This is a followup to my previous article The Mathematics of the Ph.D. Glut. To recap, yes there is a Ph.D. glut in nearly all STEM (Science, Technology, Engineering, and Mathematics) fields in the United States. As a matter of federal government policy, many more Ph.D.’s are produced than can be employed as professors or other kinds of professional researchers. The Ph.D. glut in biology and medicine is especially bad currently as discussed in Brian Vastag’s July 7, 2012 Washington Post article U.S. pushes for more scientists, but the jobs aren’t there. There has been a Ph.D. glut in most STEM fields including specific fields such as mathematics and physics where specific shortages are often either claimed or strongly implied since about 1970.
The policies that have resulted in a perpetual Ph.D. glut since about 1970 are frequently justified by explicit or implicit claims that more Ph.D.’s (or other kinds of STEM workers) will translate into more scientific and technological progress and more “growth,” a popular political mantra. Has the Ph.D. glut in biology and medicine cured cancer? Not so far — after forty years and about $200 billion in inflation adjusted dollars. Many other specific examples of lack of scientific and technological progress may be cited. In fact, the Ph.D. glut is associated with a decline in real growth rates and a slowing of scientific and technological progress in most fields other than some areas of computers and electronics.
What does the evidence show? Remarkably, both the growth rate of the US Real Gross Domestic Product (GDP) and the growth rate of the per capita US Real Gross Domestic Product were significantly higher prior to 1970 than since. Now, the decline in the US growth rate is a worrisome long term trend. It is difficult to tie to any one event or policy. It has occurred under both Republican and Democratic Presidents and Congresses. The decline has occurred despite and perhaps because of the adoption of several policies such as the overproduction of Ph.D.’s and financial deregulation that are routinely and uncritically justified as producing increased “growth.”
I look at overall economic growth for an important reason. Research and development is risky and unpredictable. For any given research field, one can argue that the problem, e.g. cancer, has proved much more difficult than expected. That may be true. Luck unquestionably plays a big role. This is economist Paul Krugman’s explanation for the general lack of progress over the last forty years: where is my flying car?
But is it all just bad luck? By looking at the total economic growth rate we can, at least partially, average out the idiosyncracies of different fields. The Ph.D. glut is a universal problem in nearly all research fields not just high profile fields like physics, mathematics, and molecular biology.
The plots below show the United States Real GDP and Real GDP per capita since 1947. The GNU Octave script and the raw data used to make the plots is provided in the appendices. The raw data is from the St. Louis Federal Reserve/Bureau of Economic Analysis (BEA) and the U.S. Census Bureau.
The analysis shows:
MEDIAN REAL GDP GROWTH RATE 1947-1970 ans = 0.040710 MEDIAN REAL GDP GROWTH RATE 1971-2011 ans = 0.027481 MEDIAN US REAL GDP PER CAPITA GROWTH 1947-1970 ans = 0.026985 MEDIAN US REAL GDP PER CAPITA GROWTH 1971-2011 ans = 0.017927
The median real GDP growth rate from 1947 to 1970 was 4.1 percent (rounded to two significant digits), as opposed to 2.7 percent from 1971 to 2011. The median real GDP per capita growth rate from 1947 to 1970 was 2.7 percent, as opposed to 1.8 percent from 1971 to 2011. The median is used to avoid the effects of outliers which can make the average or mean misleading. There is, for example, a probable outlier in the GDP data in about 1950, a growth rate of about 12 percent.
The data shows an accelerating downturn in growth rates over the last two decades. This coincides with the recent rise in many real energy prices such as gasoline. Of course, correlation does not prove causation. A number of things have risen sharply in the last two decades including the Internet, general computer use, cell phones, consumption of aspartame (the sweetener in Diet Coke), consumption of high fructose corn syrup, and diagnoses of autism (see the previous article The Mathematics of Autism), for example.
It is probable that the single biggest proximate cause of the disappointing growth over the last forty years has been limited and disappointing progress in power and propulsion technology. The plot below is from the United States Energy Information Administration and shows the real, inflation-adjusted price of a gallon of gasoline over the last forty or so years. These prices show a general reversal of the previous trend of declining real prices of gasoline during the early twentieth century (1900-1970).
The rising real price of gasoline presumably reflects that the production of gasoline and other competing energy sources has not kept up with rising global demand. Keep in mind that most demand comes from the so-called developed world: the United States, Europe, Japan, and a few other nations. It would require something like a four-fold increase in global energy production to raise the standard of living of the entire human race to US or European levels.
Why Have More Ph.D.’s Produced Less Progress and Growth?
The policy of overproduction of Ph.D.’s is based on a number of assumptions that are rarely stated or discussed. Remarkably, it is quite possible that policy makers, business leaders, and others have never thought through what they are doing and why. It is doubtful that aging policy makers, senior scientists and others would consciously sabotage attempts to find cures or effective treatments for cancer or other diseases of old age, although that is what they may well have done with the current glut of biology and medicine Ph.D.s. Similarly, very few policy makers, senior scientists, or others, except perhaps a few energy industry moguls, benefit from the lack of progress in power and propulsion technology, especially if we run out of oil, natural gas, and other hydrocarbon fuels without finding a replacement: the Peak Oil scenario.
In general, science — and with it mathematics — public policy is based on a crude physical analogy. Scientists, at least the graduate students and post-doctoral researchers, are envisioned as essentially intellectual ditch diggers. Intelligence is envisioned as a single linear scale rather like Pearson’s general intelligence and equated to the physical strength of the ditch diggers: a simple “mental horspower.” If you want more, better results, hire more and stronger intellectual ditch diggers. Perhaps, there are a few super ditch diggers who are ten times stronger than the average ditch digger. It would be nice to hire them but you would rather not pay them ten times as much.
Somewhat related is a belief that control over the intellectual ditch diggers is a good thing. The more control, the more likely you will get better results. Thus, slave labor from India and China is expected to produce better science and technology than free labor from the United States. This latter view seems especially suspect. Even in physical labor, were the free factory and farm workers of the North who could quit their jobs in disgust if mistreated really less productive than the slave labor of the antebellum South? Not only did the North forge ahead into the industrial era leaving the South far behind, but England and France became dependent on imports of food from the North, not the South, so when the Civil War came, England and France ultimately sided with the North despite “King Cotton” and the textile industry lobbies.
The Ph.D. glut combined with the heavy importation of guest workers from India, China, and other Third World nations who often face serious economic hardship if fired and sent home — a common fate — creates a situation approaching slave labor. Can slave labor really cure cancer or invent new practical energy sources? Don’t hold your breath!
The logical implication of this implicit model of scientific research is simple: throw a lot of money and manpower under centralized bureaucratic control at a problem like cancer and it will fall. There have been many attempts to do this since World War II, inspired in part by the spectacular success of the Manhattan Project (see the previous article The Manhattan Project Considered as a Fluke). Like the War on Cancer, most of these efforts have
However, if scientific and mathematical research and development is not analogous to this concept of digging ditches, one might expect the Ph.D. glut to cause serious problems. The Ph.D. glut depresses wages and working conditions in critical research areas such as health and energy. The purported “best and brightest” in the United States go elsewhere — developing pseudo-scientific mathematical models for Wall Street with disastrous consequences or, somewhat better, developing apps to sell pet food over iPhones and similar gimmicks with limited but at least positive benefits. Increased quantity cannot make up for the loss of quality.
Most importantly, the Ph.D. glut greatly reduces the independence and creative freedom that has so often proven necessary to solve extremely hard scientific problems like cancer or new energy sources. The Ph.D. glut means that egos and politics dominate. Original thinkers can easily be eliminated and compliant yes men rewarded with the few permanent positions. Near slave labor from Third World nations will only further reduce this independence and creative freedom.
As the data shows, the Ph.D. glut is associated with a long term decline in growth rates in the United States. While quantitative data are limited, it is associated with a qualitative decline in the rate of scientific and technological proress in many fields, especially power and propulsion technologies. There is the notable exception of some computer and eletronic technologies, although artificial intelligence (AI) actually shows a similar disappointing rate of progress — unlike CPU clock speeds or video compression algorithms where progress clearly has been impressive and comparable to the historical pre-1970 levels in many other technical fields.
While it is difficult to
© 2012 John F. McGowan
About the Author
John F. McGowan, Ph.D. solves problems using mathematics and mathematical software, including developing video compression and speech recognition technologies. He has extensive experience developing software in C, C++, Visual Basic, Mathematica, MATLAB, and many other programming languages. He is probably best known for his AVI Overview, an Internet FAQ (Frequently Asked Questions) on the Microsoft AVI (Audio Video Interleave) file format. He has worked as a contractor at NASA Ames Research Center involved in the research and development of image and video processing algorithms and technology. He has published articles on the origin and evolution of life, the exploration of Mars (anticipating the discovery of methane on Mars), and cheap access to space. He has a Ph.D. in physics from the University of Illinois at Urbana-Champaign and a B.S. in physics from the California Institute of Technology (Caltech). He can be reached at [email protected].
GNU Octave script usgdp_per_capita.m to generate the plots in the article.
% script to compute annual growth rate of US REAL GDP % % (C) 2012 By John F. McGowan, Ph.D. % %data = dlmread('us_real_gdp.txt'); % federal reserve data data = dlmread('us_real_gdp.txt'); % federal reserve data year = data(:,1); gdp = data(:,2); figure(1) h1 = plot(year, gdp); set(h1, 'linewidth', 3); title("US REAL GDP (CHAINED 2005 DOLLARS)"); xlabel('YEAR'); ylabel('BILLION DOLLARS'); pop1947 = dlmread('us_pop_1947_2012.txt'); len = length(year); gdp_per_capita = gdp*1e9 ./ pop1947(1:len, 2); figure(2) h2 = plot(year, gdp_per_capita); set(h2, 'linewidth', 3); title("US REAL GDP PER CAPITA (CHAINED 2005 DOLLARS)"); xlabel('YEAR'); ylabel('DOLLARS'); delta = conv(gdp, [1 -1]); growth = delta(1:end-1) ./ gdp(1:end); [p, s] = polyfit(year(2:end), growth(2:end)*100, 3); fit = polyval(p, year(2:end)); target = ones(size(fit))*6.8; % need average growth rate of 6.8% to absorb new Ph.D.'s figure(3) h3 = plot(year(2:end), growth(2:end)*100, '+', year(2:end), fit, '-r', year(2:end), target, '-g'); set(h3, 'linewidth', 3); title("US GDP ANNUAL REAL GROWTH RATE"); xlabel('YEAR'); ylabel('PERCENT'); legend('DATA', 'SMOOTHED', 'PHD GROWTH RATE'); disp("MEDIAN REAL GDP GROWTH RATE 1947-1970"); median(growth(2:23)) disp("MEDIAN REAL GDP GROWTH RATE 1971-2011"); median(growth(24:end)) delta_pop = conv(pop1947(:,2), [1 -1]); growth_pop = delta_pop(1:end-1) ./ pop1947(:,2); [ppop, spop] = polyfit(year(2:end), growth_pop(2:end-1)*100.0, 3); fit_pop = polyval(ppop, year(2:end)); figure(4) h4 = plot(year(2:end), growth_pop(2:len)*100, '+', year(2:end), fit_pop, '-r'); set(h4, 'linewidth', 3); title("US POPULATION GROWTH RATE"); xlabel('YEAR'); ylabel('PERCENT'); legend('DATA', 'SMOOTHED'); % growth rate of per capita gdp delta_gdp = conv(gdp_per_capita, [1 -1]); growth_gdp = delta_gdp(1:end-1) ./ gdp_per_capita; [p_gdp, s_gdp] = polyfit(year(2:end), growth_gdp(2:len)*100.0, 3); fit_gdp = polyval(p_gdp, year(2:end)); figure(5) h5 = plot(year(2:end), growth_gdp(2:len)*100, '+', year(2:end), fit_gdp, '-r'); set(h5, 'linewidth', 3); title("US REAL GDP PER CAPITA GROWTH RATE"); xlabel('YEAR'); ylabel('PERCENT'); legend('DATA', 'SMOOTHED'); disp('MEDIAN US REAL GDP PER CAPITA GROWTH 1947-1970'); median(growth_gdp(2:23)) disp('MEDIAN US REAL GDP PER CAPITA GROWTH 1971-2011'); median(growth_gdp(24:end)) % real price of gallon of gasoline at EIA Department of Energy % % http://www.eia.gov/forecasts/steo/realprices/ disp('ALL DONE');
US Real Gross Domestic Product (GDP) data (1947-2011) in billions of 2005 “chained” dollars (inflation adjusted) from the St. Louis Federal Reserve: us_real_gdp.txt.
1947 1793.3 1948 1868.2 1949 1838.7 1950 2084.4 1951 2192.2 1952 2305.3 1953 2314.6 1954 2379.1 1955 2535.5 1956 2582.1 1957 2589.1 1958 2654.3 1959 2782.8 1960 2800.2 1961 2975.3 1962 3097.9 1963 3262.2 1964 3429.0 1965 3720.8 1966 3881.2 1967 3977.6 1968 4174.7 1969 4259.6 1970 4253.0 1971 4442.5 1972 4750.5 1973 4948.8 1974 4850.2 1975 4973.3 1976 5187.1 1977 5446.1 1978 5811.3 1979 5884.5 1980 5878.4 1981 5950.0 1982 5866.0 1983 6320.2 1984 6671.6 1985 6950.0 1986 7147.3 1987 7451.7 1988 7727.4 1989 7937.9 1990 7982.0 1991 8062.2 1992 8409.8 1993 8636.4 1994 8995.5 1995 9176.4 1996 9584.3 1997 10000.3 1998 10498.6 1999 11004.8 2000 11325.0 2001 11370.0 2002 11590.6 2003 12038.6 2004 12387.2 2005 12735.6 2006 13038.4 2007 13326.0 2008 12883.5 2009 12873.1 2010 13181.2 2011 13441.0
US Population Data from the St.Louis Federal Reserve (1959-2012) and the US Census Bureau (1947-1958) combined: us_pop_1947_2012.txt.
1947,144126071 1948,146631302 1949,149188130 1950,152271417 1951,154877889 1952,157552740 1953,160184192 1954,163025854 1955,165931202 1956,168903031 1957,171984130 1958,174881904 1959,175818000 1960,179492000 1961,182404000 1962,185347000 1963,188113000 1964,190763000 1965,193308000 1966,195614000 1967,197814000 1968,199864000 1969,201821000 1970,203929000 1971,206567000 1972,208989000 1973,211053000 1974,213003000 1975,214998000 1976,217172000 1977,219262000 1978,221553000 1979,223973000 1980,226554000 1981,229004000 1982,231235000 1983,233398000 1984,235456000 1985,237535000 1986,239713000 1987,241857000 1988,244056000 1989,246301000 1990,248743000 1991,252012000 1992,255331000 1993,258799000 1994,262021000 1995,265157000 1996,268258000 1997,271472000 1998,274732000 1999,277891000 2000,281083000 2001,283960000 2002,286739000 2003,289412000 2004,292046000 2005,294768000 2006,297526000 2007,300398000 2008,303280000 2009,306035000 2010,308706000 2011,311019000 2012,313278000