Skill-biased technological change and the demographic composition of the US military, 1979-2008

Andrea Asoni, Andrea Gilli, Mauro Gilli, Tino Sanandaji 19 September 2021



The demographic composition of the US military is a topic of central policy and scholarly relevance. It touches upon debates such as how socioeconomic inequality affects who joins the military and hence bears the burden of American defence. 

According to the literature, the US ‘all-volunteer force’ is profoundly unfair, as it imposes the cost of American wars on the less affluent segments of the society and relieves the wealthy from contributing to US national security, an important public good (Kriner and Shen 2010). If that is true, the implications for the US military would be dramatic.

American military superiority rests as much on its advanced technology as well as on its qualified, proficient, and highly trained personnel (Biddle 2004, Biddle and Long 2004). Yet, existing debates have paid little attention to how, over the past 40 years, technological change – together with doctrinal, organisational, and tactical transformations – has affected the personnel demand from the US Air Force, Army, Marines, and Navy.

As we show in a forthcoming article (Asoni et al. forthcoming), while existing perceptions of unfairness were empirically accurate in the past, they no longer are.

Existing perceptions

The perception that the US military predominantly recruits individuals from the most disadvantaged socioeconomic backgrounds, who have little other career opportunities, has become widely accepted over the years, both among the public and academics. After the abolition of the military draft in 1973, some started to refer to the all-volunteer military as a ‘poverty draft’, as ‘economic conscription’, or as a ‘military of mercenaries’ (e.g. Binking 1982, Henderson 2005, MacLean and Parsons 2010). 

Echoing these concerns, in the late 1980s, the Democratic Leadership Council argued that the US cannot “ask the poor and under-privileged alone to defend us while our more fortunate sons and daughters take a free ride, forging ahead with their education and careers” (1988). This view became so entrenched that 15 years later, Noam Chomsky spoke about a “mercenary army of the poor” (2003) while a Congressman referred to the Iraq War as a “death tax ... on the poor” (Rangel 2004). Along the same lines, the New York Review of Books speculated in 2008 that “the military has set its sights on an especially vulnerable population” (Massing 2008) while, according to The Guardian, “these days, the US military is more like a sanctuary for racists, gang members and the chronically unfit” (Kennard 2012).

Problems of existing academic research

While intuitive at face value, the accepted wisdom that the US military predominantly recruits individuals from the most disadvantaged socioeconomic backgrounds and with little career opportunities suffers from two problems. 

Many recent works rely on aggregate data by zip code (e.g. Simon and Warner 2007, Kriner and Shen 2010, Dean 2018). Hence these works do not permit us to derive any inference about individual socioeconomic backgrounds (ecological fallacy). In fact, using the same data, some have reached the very opposite conclusions (Kane 2005, Lien 2012). 

Other studies do rely on individual data, but these data are from the past (e.g. Cooper 1977, Fredland and Little 1982, Boulanger 1981, Fernandez 1989, Geachman 1993, Lutz 2008). As such, they are no longer informative about the present because of technologic, tactical, operational, and doctrinal changes that have affected the US military (Binkin 1986, Macgregor 2003, Mahnken 2008, Tuck 2014).

We address these problems by relying on individual-level data from the National Longitudinal Survey of Youth 1979 and 1997 (NLSY79 and NLSY97), with two longitudinal nationally representative samples of individuals who were 14 to 17 when the survey started (see Asoni et al. forthcoming for details). With these data, we can address the ecological fallacy of previous works and, comparing older with more recent data, can also assess how the demographic composition of the US armed forces has changed over time.

Alternative hypotheses: Deskilling vs skill-biased technological change

To understand how the US military has evolved over the past 40 years, we investigate two sets of alternative hypotheses.

The first set tests the conventional wisdom that those from poorer backgrounds and with fewer career opportunities are overrepresented in the US military and that such inequality in serving has worsened over the past decades. These arguments are based, implicitly or explicitly, on the opportunity cost and the deskilling hypotheses. 

The opportunity cost hypothesis implies that individuals from more disadvantaged socioeconomic backgrounds lack the skills to compete in the job market, and hence they have more to gain and less to lose from joining the military, despite the risk for their life (e.g. Cohen 2001, Vasquez 2005, Bacevich 2008, Kriner and Shen 2010). 

The deskilling hypothesis implies that increasing reliance on advanced technology has relieved the military from employing highly skilled personnel, thus allowing for the recruitment of less-talented individuals (e.g. Toomepuu 1986; and Caselli 1999 for a general discussion).

The second set of hypotheses suggests that the US military has become more equal over the past decades. Our argument draws in particular from the skill-biased technological change and nurture hypotheses. We hypothesise that the change in technology, tactics, operations, and doctrines observed over the past decades has called for the recruitment of more skilled individuals (e.g. Card and DiNardo 2002, Autor et al. 2003). 

Further, drawing from the literature in child development, education, epigenetics, and neuroscience, we hypothesise that the most disadvantaged socioeconomic groups are less likely to meet the requirements of the highly capital-intensive US military (Brooks-Gunn and Duncan 1997, Heckman 2011, Heckman et al. 2013, Noble et al. 2015).

Empirical evidence

We compare different demographic indicators for the 1997 and the 1979 cohorts and observe a marked change in the demographic characteristics of those who join the military (‘veterans’) relative to those who do not (‘civilians’). Table 1 summarises our findings. Those who joined the military in the 1997 cohort have a higher family1 median and average income, higher family median wealth (for males), and higher cognitive skills than civilians, whereas in the 1979 cohort we observe the very opposite: civilians outperformed veterans along most dimensions.

Table 1 General characteristics of veterans and civilians, by gender (2017 US$)

Notes: Family income and wealth are reported in 2017 US dollars. Family wealth is not available in NLSY79. Cognitive skills are measured through an AFQT-like score calculated by the BLS. BLS provides for each individual its cognitive skills percentile across the entire population. Table reports the average AFQT percentile for each demographic group. Education is measured as the average number of years of education by 2014.

To summarise our findings, we report the non-parametric estimates of the probability of joining the military along several dimensions. Figure 2 shows the probability of joining the military against family income, separately for the 1979 cohort and the 1997 cohort. 

Figure 1 Probability of joining the military vs family income (2017 US$)

For the 1979 cohort (red line), the probability of joining the military is clearly higher for those with lower-than-average family income. However, for the 1997 cohort, the probability is much more evenly distributed, and it follows an inverse-U shape with the lowest and highest incomes less likely than the median and average ones. 

Precisely, the group most likely to join had a family income between $17,000 and $27,000 in the 1979 sample and between $69,000 and $107,000 in the 1997 sample (all figures are expressed in 2017 dollars).

We carried out the same analysis for scores in the AFQT test. This is a test of cognitive skills, taken by those in the National Longitudinal Survey of Youth samples, analogous to a similar test conducted by the military for recruitment purposes. Figure 2 reports our results. While both 1979 and 1997 samples show the same reversed-U-shape pattern, the 1997 line is translated to the right of the 1979 line, indicating that over time the US military has become more selective in its screening for cognitive skills.

Figure 2 Probability of enlisting vs. cognitive skills

Table 2 below confirms this finding: for the 1979 cohort, individuals in the 2nd, 3rd, and 4th quintile of the cognitive-skills distribution are all equally likely to join the military; for the 1997 cohort, individuals in the 4th quintile are significantly more likely to join.

Table 2 Share of people who join the military by quintile of income, wealth, and skill

We test the robustness of these findings with a more rigorous statistical analysis as well as look at different sub-samples to address alternative explanations. Our results are robust to restricting the analysis to the US Army and US Marines, to enlistees only (i.e. excluding officers), to those who joined before the 9/11 attack, and to those who joined before the Great Financial Crisis (for details, see Asoni et al. forthcoming).


We analysed existing perceptions about the demographic composition of the US military. While in the past, American armed forces predominantly recruited individuals from lower socioeconomic backgrounds and more limited career opportunities, since the late 1990s, this has no longer been the case. 

Skill-biased technological change (together with resulting changes in doctrine, operations, and tactics) account for the increase in the quality of military personnel we observe. To our knowledge, this is one of the first analyses that apply the literature on skill-biased technological change to military recruitment, and thus it shows the promise of applying the literature in economics or of investigating economic dynamics to understand military affairs (e.g. Harrison 2019, O’Brien 2019).

Our analysis also has important implications for the US military and more generally for American policymakers. First and foremost, our results support the urgency of so-called early interventions in child development, to put them on the right path so they are not screened out by potential employers when they grow up (Heckman 2011). 

Second, our findings raise the question as to whether growing economic inequality might affect patterns of recruitment in the future, and what the US should do to anticipate such a development.

Authors' note: The views expressed here are the authors' own and do not represent those of NATO or the NATO Defence College, CRA or any CRA employee.


Asoni, A, A Gilli, M Gilli and T Sanandaji (forthcoming), “A mercenary army of the poor? Technological change and the demographic composition of the post-9/11 US military” Journal of Strategic Studies.

Autor, D H, F Levy and R J Murnane (2003), “The skill content of recent technological change: An empirical exploration”, Quarterly Journal of Economics 118(4): 1279–333.

Bacevich, A J (2008), The limits of power: The end of American exceptionalism, New York: Metropolitan Books/Henry Holt & Company.

Biddle, S, and S Long (2004), “Democracy and military effectiveness: A deeper look”, Journal of Conflict Resolution 48(4): 525–46.

Biddle, S (2004), Military power: Explaining victory and defeat in modern combat, Princeton, NJ: Princeton University Press.

Binkin, M (1986), Military technology and defense manpower, Washington, DC: The Brookings Institution.

Brooks-Gunn, J, and G J Duncan (1997), “The effects of poverty on children”, Future of Children 7(2): 55–71.

Boulanger, G (1981), “Who goes to war?” in A Egendorf et al. (eds.), Legacies of Vietnam: Comparative adjustment of veterans and their peers Vol. 4, Washington, DC: US Government Printing Office, 494–515.

Card, D, and J E DiNardo (2002), “Skill-biased technological change and rising wage inequality: Some problems and puzzles”, Journal of Labor Economics 20(4): 733–83.

Caselli, F (1999), “Technological revolutions”, American Economic Review 89(1): 78–102.

Chomsky, N (2003), Understanding power: The indispensable Chomsky, London: Vintage Books.

Cohen, E A (2001), “Twilight of the citizen-soldier”, Parameters 31: 23–28.

Cooper, R N (1977), Military manpower and the all-volunteer force, Santa Monica, CA: RAND Corporation.

Dean, A (2018), “NAFTA”s army: Free trade and US military enlistment”, International Studies Quarterly 62(4): 845–85.

Fernandez, R L (1989), Social representation in the US military, Washington, DC: Congressional Budget Office.

Fredland, E J, and R D Little (1982), Socioeconomic characteristics of the all volunteer force: Evidence from the National Longitudinal Survey, 1979, Annapolis, MD: US Naval Academy.

Harrison, M (2019), “Economic warfare: Insights from Mançur Olson”,, 14 November.

Heckman, J (2011), “The economics of inequality: The value of early childhood education”, American Educator 35(11): 31–47.

Heckman, J, R Pinto and P Savelyev (2013), “Understanding the mechanisms through which an influential early childhood programme boosted adult outcomes”, American Economic Review 103(16): 2052–86.

Henderson, D R (2005), “The role of economists in ending the draft,” Economic Journal Watch 2(2): 362–76.

Kane, T (2005), “Who bears the burden? Demographic characteristics of US military recruits before and after 9/11”, Center for Data Analysis Report, Washington, DC: Heritage Foundation, 5–8.

Kennard, M (2012, “The modern US army: Unfit for service?”, The Guardian, 31 August.

Lien, D S (2012), An investigation of FY10 and FY11 enlisted accessions’ socioeconomic characteristics, Arlington, VA: Center for Naval Analyses.

Lutz, A (2008), “Who joins the military? A look at race, class, and immigration status”, Journal of Political and Military Sociology 36(2): 167–88.

Macgregor, D A (2003), Transformation under fire: Revolutionizing how America fights, Westport, CT: Praeger.

MacLean, A, and N L Parsons (2010), “Unequal risk: Combat occupation in the volunteer military”, Sociological Perspectives 43(3): 347–72.

Mahnken, T G (2008), Technology and the American way of war since 1945, New York: Columbia University Press.

Massing, M (2008), “The volunteer army: Who fights and why?”, New York Review of Books, 3 April.

O’Brien, P P (2019), “How the war was won,”, 3 September.

Noble, K G, S M Houston, N H Brito, H Bartsch, E Kan, J M Kuperman, N Akshoomoff et al. (2015), “Family income, parental education and brain structure in children and adolescents”, Nature Neuroscience 18: 773–78.

Rangel, C (2004), Representative Rangel remarks at the National Press Club, Washington, DC (15 April 2004), cited in Edward Epstein, “Draft unlikely soon but hot topic now”, San Francisco Gate, 2 May.

Simon, C J, and J T Warner (2007), “Managing the all-volunteer force in a time of war”, Economics of Peace and Security Journal 2(1): 20–29.

Tuck, C (2014), Understanding land warfare, London: Routledge.

Toomepuu, J (1986), Costs and benefits of quality soldiers: A critical review of the CBO report, “Quality soldiers: Costs of manning the active army”, Fort Sheridan IL: US Army Recruiting Command.

Vasquez, J P III (2005), “Shouldering the soldiering: Democracy, conscription, and military casualties”, Journal of Conflict Resolution 49(6).


Family income and wealth are those of the potential recruits’ families of origin, measured at a time when most individuals in the sample were still living with their parents. Therefore, our analysis does not suffer from a ‘reverse causality’ problem, i.e. does not capture the effect of joining the military on income and wealth.



Topics:  Labour markets Poverty and income inequality

Tags:  deskilling effect, Inequality, military, skill biased technological change, skill-biased technical change, US

Principal, Charles River Associates

Senior Researcher, NATO Defence College

Senior Researcher, Swiss Federal Institute of Technology in Zurich (ETH-Zurich)

Researcher, Institute for Economic and Business History Research, Stockholm School of Economics


CEPR Policy Research