18–44 who served in the military between 1940 and 1945; the percent change in female employment gives the log change in the total number of nonfarm weeks worked by women aged 14–64; the percent change in the female wage gives the (deflated) change in the log weekly wage of women employed full time multiplied by 100.
State
Mobilization Rate
(%)
Change in Female Employment
(%)
Change in Female
Wage (%)
Alabama 43.6 20.3 81.0
Arkansas 43.6 19.2 79.5
Arizona 49.4 70.2 38.4
California 50.0 65.7 31.3
Colorado 49.7 54.5 50.2
Connecticut 49.4 27.9 34.5
Delaware 46.9 39.4 24.6
Florida 47.7 35.2 69.9
Georgia 41.2 16.7 65.2
Idaho 49.8 53.3 58.1
Illinois 47.6 26.2 42.0
Indiana 45.3 31.6 48.3
Iowa 45.3 2.9 51.2
Kansas 49.0 18.8 55.6
Kentucky 45.2 15.1 51.1
Louisiana 43.5 19.5 69.4
Maine 50.3 19.1 38.4
Maryland 46.9 22.1 48.9
Massachusetts 54.5 24.8 26.9
Michigan 45.3 39.1 48.6
Minnesota 46.8 23.9 47.5
Mississippi 43.7 2.2 73.0
Missouri 45.5 13.2 48.2
Montana 49.4 10.1 44.2
Nebraska 46.3 30.4 49.0
New Hampshire 53.0 20.1 41.8
State
Mobilization Rate
(%)
Change in Female Employment
(%)
Change in Female
Wage (%)
New Jersey 49.7 24.3 35.7
New Mexico 47.8 51.1 50.6
New York 48.4 24.9 33.7
North
Carolina 42.1 23.3 51.6
North
Dakota 41.8 –12.5 51.8
Ohio 47.8 32.4 41.1
Oklahoma 49.0 25.9 55.1
Oregon 53.1 66.5 42.3
Pennsylvania 52.6 31.9 37.9
Rhode Island 54.1 27.8 28.6
South
Carolina 42.7 31.1 80.0
South
Dakota 42.2 6.5 52.5
Tennessee 44.9 19.5 52.4
Texas 46.0 48.5 66.8
Utah 52.8 56.9 35.3
Vermont 47.3 21.9 62.6
Virginia 44.7 34.5 56.1
Washington 52.4 72.8 39.2
West
Virginia 48.4 27.3 47.5
Wisconsin 43.3 27.3 44.4
Wyoming 48.9 36.2 39.6
FIGURE 3-24 The Impact of Wartime Mobilization of Men on Female Labor Supply and Wages
% Change in Employment, 1939–49
0
−20 40 20 40 60 80
45 50 55
Mobilization Rate
Regression line has slope +2.62
% Change in Weekly Wage, 1939–49
50
20 40 40 30 60 70 80 90
45 50 55
Mobilization Rate Regression line has slope −2.58
(b) Mobilization Rate and Changes in Female Wages, by State (a) Mobilization Rate and Changes in Female Employment, by State
relation between the 1939–49 percent change in the female wage and the mobilization rate.
In other words, female wages grew least in those states where a larger proportion of men went off to war. In fact, the regression line relating these two variables is
Percent change in female wage =
171.69 - 2.58 Mobilization rate (3-24)
(21.45) (0.45)
The slope coefficient of this regression line indicates that a 1 point increase in the mobili- zation rate is associated with a 2.58 percent drop in the female wage.
The regression models reported in equations (3-23) and (3-24) can now be used to esti- mate the labor demand elasticity. The data tell us that for every 1 point increase in the male mobilization rate, female employment increased by 2.62 percent and female wages fell by 2.58 percent. Put differently, a historical event that reduced the female wage by 2.58 percent was associated with a 2.62 percent increase in female employment. Therefore, the labor demand elasticity is given by the ratio of these two numbers, or
=
Percent change in female employment
Percent change in female wage = 2.62
-2.58 = -1.02 (3-25) The historical experience of female wages and employment during World War II thus sug- gests that the labor demand elasticity for women is around 1.0.
The methodological approach summarized visually in Figures 3-24 a and 3-24 b can be expanded to control for other factors that might shift the labor supply or labor demand curves differently in different states, such as the educational attainment and age distribu- tion of female workers. Although this multivariate approach cannot be easily illustrated, the method of instrumental variables relies on the same basic logic: the availability of an instrument that shifts only the labor supply curve allows us to use the resulting data on wages and employment to trace out the labor demand curve.
The discussion also illuminates the main weakness of the instrumental variable approach:
The legitimacy of the entire exercise hinges on finding a valid instrument, a variable that shifts only one of the curves in the supply-demand framework. A great deal of the disagreement over the interpretation of many empirical results in labor economics often hinges on whether the researcher is using a valid instrument that allows her to trace out or identify either the labor supply or the labor demand curve. The ratio in equation (3-25) is a labor demand elasticity only if interstate differences in the mobilization rate generated interstate differences in female labor supply but did not generate interstate differences in female labor demand. As we have seen, the labor demand curve is given by the value of marginal product curve. The mobilization rate would then be a valid instrument only if it is uncorrelated with both interstate differences in the price level and interstate differences in female productivity.
Summary
• In the short run, a profit-maximizing firm hires workers up to the point where the wage
equals the value of marginal product of labor.
• In the long run, a profit-maximizing firm hires each input up to the point where the
price of the input equals the value of marginal product of the input. This condition im- plies that the optimal input mix is one in which the ratio of marginal products of labor and capital equals the ratio of input prices.
• In the long run, a decrease in the wage generates both substitution and scale effects.
Both of these effects spur the firm to hire more workers.
• Both the short-run and long-run demand curves for labor are downward sloping, but the
long-run demand curve is more elastic than the short-run curve.
• The short-run labor demand elasticity may be on the order of 0.4 to 0.5. The long-run
elasticity is on the order of 1.
• Capital and skilled workers are complements in the sense that an increase in the price of
capital reduces the demand for skilled workers. Capital and unskilled workers are substi- tutes in the sense that an increase in the price of capital increases the demand for unskilled workers.
• The imposition of a minimum wage on a competitive labor market creates unemployment
because some workers are displaced from their jobs and because new workers enter the labor market hoping to find one of the high-paying (but scarce) jobs.
• The elasticity of teenage employment with respect to the minimum wage is on the order
of 0.1 to 0.3.
• The presence of variable adjustment costs implies that firms adjust their employment
slowly when the wage changes. If fixed adjustment costs are important, employment changes in the firm are large and sudden, if they occur at all.
• An instrument is a variable that shifts either the supply or demand curve. The variation
caused by this shock can be used to estimate the labor demand or labor supply elasticity.
adjustment costs, 127 average product, 87 capital-skill
complementarity hypothesis, 114 cross-elasticity of
factor demand, 112 demand curve for
labor, 90
Key Concepts
elasticity of labor demand, 91 elasticity of
substitution, 106 equilibrium, 114 instrument, 135
instrumental variable, 135 isocost, 96
isoquant, 94
law of diminishing returns, 87 marginal cost, 92 marginal product of
capital, 85 marginal product of
labor, 85 marginal productivity
condition, 92