BACKGROUND ON THE DRAFT AND DRAFT LOTTERY

As of the end of 1965, a system of conscription was nominally in place, but the armed forces were almost entirely manned by volunteers. That changed with the escalation of the Vietnam War in 1966, just as the high school class of 1965 was reaching the age of draft eligibility (19 years).Footnote ¹ For the next 3 years, call-up, exemption, and deferral decisions were made by local draft boards loosely operating under federal guidelines. As Baskir and Strauss (Reference Baskir and Strauss1978, 24) put it, “the four thousand draft boards developed four thousand very different policies.” Deferments of many forms were, in principle, available, the most common of which involved the existence of dependents, especially children, and the college student deferment, which required satisfactory academic performance and progress toward a degree. Older men within the 19- to 26-year-old age range were called up before younger, reducing the vulnerability to the draft faced by the class of 1965 in the war's early years.

With the expansion of the draft from 1966 to 1968 came the draft resistance movement and the growth of dissatisfaction with draft procedures among those not opposed to the draft itself. Blue ribbon commissions were set up to study Selective Service reform (e.g., Marshall Reference Marshall1967). Draft policies were tweaked through Executive Orders and new legislation. In 1967, for example, the deferment for graduate study was eliminated. The idea of a national draft lottery gained salience. The abolition of college student deferments was debated, in part as a response to the perceived low socioeconomic status (SES), nonwhite bias of the draft. At the same time, public opinion remained largely against the draft resisters and, until mid-1968, ambivalent about or in favor of the war.

Soon after taking office, in March 1969, President Nixon sent to Congress his plan for reform of the draft, which called for a national lottery, a continuation of the college student deferment, the creation of a 1-year window of maximum draft vulnerability, and a shift to prioritizing younger over older men within the 19- to 26-year-old range—with the important proviso that those with a deferment would have their year of maximum vulnerability begin whenever their deferment ended, if it ever did.Footnote ² By the end of the year, the new system was in place. On December 1, 1969, on national television, lottery numbers (officially called “random sequence numbers” or RSNs) were assigned to birth dates by picking birth date−stamped capsules from an urn and numbering them, sequentially, from 1 and 366. The resulting numbers were to apply to potential draftees born between 1944 and 1950, which of course included the high school class of 1965.Footnote ³ Starting with the number 1, assigned to September 14, draft-eligible men would be called up to the extent required to fulfill military need.

Thus, as 1969 came to a close, nonexempted members of the class of 1965 faced a new draft regime, where their vulnerability to the draft was largely dictated by their draft number unless they could obtain and maintain a deferment. For those who had already taken advantage of the student deferment, the time in that refuge was running out. Those newly seeking out the college refuge would find it short lived because student deferments were ended in 1971. Even sooner to disappear were the “fatherhood” deferments, which President Nixon abolished by Executive Order in early 1970.Footnote ⁴ Lowering the stakes somewhat was the fact that forces were being gradually withdrawn from Vietnam and draft rates were on the decline. Yet, new draftees were increasingly sent to combat duty and casualty rates were high.

Over the ensuing 3 years, there was a good deal of uncertainty about just how high in the 1-to-366 sequence the draft call would go. The widely reported and repeated claim was that numbers in the lower third (1–122) were likely to be called, those in the top third (245–366) were unlikely to be called, and those in the middle third either would or would not be called depending on how the war progressed. Because induction requirements remained decentralized and local draft boards continued to oversee draft status classifications and call-ups, the lottery numbers actually called varied substantially across states and localities. In the first few postlottery months, some local boards went very high in calling up men for induction, something that Director of the Selective Service Curtis Tarr (Reference Tarr1981, 38) ruled as “contrary to the intent of the law” and that prompted the setting of national ceilings. The ceiling was set at 115 in April 1970, 145 in May, 170 in June and July, 195 for the rest of the 1970, and 125 for 1971 and 1972.Footnote ⁵

POLITICAL PSYCHOLOGY OF THE DRAFT LOTTERY THREAT

The primary pathway by which an adverse draft number could induce change attitudes toward the war is self-interest. Although self-interest effects have been notoriously elusive in public opinion research, the consensus is that strong self-interest effects are most likely when what is at stake is “1. visible, 2. tangible, 3. large, and 4. certain” (Citrin and Green Reference Citrin and Green1990, 18; see also Chong, Citrin, and Conley Reference Chong, Citrin and Conley2001; Green and Gerkin Reference Green and Gerkin1989; Sears and Funk Reference Sears, Funk and Mansbridge1990). Those with low draft numbers were facing a situation that would meet these four criteria handsomely—a (relatively) high likelihood of being forced to abandon all personal plans and undertakings and to take part in a potentially life-threatening war. As one's lottery number increased, one's vulnerability decreased. The potency of the self-interest motive would have been enhanced by the fact that the risk of losses, rather than of gains, was at issue (Cacioppo and Gardner Reference Cacioppo and Gardner1999; Mercer Reference Mercer2005).

Self-interest effects could have played out two ways. Those with lower lottery numbers were more likely to have been drafted or to have enlisted expecting call-up, and thus to have directly paid the costs of serving in the war (while nevertheless living to provide survey responses post war). In this case, the reasoning would be that low lottery number holders, having been dragged into military service, ended up more hostile to the Vietnam War than those who, holding high lottery numbers, escaped military service altogether. We consider the evidence for this mediating path in a later section.

However, simply facing the risk of being drafted, even if that possibility did not actually materialize, would have imposed direct and, in many cases, large costs on draft-eligible men. Studies focused on the draft and draft resistance have documented the psychological, material, and opportunity costs young men faced as they attempted to elude the draft (e.g., Baskir and Strauss Reference Baskir and Strauss1978; Foley Reference Foley2003). Stories from those made vulnerable to conscription due to the lottery likewise testify as to its disruptive consequences.Footnote ⁶ Low lottery number holders were required to report for preinduction physical examinations. Anxiety and fear were commonplace.Footnote ⁷ Employment opportunities for low lottery number holders were limited (Frank Reference Frank2007; Tarr Reference Tarr1981, 37). Future plans had to be put on hold, revised, or abandoned altogether (Mann and Dashiell Reference Mann and Dashiell1975). Attempts to elude the draft took time, energy, and financial resources. Thus, even if a low draft number did not actually result in military service, the negative personal consequences following from an unlucky lottery draw should have fueled opposition to the draft and to the war.

There is also an indirect path through which these lottery-induced circumstances may have come to shape political views. Feelings of fear and anxiety direct attention to the threatening stimulus and prompt learning (e.g., Cacioppo and Gardner Reference Cacioppo and Gardner1999; Marcus, Neuman, and MacKuen Reference Marcus, Neuman and MacKuen2000). Similarly, having a direct personal stake in an issue heightens the attention it receives (Krosnick Reference Krosnick1990; Lau, Brown, and Sears Reference Lau, Brown and Sears1978). Thus, the lottery should have prompted those who were most vulnerable to being drafted to pay greater attention to the war and the politics surrounding it. And, what anyone would learn when paying attention to the war in the postlottery period was that casualties were mounting, the war was going badly, and the majority of the public and many political elites had turned against it (Hallin Reference Hallin1984; Mueller Reference Mueller1973; Schuman Reference Schuman1972). Thus, greater opposition to the war among low lottery number holders could have emerged as an indirect consequence of information seeking sparked by vulnerability to the draft.

Evidence that draft lottery status affected attitudes toward the Vietnam War would in one sense speak to a void in the literature and in another sense challenge the received wisdom. Studies of opinions toward the Vietnam War have rarely considered draft status.Footnote ⁸ The only studies that looked explicitly at draft number and attitudes employed small convenience samples of college students holding deferments in the immediate postlottery period. Longino (Reference Longino1973) found more negative attitudes toward the war among those with low lottery numbers, but not to a statistically significant extent (see also Bergen Reference Bergen2009), whereas Apsler (Reference Apsler1972) found more disapproval of the draft. Yet, civilian attitudes toward (or related to) the war in Vietnam have been shown to be at best weakly related to self-interest indicators such as whether one has a family member or friend serving in Vietnam (Lau, Brown, and Sears Reference Lau, Brown and Sears1978; Mueller Reference Mueller1973). Indeed, Lau, Brown, and Sears (Reference Lau, Brown and Sears1978, 479) find “no evidence that the self-interested had distinctly self-serving attitudes toward the war.” As we will see, that is decidedly not the case here.

DATA AND MEASURES

LOTTERY NUMBER AND VIETNAM ATTITUDES

The pivotal question that this article addresses is whether 1969 lottery numbers affected Vietnam attitudes in the 1973 survey. To measure Vietnam War attitudes, we construct a Vietnam Dove–Hawk index using items from the 1973 panel wave. The first component is the standard question of whether the war was a mistake. The measure has three possible scores: yes (Dove), in-between, and no (Hawk). The second is also a 3-point measure, derived from open-ended responses (in 1973) regarding what should have been done. Respondents were first asked:

Those who denied that the government handled the war well were then asked for up to two answers to the following question:

First mentions were coded dovish (codes 20–39) or hawkish (1–19). If the first mention was neither a Dove nor a Hawk response, then we scored their second mention. Those who failed to offer either a Dove response or a Hawk response were coded as in-between. The minority who responded to the first question by agreeing that the government handled the war “as well as it could have” were coded as Hawks.

We averaged the closed-ended “mistake” responses and the manufactured “what should we have done?” responses (each scaled 0–1) to create the Dove–Hawk index. The composite index had five possible responses. The scale ranges from 0 (war was a mistake, should have retreated) to 1 (war justified, no errors, or should have escalated).

Table 1 estimates the effect of lottery number on war attitudes using the Dove–Hawk scale. Four equations are presented, two for our draft-eligible, college-bound respondents, and for comparison, two more for the draft-eligible, non–college bound. For each group, we regress the Dove–Hawk index on lottery number (rescaled to range from 0 = lowest through 1 = highest), both in a bivariate equation and with controls for 1965 (high school) party identification and a four-item index of political attitudes, also from the 1965 survey wave.Footnote ¹⁴ We expect college-bound respondents with safe (high) lottery numbers to be the most hawkish post treatment. Little or no effect is anticipated for the non–college bound, for whom the matter of possible military service had usually been settled by 1969. The note to Table 1 also reports on the results of placebo tests considering lottery number effects among college-bound women. Because women were not eligible for the draft, once again we should find zero lottery number effects.

TABLE 1.

Effect of 1969 Lottery Number on Attitudes toward Vietnam War, 1973

Notes: The dependent variable is the composite Vietnam War attitude index, scaled to run from 0 (Dove) to 1 (Hawk). Lottery number is rescaled from 1 to 366 to 0 to 1. Entries are ordinary least squares (OLS) unstandardized coefficients. Robust standard errors (SEs), which take into account the clustering (by school) in the data, are shown in parentheses (see Nichols and Shaffer Reference Nichols and Shaffer2007). Cases are male respondents who had not served in the military as of 1969. “College bound” are those taking college preparatory courses in 1965. Placebo test results: Coefficients on lottery number for college-bound women are as follows: −0.00, p = 0.97 (bivariate, n = 295) and −0.01, p = 0.88 (multivariate, n = 290).

The results of Table 1 fulfill our expectations. For the college bound, the Dove–Hawk index of Vietnam attitudes shows a positive coefficient that is statistically significant beyond the 0.01 level. For the non–college bound, the coefficients are small, actually negative, and insignificant. Null results are uniformly evident for the placebo tests involving women.

The positive coefficients for the college bound were 0.24 in the bivariate case and 0.22 with controls. The implication is that the difference between holding the lowest and highest lottery number is about 20% to 25% along the Dove–Hawk continuum. Thus, we see a major attitudinal shift lasting as long as 3+ years (from late 1969 to 1973) in attitudes toward the war, with individual fates determined by the luck of the draw.Footnote ¹⁵

The parameter estimates for the control variables are also of interest. The non–college bounds’ 1973 Vietnam attitudes were shaped somewhat by their earlier party identifications, but not by their issue attitudes. For our presumably more sophisticated college bound, the pattern was the reverse. Their Vietnam attitudes were influenced somewhat by their earlier issue attitudes on a left–right scale, but not at all by their earlier party identifications.

Figure 1 presents the raw data for the bivariate relationship between lottery number and Dove–Hawk attitudes among the college bound.Footnote ¹⁶ Visual inspection shows that those with hawkish attitudes tended to be especially concentrated in the safe (high) lottery number ranges. The curve imposed on the graph is the LOWESSFootnote ¹⁷ curve predicting composite Vietnam attitudes from lottery number, based on the bivariate equation. Although there is some inevitable curvature to the estimate, the shape of the LOWESS regression suggests that the relationship between draft number and attitude is decidedly linear rather than a step function.Footnote ¹⁸

FIGURE 1.

LOWESS Curve Displaying War Attitude as a Function of Draft Lottery Number: College Bound Only

Notes: Observations are slightly jittered in order to enhance the visual display. The LOWESS curve predicts the Dove–Hawk score based on lottery number for all observations in the graph, regardless of actual military service. Bandwidth = 0.8.

Causal Process: Via Anticipation or Reality of Military Service?

Given the randomness of draft lottery assignments, we can be quite confident that the statistically significant relationship between lottery number and Vietnam attitudes was causal. But what was the causal mechanism? As described previously, one possibility is that it was a consequence of the lottery's impact on the probability of getting drafted, which proved traumatic and disruptive for low lottery number holders. A plausible rival hypothesis is that the response arose to the actuality rather than the expectation of military service. Consider that even if the lottery draw affected political attitudes initially, the final attitudinal resolution could have depended on whether the subject was actually compelled into military service. If some believed that their lottery number was “safe,” then their continued support for the war could have been conditional on escaping the draft. If others opposed the war because their draft number made them feel vulnerable, then their opposition might have evaporated once they passed their year of vulnerability without getting drafted. Why, we might ask, should initial reaction to the draft number trump the actuality of whether their lives were disrupted by military service?

Clearly, the draft number affected the likelihood of military service, and military service can affect political attitudes. One possibility is that the military service intervening variable works against the vulnerability hypothesis, with getting a low draft number leading to military service, which then caused hawkish views. However, the opposite is also possible—that getting a low number led to military service, which then produced alienation from the military and the war effort. In fact, we would suspect the latter to be more plausible because among those who entered the military in 1969 or later, the lower the lottery number, the more dissatisfaction with their military service,Footnote ¹⁹ and low lottery number has a strong adverse effect on feelings toward the military.Footnote ²⁰ Thus, some of the effects we have observed for lottery number may be due to service itself rather than anticipation of military service.

The obvious first test is to model 1973 Dove–Hawk attitudes as a function of both lottery number and military service. The hypothesized effect of military service would have it predict opposition to the war, which would show up as a negative coefficient. However, as the equation in the first column of Table 2 shows, the estimated military effect is positive, although trivial and insignificant. The “direct” impact of lottery number remains intact at about its original value. The second equation in Table 2 separates military service as either voluntary or involuntary, based on self-reports. Again, there is no military effect, not even from getting drafted. The third and fourth equations analyze the effect of lottery number separately for those who served and those who did not. The lottery coefficient actually increases among those who did not serve and drops for those who served, although the difference is not statistically significant. If real, this difference is readily explained if those who were more acceptant of military service were relatively indifferent to their lottery number fate, whereas those resistant to the idea of being called to duty were more affected than others by their lottery draw. Even if so, what is crucial here is that lottery number effects are strongly evident among those who never ended up serving, where military service as the mediating variable is not a possibility.Footnote ²¹

TABLE 2.

Evaluating Possible Mediating Role of Military Service: College Bound Only

Notes: All results for college-bound (those whose 1965 high school curriculum was college preparatory) males who did not enter military service prior to 1969. Standard errors are clustered standard errors. All variables are scaled 0 to 1.

^a Instrument for Military Service is a 0-to-1 dichotomy, whether the draft number was 196 and above or 195 and below.

We might conclude from this that the lottery effect is all or almost all via the expectation rather than the actuality of military service. However, some caution is necessary. Unlike the random draw of a lottery number, whether one served in the military involves a degree of self-selection. Even though the originating independent variable (lottery number) is randomly assigned, separating out the direct effect from the indirect effect of mediating variables is not as clear-cut as it might appear.Footnote ²² The challenge is that the true effect of military service on attitudes might be biased from a possible reverse effect, whereby respondents who enlisted were more acceptant of the war. A biased estimate of the military effect would also bias the OLS estimate of the direct effect of the draft number. Undoubtedly, enlistments were driven to some degree by political attitudes—for example, enlistees scored higher on our conservatism scale from high school in 1965, although not to a statistically significant degree.

How serious is this possible bias? We conducted a sensitivity analysis to explore the plausible conditions under which serious bias would arise. We ask the following: What is the plausible range of downward bias to our OLS-based estimate of the military effect on war attitudes? And how seriously could this bias upward the presumed direct effect of lottery number?

Answering these questions requires that we evaluate the plausible causal chain from draft number to military service to Vietnam War attitudes. We know that the link from draft number to military service is weak. For our college-bound respondents, the actual impact of draft numbers on the likelihood of actual military service was slim and almost certainly less than our subjects expected at the time of the draft. As noted previously, the difference in the probability of serving for those with numbers at or below and those above the 195 ceiling is only 15 percentage points. Similarly, the differential “effect” on military service of having the lowest versus the highest number both via either a linear OLS model or a probit model is a probability differential of only 0.18. With the actuality of military service so weakly dependent on lottery number, the link from military service to attitude would have to be extremely large in order for the lottery effect to operate mainly via military service.

How large would this military effect have to be? Let us consider the hypothetical case where the direct effect of the lottery number on Vietnam War attitude is exactly half of our 0.24 estimate, with the other half via a direct military effect. With the direct effect cut to 0.12, the product of the lottery number effect on military service and the military service effect on Dove–Hawk attitudes would also be 0.12. For this to occur, the military effect on attitudes would need to be a stunningly negative −0.67, as if getting snared into military service caused our respondents to become more antiwar by moving two thirds of the range on the composite war scale (−0.18 × −0.67 = 0.12).

The unrealistic nature of such a large military effect on Vietnam attitudes is amplified when one considers that the observed correlation between military service and Vietnam attitudes is virtually zero. For the causal accounting to add up to a zero correlation, any negative effect of military service on support for the war would have to be balanced by an equal-sized positive effect of war attitude on military service. A large net effect both ways would require one set of college-bound military men joining the armed services against their will and then turning strongly against the war, whereas another set joins up out of extreme enthusiasm for the war. Any among our sample who wanted to select into the military had ample opportunity to join during the 4 years prior to the lottery. That, combined with the unpopularity of the war from 1969 to 1973, surely limits the magnitude of the hawkishness-driven selection effects on military service one would expect to be operating among recent college graduates exposed to the draft lottery.

Finally, we can leverage the fact that nobody with draft numbers over 195 should have been drafted. This allows a two-stage least squares (TSLS) analysis using as an instrument for military service whether the respondent's draft number was above or below the 195 cutoff. Being above or below the195 cutoff passes the test for plausible instrumental variables. It is not caused by Vietnam attitude and any independent effect on war attitude (controlling for lottery number) would be via military service.

The TSLS equation is shown in the final column of Table 2. Regressing war attitudes on lottery number and military service (proxied by being above or below the cutoff) yields a seemingly “large” but insignificant coefficient in the “wrong” direction, as if being called to military duty made one more prowar.Footnote ²³ Meanwhile, the direct effect of draft number appears as strong as ever, with the coefficient even larger at 0.35. Thus, although the draft number, as a marker for the perceived likelihood of getting drafted in response to the 1969 draft, is related to Vietnam War attitudes, the actuality of getting called (instrumented by whether the draft number is above or below the cutoff) did not generate dovish attitudes.

To summarize, although involuntary military induction conceivably could have by itself turned some young men against the war, this effect could not have been large enough to challenge the thesis that the source of the lottery number effect was the disruption, uncertainty, and anxiety generated by the lottery number itself. If a large military effect had been present, then we would see a different pattern in the data. Instead, we conclude that it was how the lottery transformed one's vulnerability to military service—with the psychological, material, and opportunity costs entailed—that is behind the lottery effects we observe.Footnote ²⁴ That such effects could be observed years later in the 1973 wave of the Political Socialization Survey testifies to the persistence of their impact.

LOTTERY NUMBER AND VOTE CHOICE, POLICY ATTITUDES, IDEOLOGY, AND PARTISANSHIP

We consider next whether the luck of the draw in the 1969 draft lottery had effects that extended beyond attitudes toward the war. When the lottery number affected men's Vietnam attitudes, did the alteration of views extend to related attitudes involving partisan choice, policy direction, political personalities, and vote choice? That is, did those with high numbers who became Hawks also become (for instance) more Republican and conservative, while those with low numbers who became Doves turned more Democratic and liberal?

There are several reasons to expect that lottery numbers also affected the liberalism-conservatism of the respondents and their voting behavior and participation in partisan activities. Dynamic attitudinal constraint could have been at work: As Vietnam attitudes were “experimentally” induced via lottery number, this change could have triggered other attitudes to become more consistent with Vietnam views. Being against the war might have by itself activated cognitions prompting citizens to turn liberal, vote Democratic, or identify with the Democratic Party. The agent of change could also have been social networks, as antiwar sentiments could have pushed people to associate with antiwar crowds and adopt their constellation of leftist attitudes. Changes in life circumstance prompted by a low draft number (e.g., problems finding employment) could also have been the immediate source of attitude change on a variety of fronts. Finally, the lottery number itself could have exerted an effect directly, as those unlucky enough to draw an adverse draft number could have simply blamed the president and his party for their plight, apart from any antiwar sentiment engendered.

To estimate these effects, our dependent variables include vote for president in 1972 (reported in 1973), relative Nixon-McGovern thermometer scores, an index of partisan political activity performed from 1970 to 1973, a three-item ideology index, an eight-item policy issue index (left–right), plus 1973 party identification. Vote was coded 0 for McGovern and 1 for Nixon. The remaining variables ranged continuously between 0 (liberal/Democratic) and 1 (conservative/Republican).Footnote ²⁵

Table 3 shows the results, limited to the college bound. For each dependent variable, we present multivariate versions, adding our two exogenous control variables from the 1965 interviews: the four-item issue index and respondent party identification. The first column shows the probit equation predicting presidential vote choice in 1972 (as recalled in 1973).Footnote ²⁶ Other results are estimated via OLS. With one prominent exception, they all show positive and significant or near significant effects.Footnote ²⁷ Those whose lottery number made them vulnerable to the draft show a broad pattern of attitudinal and behavioral differences as of 1973—they were more likely to have rejected Nixon in the voting booth, to express attitudes that favored McGovern over Nixon, and to have participated in acts that showed that same partisan bias. They were also more likely to align themselves with the liberal rather than the conservative end of the ideological continuum, and tended to express more liberal attitudes on a wide array of issues.

TABLE 3.

Multivariate Analysis of 1972 Vote Choice, Presidential Candidate Evaluations, and Issue Attitudes: College Bound Only

Notes: Lottery number is rescaled from 1 to 366 to 0 to 1. The dependent variables are scaled to run from 0 (liberal/Democratic) to 1 (conservative/Republican). The vote choice equation was estimated with probit. Shown for that dependent variable is the estimated change in the probability of a Nixon vote if the X in question changed from 0 to 1, holding the other two Xs at their means. The p value is from the test on the probit coefficient. Probit coefficients and robust-clustered standard errors (SEs) for the three predictors, in turn, are 1.00 (0.35), 0.79 (0.23), and 1.11 (0.43). Entries shown for the other dependent variables are unstandardized coefficients from ordinary least squares (OLS), with robust-clustered SEs in parentheses. Cases are college-bound (those whose 1965 high school curriculum was college preparatory) male respondents who had not served in the military as of 1969. The probit pseudo R ² is the McKelvay-Zavoina estimate of the proportion of the variance in the latent variable that is explained. Placebo test results: Coefficients on lottery number among college-bound women are vote: 0.02, p = 0.834; candidate ratings: −0.03, p = 0.370; partisan activity: 0.00, p = 0.988; issue index: −0.01, p = 0.851; ideology: −0.05, p = 0.231; and party ID: −0.02, p = 0.790. All dependent variables are measured in the 1973 panel wave.

For no dependent variable was the effect clearer than for reported vote in 1972. The probit equation predicting the vote reveals an average effect in terms of the probable vote of 38 percentage points as the differential from the lowest to the highest lottery number. Holding the other variables at their means, the projected percent voting for Nixon is 37% with lottery number 1 and 75% with number 366.

The exception is party identification. Although the party identification coefficient is positive (indicating those with safe numbers were more Republican), it does not achieve statistical significance.Footnote ²⁸ We consider this exception in further depth later in the article.

Inclusion of the control variables allows us to compare the lottery number effects with the coefficients on prior predispositions. Perhaps the most interesting finding in this respect is the importance of issue attitudes from the high school years. An index based on attitudes toward school prayer, racial segregation, and the United Nations, plus tolerance of communists and atheists, dominates party identification as a predictor of the attitudes held 8 years later. In fact, for all variables except 1973 party identification, the respondent's 1965 party identification is statistically dominated by lottery score.Footnote ²⁹ It is worth mulling this remarkable fact about these 25- and 26-year-old men in 1973, who had a college preparatory high school education. With their exposure to the 1969 draft and with an early adulthood spent during the turmoil of the Vietnam War years, their lottery number was a stronger influence on their political outlook than their late-childhood party identification.

DRAFT LOTTERY EFFECT AS A GENERAL POLITICAL SHOCK

So far, we have dealt with the draft number effect as a directional effect whereby getting a high or a low number induces specific changes in attitude or behavior. In addition, the draft lottery could have affected respondents by shocking the attitude structure, prompting the development of political views at odds with positions expressed in the past. After all, the young men subject to the draft were going through their early twenties, a life stage during which preadult political orientations are often revised in light of the new circumstances and experiences encountered in early adulthood (Sears Reference Sears, Wheeler and Shaver1983). This way of thinking focuses attention on how postlottery attitudes compare to those expressed before, with the expectation that the attitudes of those with low lottery numbers would have especially been transformed. Those with safe numbers could go about their business without rethinking their politics, whereas this may not have been true for those with low numbers who were confronted with potential life disruptions (or worse).

From such considerations, the general prediction is that getting a low (adverse) draft number leads to a lesser continuity of attitudes from the 1965 to the 1973 panel. Table 4 shows this expectation to be correct. It displays the correlation between 1965 party identification and each of the 1973 postdraft variables, calculated separately for those with draft numbers in the lowest third (most vulnerable) and for those with numbers in the top third (safest).

TABLE 4.

Correlation between 1965 Party Identification and 1973 Political Attitudes by Lottery Number: College Bound Only

Notes: Correlations were based on pairwise deletion of missing data. Cases are college-bound (those whose 1965 high school curriculum was college preparatory) male respondents who had not served in the military as of 1969. Ns ranged from 57 to 75 for the low lottery number group and from 60 to 84 for the high lottery number group.

*p < 0.05; **p < 0.01; ***p < 0.001.

Except for attitudes toward the war, prior partisanship is much more predictive of the 1973 attitudes of those holding safe lottery numbers than it is for those who were vulnerable to being drafted. In fact, among the vulnerable group, 1965 party ID is essentially uncorrelated with each of the 1973 variables except party ID. Meanwhile, those with relatively safe numbers show at least a modest degree of continuity in their 1965 to 1973 correlations. Thus, for example, the correlation between preadult party identification and the vote one cast in 1972 is −0.03 among those with adverse lottery numbers but 0.43 among those with safe ones. The most important difference is for party identification predicting itself. The over-time correlation for party identification (1965 vs. 1973) for the vulnerable group is less than half the size of that found for the safe group (0.27, p < .05 vs. 0.56, p < 0.001).

We see, then, that a low draft number had the consequence of upending prior party identification. The higher the draft number, the more party identification in 1973 resembled party identification from high school in 1965. As a secondary consequence, other attitudes in 1973 also became less tied to 1965 partisanship. Another way to see this is presented in Table 5. Here, we regress our 1973 variables (including party identification) on the 1965 party identification, lottery number, and interaction term representing the product of lottery number and 1965 party identification.

TABLE 5.

Interaction of Lottery Number and Prior Party Identification: College Bound Only

Notes: Lottery number is rescaled from 1 to 366 to 0 to 1. The dependent variables are scaled to run from 0 (liberal/Democratic) to 1 (conservative/Republican). The vote choice equation was estimated with probit. Shown for that dependent variable is the estimated change in the probability of a Nixon vote if the X in question changed from 0 to 1, holding the other two Xs at their means. The p value is from the test on the probit coefficient. Probit coefficients and robust-clustered standard errors (SEs) for the three predictors, in turn, are −0.33 (0.60), −0.71 (0.61), and 3.21 (1.18). Entries shown for the other dependent variables are unstandardized coefficients from ordinary least squares (OLS), with robust-clustered SEs in parentheses. Cases are college-bound (those whose 1965 high school curriculum was college preparatory) male respondents who had not served in the military as of 1969. The probit pseudo R ² is the McKelvay-Zavoina estimate of the proportion of the variance in the latent variable that is explained.

As seen, the higher (safer) the lottery number, the more are 1973 attitudes affected by 1965 party identification. This pattern is seen in all of the results, although the interaction term is insignificant for war attitudes and partisan activity. Most tellingly, the additive coefficients for 1965 party identification tend toward estimates of zero effects. The straightforward interpretation is that, among the must vulnerable (zero on the interaction variable), 1965 party identification had virtually no relationship with future attitudes. The comparison for the least vulnerable (interaction term equals 1) is the addition of the party identification coefficient and the interaction coefficient. For 1973 party identification, the implied effect of lagged party identification from 1965 is a mere 0.13 with the lowest lottery number but 0.56 with the highest number (0.13 + 0.43).Footnote ³⁰

The draft lottery spurred the most draft-vulnerable men to rethink their party identification, which resulted in new affiliations weakly tied to those expressed in the past. Consideration of the predicted values shows that much of the movement was from erstwhile Republicans, who, by 1973, had become decidedly more Democratic or Independent in their identification. Those with Democratic leanings in 1965, in contrast, tended to stick with those views, although they sometimes gravitated toward a more independent affiliation. A similar pattern holds when considering the other political attitudes expressed in 1973. The young men hurt by Nixon's policy who had left childhood with Republican leanings ended up more sympathetic to Democratic and liberal causes by their mid-twenties.

LONG-TERM EFFECTS

Our cohort of 1965 high school college-bound senior men faced their pivotal draft lottery in December 1969. The observed political responses are from early 1973. Because we believe the intervening causal variable was trauma and disruption (or relief) induced by the lottery number, we observed causal impacts approximately 3 years after the initial stimulus. As political attitude studies go, this is a long duration. Rarely do we study attitudinal change over a span of years.

We also have the means to study the possibility of the persistence of the effect over the course of a political lifetime. We refer, of course, to the opportunity to examine responses from the third (1982) and fourth (1997) waves of the study. Here, we offer a brief assessment of long-term effects. First, we examine the possible additive effects. Did getting a low lottery number lead to a persistence of antiwar sentiment, Democratic voting, and liberalism beyond 1973? We also examine the possible continuation of the shock effect, whereby low lottery numbers reduced the effects of preadult party identification. For draft-vulnerable respondents, was this shock temporary? Did their party identifications revert back to their earlier party identification, or did they carry forward based on their new (as of 1973) partisanship?

Table 6 is our guide for looking for the persistence of additive effects.Footnote ³¹ For each of the selected items shown in Table 6, the analysis is based on the constant set of respondents with responses in all three postlottery waves. In general, effects appear to fade. This is quite clear for political ideology and the composite issue index. The effects of lottery number on candidate evaluations and the vote, which were so prominent in 1972, also dissipate by 1980.Footnote ³²

TABLE 6.

Long-term Additive Effects of 1969 Lottery Numbers on Political Attitudes: College Bound Only

Notes: Data are from the four-wave youth panel file. The dependent variables are scaled to run from 0 to 1, as described in footnote 31. Cases are college-bound (those whose 1965 high school curriculum was college preparatory) male respondents who had not served in the military as of 1969. The vote choice equations were estimated with probit. Shown for those dependent variables is the estimated change in the probability of a Republican vote as lottery number ranged from 1 to 366. The p value is from the test on the probit coefficient. Probit coefficients and robust-clustered standard errors (SEs) for the three vote-dependent variables, in turn, are 0.99 (0.44), 0.44 (0.43), and 0.25 (0.37). The probit pseudo R ² is the McKelvay-Zavoina estimate of the proportion of the variance in the latent variable that is explained. Entries shown for the other dependent variables are unstandardized coefficients from ordinary least squares (OLS), with robust-clustered SEs in parentheses.

However, an exception is with the central variable itself—Vietnam attitude. Here, we use responses to the “mistake” question because that was asked in each postlottery wave. The lottery effect on responses to the question about Vietnam being a mistake maintains most of its initial magnitude into the 1990s. Even in 1997, 28 years after the precipitating event, the difference between the lowest and highest lottery number was about a quarter of the range of the Dove-Hawk scale. It seems, then, that some “immediate” effects (e.g., at least 3 years in duration) faded later in life. But the central attitude of our study—attitude toward the Vietnam War—remained shaped by the luck of the draw in 1969.

The clearest case of another draft number effect persisting into late adulthood is the continuing interaction effect between lottery number and 1965 party ID (i.e., the upending of preadult partisanship for respondents with low lottery numbers). We saw that by 1973, the most draft-vulnerable respondents were particularly prone to abandon their partisanship from high school. In later panel waves, they could have reverted to their 1965 partisanship, as if their draft-induced partisan conversions were temporary aberrations. However, they did not, instead sticking with their new (1973) party identifications.

As illustration, Table 7 displays the correlations between 1965 and 1973 party identifications, on the one hand, and our selected attitudinal measures, on the other hand. The first set of correlations presented is for the bottom third of lottery numbers (most vulnerable) with the second set for those in the top third (relatively safe). For those with low draft numbers, 1965 partisanship correlates at near zero with all the relevant indicators. For instance, 1965 party identification has no correlation whatsoever with the frequency of Democratic voting in presidential elections from 1976 to 1996. Meanwhile, those with high draft numbers show correlations between 1965 partisanship and variables measured later that, although they decay, are decidedly higher than those for the vulnerable group. Thus, 1965 high school partisanship retained some predictive power into middle age for those with safe lottery numbers, but not for those drawing adverse draft lottery numbers. For both groups, however, the new postdraft 1973 party identification predicts adult attitudes fairly well.

TABLE 7.

Correlation of 1965 and 1973 Party Identification with 1973, 1982, and 1997 Political Attitudes by Lottery Number: College Bound Only

Notes: Data are from the four-wave youth panel file. Cases are college-bound (those whose 1965 high school curriculum was college preparatory) male respondents who had not served in the military as of 1969. Respondents were 26, 35, and 50 years old in 1973, 1982, and 1997, respectively. Minimum cell entry = 41. Correlations greater than 0.30 are statistically significant at 0.05 or better.

To summarize, the effects of lottery numbers persisted long beyond their immediate impact on possible military service. It is remarkable enough that the effects persisted from 1969 to 1973 when the first postlottery survey wave was conducted. In the case of attitude toward the war itself, the effect of lottery status persisted into the 1990s, when the respondents were middle aged. Once initial 1965 party identifications were destabilized, as we saw from the 1965 to 1973 panel analysis, respondents stuck with their new partisanship rather than revert to 1965 values. A prominent effect of getting a poor outcome in the draft lottery was to cause reevaluations of party loyalties. And these revised loyalties persisted into later adulthood.