Forecasting Demand and Supply of Doctoral Scientists and Engineers: Report of a Workshop on Methodology (2000)

Key questions for understanding markets for doctoral scientists and engineers concern the possibilities for substitution and how long it takes for substitution to occur in the presence of shortages or surpluses. Substitution is the process by which supply and demand adjust. Employers may revise job descriptions, reassign tasks, reorganize laboratory operations, and modify training programs. Employees may switch activities or fields and retrain to qualify in new areas. For example, when Ph.D. engineers are plentiful, they may be shifted to administration or sales or be asked to work with less equipment or technical support. When they are scarce, technicians may be asked to take on some of their tasks, and may receive additional training to do this work. Substitution occurs in response to shortages or surpluses or to changes in wages that result from the process of bidding away jobs or workers when the two are not in balance. When substitution occurs slowly, another margin of adjustment will be worker quality. Facing a shortage, employers may accept less talented or qualified employees. If the quality of scientists declines when substitution occurs, the probability of scientific breakthroughs may decrease and development time, errors that require correction and product recall, or the need for quantity of managerial oversight may increase.

Substitution is important on both the supply and demand sides of markets for doctoral scientists and engineers. On the demand side, for example, if employers find it difficult to hire qualified Ph.D.s and the wages of Ph.D.s rise relative to the cost of software, employers may substitute capital or software for labor.

They can also move research operations overseas or encourage foreign scientists to immigrate if enough U.S. scientists cannot be found. On the supply side, students can change fields. Employed scientists with a fundamental skill like mathematics can use that skill working for employers of different descriptions (e.g., in recent years, physicists who were unable to find appropriate jobs in physics have found jobs in the financial services industry).

An understanding of the time period over which substitution can occur is important to the formation of policy designed to encourage or discourage production of new doctoral scientists and engineers. If undergraduates change majors rapidly in order to move into the hottest fields, and if doctoral students can easily change their area of research concentration when new areas emerge, shortages are likely to be short lived. Conversely, if students lack information about what fields are opening up, if significant “retooling” to change fields is required, or if institutional factors inhibit changes of wages, substitution may take a long time to occur, and the result will be a relatively inelastic, short-run, labor supply curve.

The workshop paper by Sherwin Rosen and Jaewoo Ryoo addressed some of these issues in the engineering labor market. Their model consisted of four equations:

1. A demand equation presented as an inverse function, in which the wage an engineering graduate receives depends on the stock of engineers at graduation (the end of the production period— four years for a B.S. in Engineering) and on variables that reflect shifts in demand.

2. A supply equation for new entrants to engineering, in which the number of new entrants depends on the discounted present value of future wages that entrants expect, on variables that reflect the relative attractiveness of other professions, and on the number of new entrants a year earlier.

3. A demographic identity equation, in which the current stock of workers depends on the stock of workers a year earlier less attrition, plus new entrants in the current year.

4. An expected career-prospects equation, in which entry decisions depend on discounted expected future earnings in engineering.

In solving the model, the authors (Ryoo and Rosen, 1998) found that:

[E]ntry into school of any cohort is negatively related to the stock of practitioners they expect to encounter upon entry at graduation. For example, if many students are currently enrolled in engineering schools, then entry of freshmen in the current period is deterred, ceteris paribus. Of course enrollments are encouraged by greater expected future demand conditions and discouraged by greater expected career prospects in alternative occupations.

It should be noted, however, that changes in direction are usually much harder to forecast than longer-term trends. The relative usefulness of such an expectations-based model will depend on the relative accuracy of expectations.

Using data from the Current Population Survey, Engineering Manpower Commission, and NSF, ¹ Ryoo and Rosen estimated the parameters for the system of equations in the model after normalizing the variables (e.g., the number of engineering graduates relative to college graduates, investment in R&D per unit of GDP, and so forth). They found that supply is very responsive to changes in

demand. Demand is considerably less responsive to changes in R&D expenditures relative to GDP. The authors tried different ways of modeling expectations. Prospective entrants can base their expectations on future wages (assuming that actual future values are the same as estimated future values), on past wages, on both past and estimated future wages, and only on current wages (leading to cobweb behavior). The authors concluded that expectations appear to be forward looking, and this implies that supply behavior will be less volatile than a simple myopic (cobweb) model might suggest.

In the discussion that followed, the Ryoo-Rosen model was commended for explicitly modeling wages and expectations about wages. There may be greater problems, though, in applying such a model to markets for Ph.D.s. Lags are much longer, and there are opportunities for substitution not just between engineering and other professions, but among those who have earned different levels of degrees. Cycles are long, and what may matter most to prospective enrollees are changes in demand. When will demand begin to decline? Lags in the data are such that these turning points are identified in the data only years after they have occurred. A better, but probably unachievable, goal would be to construct models that would predict turning points with some accuracy. Such findings could then be publicized and affect student decisions. In this dimension, science and engineering students may not be much different from majors in art, theater, music, and literature.

Other responses to the model noted that immigration is an important alternative supply source for engineering markets, but one that does not appear explicitly in the model. Further, as Rosen pointed out, much of the demand for engineers comes from the industrial sector and the bulk of demand is for people with baccalaureate degrees. Demand for doctoral scientists is generated to a much greater degree by the educational sector. In the current system, research funding generates a demand for both Ph.D.s and for doctoral and postdoctoral students to staff research projects. In the life sciences, in particular, there may be a long lag between the

career decision and undertaking the career itself. Forecasting lifetime earnings, or even earnings seven years into a career, may be far more problematic in these fields than in the market for baccalaureate engineers. An increase in federal spending for grants may create a demand for graduate students that is divorced from market demand for Ph.D.s.²

The effect of wage expectations on the supply of Ph.D.s should also consider the uncertainty of the career wage profile, including uncertainties about an individual's ability to attain breakthroughs (discoveries, patents, prizes, tenure, etc.). Any systematic misperception of these probabilities by students, owing to overestimation of their ability relative to others, may lead to unrealistic supply decisions even in the face of evidence that median wages will be low.

The role postdoctoral study plays in the adjustment of the Ph.D. labor market also needs to be examined. There are substantial differences by field in the utilization of postdoctorals. For example, there are very few postdoctorals in the social sciences, whereas in the biological sciences, more than 80 percent of new Ph.D.s go on to postdoctoral study. A period of postdoctoral study postpones entry into the “regular” labor market, and evidence exists that the length of this delay is increasing (NRC, 1998). For new Ph.D.s, a postdoctoral appointment delays entry into the labor market, thus disrupting supply and shortening the length of careers in regular employment. Also, the postdoctoral pool provides a ready supply of highly trained Ph.D.s that is available for regular employment if academic or industrial demand increases, thus shortening the period of adjustment until such time as the postdoctoral pool is drained. Models of supply that take into account only enrollments and graduates will miss an important margin of adjustment if they ignore postdoctorals in those fields where the numbers of these positions are large or growing.

Finally, it was noted that possibilities for substitution may vary by employment sector. Scientists and engineers in industry may be assigned to a variety of projects that blur distinctions based on field. By contrast, in academia a faculty member rarely changes departments and is hardly ever hired in a field outside the field of training.