«I Was Only Nineteen, 45 Years Ago: What Can we Learn from Australia’s Conscription Lotteries? Associate Professor Peter Siminski School of ...»
Economics Working Paper Series 2012
I Was Only Nineteen, 45 Years Ago: What Can we Learn from
Australia’s Conscription Lotteries?
Associate Professor Peter Siminski
School of Economics
University of Wollongong
Professor Simon Ville
School of Economics
University of Wollongong
WP 12 - 06
I Was Only Nineteen, 45 Years Ago:
What Can we Learn from Australia’s Conscription Lotteries? 1 Peter Siminski and Simon Ville University of Wollongong April 2012 Abstract The Australian conscription lotteries of 1965-1972 are a unique and underutilised resource for studying the effects of army service and veterans’ programs. Drawing on many data sources and 25 years of related US literature, we present a comprehensive analysis of this natural experiment, examining indicators of health, personal economic outcomes, family outcomes and educational attainment. We discuss the numerous potential mechanisms involved and the limitations of available data.
Keywords: veterans, conscription, lottery, Australia, natural experiment JEL codes: H55; H56; I38; I12; I21 The title is a reference to the song “I Was Only Nineteen” by Australian band Redgum. We are grateful to Joshua Angrist, Jason Lindo and participants of the 3rd Workshop on the Economics of Health and Wellbeing for useful discussions and comments on an earlier draft. We thank Alison Haynes and Louise Rawlings for excellent research assistance; the Australian Bureau of Statistics, Department of Veterans’ Affairs, Australian Tax Office and Australian Institute of Health & Welfare for access and assistance with de-identified data; the Department of Veterans’ Affairs and the Australian Research Council for grant support (LP100100417 and DE120101642). The views expressed in this paper are the authors’ alone, as are any errors of fact or omission.
1. Introduction The economic and social consequences of war have exercised the minds of policy makers and scholars over many years. World War Two has been a particular focus as the apogee of ‘total war’ where mass mobilisation of universally conscripted armed services and the aerial bombing of civilian communities inflicted massive human and physical destruction, with immense economic and social consequences (Marwick 1974, Milward 1977). The second half of the twentieth century, by contrast, witnessed limited regional conflicts with effects largely limited to battle locations. From the perspective of Western countries, these wars have been fought by regular military personnel, sometimes supplemented by selective conscription. Nevertheless, the economic and social consequences of military service remain pertinent policy-relevant questions. They inform assessments of the costs of modern warfare, an issue of central concern in the context of ongoing conflicts in Afghanistan and Iraq (Bedard and Olivier, 2006; Stiglitz and Bilmes, 2008). They also inform appraisements of the adequacy and design of veterans’ compensation. The Australian Government Department of Veterans’ Affairs (DVA) has an annual research budget of around AU$2.5 million, some of which has been dedicated to assessing the health and mortality effects of military service and subsequent rehabilitation and reintegration (examples of DVA-funded projects include Fett et al. 1984; Harrex et al. 2003; Sim et al. 2003; Sim et al. 2005; Wilson et al. 2005a, 2005b).
Our focus is Australia’s involvement in the Vietnam War, one of the most significant events in Australia’s recent history. When Australia’s initially minimal commitment to the war expanded from 1965, the regular army was supplemented by conscripts selected by a ballot based on date of birth (DOB). Sixteen bi-annual ballots were conducted from 1965 to 1972. The randomness of the conscription ballots, designed to create a sense of fairness, but increasingly regarded as arbitrary and lacking legitimacy as the war progressed, has proved ideal for researching the causal effects of military service. It has allowed us to solve the problem caused by non-random selection, which typically affects observational data analyses. Men accepted into the army are typically much healthier than the general population. There are likely to be many other differences between army servicemen and other men. Some of these differences will be unobserved and hence ordinary least squares (OLS) regression cannot adequately overcome this selection problem. The conscription lotteries, however, solve the selection problem. They are almost identical to randomised controlled trials with imperfect compliance, which can be addressed using instrumental variable techniques.
The intuition of the approach is to compare the outcomes of all men (not just veterans) whose DOBs were ‘balloted in’ to the outcomes of all men whose DOBs were balloted out. Any differences in their outcomes likely reflect the effects of army service or its downstream consequences such as veterans’ compensation.
The United States had a broadly comparable conscription system in a similar era. Over the last 25 years, many US studies have exploited this opportunity to study the effects of military service on health, economic and social outcomes (these include Angrist 1990; Angrist et al. 1996; Angrist et al.
2010; Angrist et al. 2011; Angrist and Chen 2011; Conley and Heerwig 2009, 2011; Dobkin and Shabani 2009; Hearst et al. 1986; Hearst et al. 1991; Lindo and Stoecker 2010). Until recently, Australia’s lotteries have not been utilised in this way. In recent years, several high quality data sources have emerged that facilitate such research. These include the Census of Population and Housing (which collected DOB for the first time in 2006), personal income tax data, and administrative data on mortality and veterans’ compensation. We use these sources to study the effects of army service on a broad range of outcomes, including health, personal economic outcomes, family outcomes and educational attainment. We have previously published studies on mortality (Siminski and Ville 2011) and employment (Siminski forthcoming). Here, our purpose is broader. We aim to provide a comprehensive account of what we has been learned thus far from Australia’s conscription lottery ‘natural experiment’ in the Vietnam War era. The majority of the results we show have not been published previously. 2 We make several contributions to the Australian and international literature. We utilise Australia’s conscription lotteries for identification. Only our own recent studies have done so previously. To our knowledge, this is also the first Australian study to estimate the effects of army service on a broad set of economic and social outcomes using any quantitative method. The institutional contexts of Australia’s conscription lotteries also allow us to separately identify the effects of army service for those who served only in Australia and the effects of service for those who served in Vietnam.
Several other advantages of the Australian conscription lotteries are detailed in Siminski (forthcoming).
The remainder of the paper is structured as follows. The following section describes the selection process for national service, compares the service experience of those who went to Vietnam with those who completed their service in Australia, and examines the range of repatriation benefits available to returning soldiers. Section 3 explains the methodology used to exploit the randomness of the ballots and describes the data used to examine effects. The subsequent sections present The exceptions are results for mortality (Siminski and Ville, 2011), employment and disability compensation (Siminski, forthcoming). Brief discussions of effects on mean income, mean earnings and education are also contained in Siminski (forthcoming).
results for health and disability compensation (Section 4), employment, income and earnings (Section 5), domestic life (Section 6) and educational attainment (Section 7). Section 8 summarises and provides overall conclusions. Comparisons are drawn with the American context and results throughout the paper.
2. Institutional ContextSelection
Conscription into the armed services has been practised widely over a long period of history (Ville and Siminski 2011). In Australia there had been three conscription schemes prior to the Vietnam war era, in 1911-29, 1939-45, and 1951-9. Each of these was developed as a universal obligation for men of a specified age. However, in the last two years of the final scheme (1957-9), there was a shift to selective conscription that reflected reduced manpower needs and the desire to focus on more intensive training of fewer men. The process of selection used, a date of birth ballot, was revived in 1965. The context of its use differed somewhat second time around – the target group was 20 year old males (previously 18 year olds) and, rather than being solely a training scheme, it involved service in the regular army that could include wartime deployment in Vietnam. Bi-annual conscription ballots were conducted continually for eight years up to 1972, representing 16 natural experiments from which to analyse the effects of army service.
The nature of the selection process was largely unchanged throughout this era. Males were required to register for the ballot in January or July in the half-year of their twentieth birthday. Marbles, each marked with a number corresponding in serial order to a date in that six-month period, were placed in a barrel. Successive marbles were then drawn randomly from the barrel until a sufficient number had been selected to correspond with estimates of the army’s manpower needs taking account of the size of the birth cohort and the extent of likely exemptions and deferments.3 Those whose dates of birth were withdrawn from the barrel were ‘balloted-in’, that is, liable to be enlisted in the army. Those balloted-out, the remainder, were permanently exempt from conscription. The probability of being balloted-in varied quite significantly across the Langford (1997) provides an excellent summary of the Vietnam era National Service scheme.
16 ballots, from a peak of 53 per cent in the first 1965 cohort to a nadir of 16 per cent in the second half of 1969. Being balloted-in did not, however, lead inevitably to army service. Far from it: of an estimated 224,706 men with balloted in dates of birth, only 63,735 (28 per cent) served in the army, a proportion that also varied between cohorts (Ville & Siminski 2011). A variety of exemptions and deferments explains why little more than a quarter of those balloted-in served. The main grounds for exemption were the failure of medical, psychological or aptitude tests. Members of religious orders and conscientious objectors to war were also exempt. Married men, volunteers, and serious criminals were offered indefinite deferments, while students, apprentices and trainees were granted temporary deferments for the duration of their course, subject to satisfactory progress (Fett et al: 191Comparisons with the American conscription lottery suggest that Australia’s procedure provided a cleaner natural experiment. The longer lead times of the annual American ballot provided greater scope for post-ballot draft avoidance behaviour. It seems likely that the use of a progressive system of selection through ‘Random Sequence Numbers’ may have influenced behaviour (either motivated by draft avoidance or in anticipation of being drafted) amongst men with dates of birth that fell in the uncertain middling section of the sequence of numbers. Finally, while dates of birth drawn remained unpublished until 1997 for all but the last five Australian lotteries, the American ballots were televised live. In the US context, this may have enabled discrimination against draft eligible men who had avoided conscription (especially by potential employers).
The basis on which National Servicemen were selected to serve in Vietnam is complicated, but was far from random (Fett et al., 1984). Units in the Infantry, Artillery, Engineers and Armour were most needed in Vietnam so a serviceman’s allocation to a particular corps was a critical factor. Corp allocation took into account skills, further psychological aptitudes and personal preference, amongst many other corp-specific factors. It has been suggested that many who expressed unwillingness to serve in Vietnam were able to avoid doing so (Fett et al., 1984).