Assignment 2: The Influence of Organized Crime on Political Leaders **FOR SOLUTIONS PRO ONLY***

Public Choice (2007) 130:251–266

DOI 10.1007/s11127-006-9016-1

O R I G I N A L A R T I C L E

Organized crime and the 1960 presidential election

John J. Binder

Received: 20 May 2005 / Accepted: 6 February 2006 C© Springer Science + Business Media B.V. 2007

Abstract It has been asserted that, based on a pre-election agreement promising them fa- vorable federal treatment, the Chicago Mob (known as the “Outfit”) was responsible for

John Kennedy’s election in 1960. An examination of these claims indicates that the sources

generally lack credibility and their accounts are implausible. Additionally, there is no evi-

dence Outfit controlled wards/suburbs around Chicago or members of Outfit influenced labor

unions voted unusually heavily Democratic in the 1960 presidential election. Therefore, if

anything the Outfit “double crossed” the Kennedys by not delivering the promised votes, as

opposed to vice-versa.

“Someone forgot to tell the horse.” – An old saying in horse racing.

1 Introduction

The 1960 presidential election is one of the most widely discussed events in the history of

American organized crime. The discussion focuses on the role of the Cosa Nostra crime

family in Chicago, known as the Chicago Outfit (or the Mob or Syndicate), in electing John

F. Kennedy. The initial claims by Brashler (1977) and Roemer (1989) are mild, suggesting

the Outfit may have worked for Kennedy in certain wards, but it played a minor role at best

compared to Mayor Richard J. Daley and the Democratic Party Machine.

More recent works (see Giancana & Giancana, 1992; Hersh 1997 & Russo, 2001) argue

that the Outfit made a massive effort which resulted in heavy voting for Kennedy (1)in various

parts of Chicago and/or (2) by labor union members, either in specific states where the Outfit

had influence or around the entire country. These authors assert that Joseph Kennedy, the

candidate’s father, struck a deal wherein John Kennedy would “lay off” organized crime

in return for the Outfit’s support in the election. Based on these stories, it is claimed the

J. J. Binder College of Business Administration (MC 168), University of Illinois at Chicago, 601 S. Morgan, Chicago, IL 60607, USA e-mail: [email protected]

Springer

252 Public Choice (2007) 130:251–266

Kennedys “double crossed” the Outfit by intensifying the federal attack on it and Giancana

and Giancana assert that the Chicago Mob retaliated by killing both John and Robert Kennedy.

Conversely, O’Connor (1975) argues the Outfit was negatively disposed toward John

Kennedy in 1960. He compares the voting by certain wards in the 1960 presidential election

to the outcomes in the 1955 mayoral election in Chicago and concludes that wards where

the Outfit controlled the Democratic party organization or the ward boss was friendly with

them voted unusually heavily against Kennedy. He also states that Outfit associated unions

greatly disliked the Kennedys due to the treatment they received at the McClellan Committee

Hearings.

This paper examines the role of the Chicago Outfit in the 1960 presidential election by

analyzing (1) the plausibility of several of the preceding arguments and the credibility of their

sources and (2) voting data for Chicago’s 50 wards and suburbs of interest, as well as for the 50

states. Although the Outfit was quite capable of delivering votes in certain Chicago area polit-

ical districts, when normal voting behavior is controlled for there is no evidence the Syndicate

provided unusual support for John Kennedy. Rather, it was mainly concerned with defeating

the incumbent Republican State’s Attorney for Cook County, Benjamin Adamowski, who

was running for re-election at the same time. Some of this effort, due to straight ticket voting,

spilled over to Kennedy, resulting in slightly stronger support than usual for him in the Outfit

controlled areas in 1960. There is also no evidence that union members voted unusually

heavily for Kennedy nationally or in states where the Outfit had strong influence. In fact,

union members in states where the Outfit controlled organized crime voted unusually heavily

against Kennedy. Based on these results, it is unlikely the Kennedys and the Outfit had an agreement before

the election, because the Chicago Mob had no reason to make a deal and then break it.

Certainly there was no “double cross” by the Kennedys when they attacked organized crime,

because the Outfit never meaningfully delivered votes for Kennedy, nor any reason for the

Outfit to retaliate against the Kennedys. If anything, the Outfit – if there was an agreement

before the election as has been claimed – double crossed the Kennedys.

2 The 1960 presidential election

The federal government first took concerted action against the Cosa Nostra, the organized

crime families in the United States which grew out of the Prohibition Era bootlegging gangs,

after the 1957 Appalachin (New York) meeting of gangsters from around the country was

exposed by the New York state police.1 At the same time, the McClellan Committee of

the U. S. Senate, including Senator John F. Kennedy and chief counsel Robert F. Kennedy,

probed organized crime in various cities, including Chicago.2 Robert Kennedy took particular

delight in annoying mobsters, including Sam Giancana, the operating boss of the Outfit. In

one exchange, Kennedy asked the smirking Giancana, “Are you going to tell us anything or

just giggle. I thought only little girls giggled.”3

Several writers claim the Outfit played a role in the election of John F. Kennedy in 1960.

The earliest statement, by William Brashler (1977), is quite mild. He argues that Frank Sinatra

approached Sam Giancana and asked him to help elect Kennedy. However, the Outfit’s efforts

1 See Binder (2003) for a general history of the Chicago Outfit as well as the Cosa Nostra. 2 See McClellan (1962). 3 See Brashler (1977, p. 157).

Springer

Public Choice (2007) 130:251–266 253

were secondary – the icing on the cake – to those of the Chicago Democratic Machine, which

went all out for an Irish-Catholic strongly supported by Mayor Richard Daley. According to

Brashler (p. 195), “an order from the mob to work for Kennedy only insured a total Chicago

effort of the kind that historically [italics added] had been known to work miracles in the early-morning hours of vote counting.” In other words, the Chicago Democratic Machine

delivered on election day as it usually did, with the Outfit controlled wards doing little/nothing

for Kennedy beyond what was done by other wards.

Similarly, former FBI agent William Roemer (1989, pp. 150–151) states:4

Whether or not Giancana got that commitment from Sinatra [that Kennedy would

halt the FBI investigation of Giancana], Giancana exercised influence through the

group of politicians known as the “West Side Bloc” [from political districts just

west of and including the Loop] . . . But much more important than Giancana’s

influence was that of Mayor Richard J. Daley, whose interests at the time just

happened to jibe with Giancana’s. To say that Giancana’s influence swung the

election to Kennedy would be farfetched; to say that Daley delivered Chicago

and Illinois may be true . . .

Roemer, it should be noted, had developed two high placed informants in the Outfit and was

uniquely aware of what happened in that world.

A greatly amplified version of these events, compared to the accounts by Brashler and

Roemer, appears in the book Double Cross by Sam and Chuck Giancana (1992), the half- nephew and half-brother, respectively, of Sam Giancana, the Outfit boss from 1957 to 1966.

According to the Giancanas, John Kennedy’s father, Joseph, contacted Sam Giancana before

the election. After several meetings, Giancana and the elder Kennedy struck a deal. “I help

get Jack elected and, in return, he calls off the heat,” Sam Giancana (pp. 279–280) is reputed

to have said.

Giancana and Giancana (pp. 289–290) claim the Outfit did everything possible for

Kennedy in the wards they controlled. Not only was there massive fraudulent voting, but

hoods inside polling places intimidated voters, making sure all ballots were cast for Kennedy

by breaking the arms and legs of those who refused to comply. The Kennedys, however,

“double-crossed” Giancana and the Outfit by ordering increased pressure on the Chicago

Mob, despite Giancana supposedly meeting with John Kennedy at the White House. This in

turn led the Outfit to assassinate both John and Robert Kennedy.5

Seymour Hersh (1997), in a chapter entitled “The Stolen Election”, also maintains there

was a pre-election deal. Former Chicago lawyer Robert McDonnell claims he arranged a meet-

ing between Joseph Kennedy and Sam Giancana in Chicago, which (curiously) McDonnell

saw take place but did not attend. McDonnell asserts the Outfit delivered votes at the ward level

in Chicago for Kennedy and also influenced various unions (although it is unclear whether he

means locally or nationally) to support Kennedy. Furthermore, the wife of Chicago mobster

Murray Humphreys, who in 1960 was the Outfit’s point man on political corruption, alleges

the Outfit delivered Teamster union votes at the national level. She claims to not only have

witnessed her husband co-ordinating this effort, but to have worked with him as Humphreys

directed Teamster leaders from around the country.

4 This is not surprising, because Roemer and other FBI agents were primary sources for Brashler’s biography of Sam Giancana. 5 See Giancana and Giancana (1992, chs. 20–22).

Springer

254 Public Choice (2007) 130:251–266

Gus Russo (2001) largely repeats the stories told by the Giancanas and Hersh. In his version

of the events, the Outfit used extreme measures to deliver the vote for Kennedy locally and,

through Murray Humphreys, made sure that union members nationally voted Democratic.6

Although he quotes Mrs. Humphreys’s Teamster focused account, Russo (p. 379) asserts

that non-Teamster union members around the country were influenced to vote for Kennedy. Russo (p. 401) places particular emphasis on the Outfit’s ability to deliver union votes in

four states: Illinois, Michigan, Missouri and Nevada. He believes (p. 401) the “contention

that it [the Outfit] ‘elected Jack’ is not without merit” and states more strongly (p. 1) that

Humphreys, Tony Accardo and other Chicago mobsters met in June 1960 to “decide who

would become the next president of the United States.”

On the other hand, Len O’Connor (1975), the dean of Chicago’s political commentators,

remarks:

The power of the Daley Machine was evident throughout the city, only the

two crime syndicate wards, the First and the Twenty-eighth, delivering a low

count, fewer votes [in terms of plurality] for Kennedy in 1960, in fact, than they

had delivered for Daley in 1955. The Machine interpreted this disappointing

performance as a mild rebuke by the syndicate people who had been mercilessly

pounded by the presidential candidate’s brother, Robert [during the McClellan

Committee hearings].7

O’Connor (1975, p. 151) notes that Charlie Weber, the Democratic 45th Ward alderman,

was persuaded by his friend Murray Humphreys to openly oppose Kennedy’s candidacy. In a nutshell, O’Connor’s view is that the Outfit controlled wards and the 45th Ward worked

against Kennedy in 1960. O’Connor was certainly very well informed about Chicago politics,

counting aldermen such as Weber among his sources and examines, at least in a cursory

fashion, voting data.8 He further notes (p. 150) that labor unions tied to the Outfit were very

displeased with Robert Kennedy and the McClellan Committee.

When closely scrutinized, the extreme versions put forth by the Giancanas, Hersh and

Russo are highly implausible and/or are based on sources who lack credibility. For example,

there is not one word in any of Chicago’s four major daily newspapers about violence directed

at voters in November 1960, much less of a 1920s style wave terror.9 More generally, the

Outfit did not have the ability to deliver meaningfully for Kennedy in Chicago.10 As discussed

below, it controlled the (Democratic party) political machinery in only five of Chicago’s 50

wards. Even if it had delivered unusual pluralities in those few wards, it is unlikely this would

have tipped the scale in a national election.

6 Russo (2001, p. 398) claims that “scores of Giancana’s ‘vote sluggers’ or ‘vote floaters’ hit the streets to ‘coerce’ the voters.” 7 See O’Connor (1975, p. 158). 8 For example, O’Connor (1975, p. 80) recounts a direct conversation in which Weber and fellow alderman and drinking companion Matthias “Paddy” Bauler disclosed behind the scenes details about Chicago politics to him. 9 Legendary crime reporter Ray Brennan, writing the day after the election, described it as “sissified” and “bland” in comparison to the violent and fraudulent primary election of April 6, 1928. See Brennan (1960a). 10 I am grateful to William Brashler for sharing his insights on these two points, some of which appear in a condensed form in his review of the Russo book. See Brashler (2002).

Springer

Public Choice (2007) 130:251–266 255

Moreover, the Outfit did not have the manpower to deliver in even those five wards. In 1960

there were 279 precincts/polling places in the five “Outfit wards.”11 To effectively intimidate

voters at a polling place, it would have taken at least four or five goons – a smaller number

would have allowed irate voters to possibly pummel the “intimidators”. With some 300 full

members in 1960, many of whom were advanced in age, the Outfit would at best have been

able to (if it so desired and the police did not intervene), coerce voters in one of these wards, each ward having between 46 and 63 precincts.12 On the latter point it is noteworthy that

when Al Capone used violence in 1924 to help elect the Republican candidates in the suburb

of Cicero, which had about 70,000 residents at the time, he needed to get additional men from

Dion O’Banion’s North Side gang.13 Certainly Capone’s gang was larger than the Outfit in

1960 while Cicero was much smaller in population than the five Outfit wards.14

Allegations the Outfit manipulated the Teamsters or other unions nationally are equally

implausible. Individual Cosa Nostra crime families generally controlled local chapters of

unions, rather than the national union.15 Therefore, the Outfit could not command union

officials from the entire country to do its bidding.16 More important, Teamsters’ boss Jimmy

Hoffa despised the Kennedys and publicly endorsed Richard Nixon, eliminating the possi-

bility this union worked for John Kennedy.17

It is also difficult to believe that Joseph Kennedy met with a notorious gangster under

investigation by a Senate committee his two sons were associated with. If he had been seen,

overheard or linked to Giancana, the damage to John Kennedy’s campaign would have been

immeasurable. Even a hint of such a meeting, leaked by someone involved, would have

been damaging. Also, it is hard to comprehend how the Outfit, having been attacked by the

McClellan Committee, would trust the Kennedys. In fact, Brennan (1960b) reports two days

after the election that John Kennedy was going to crack down hard on organized crime,

especially the Outfit in co-operation with Chicago’s police superintendent, as an outgrowth

of his activities with the McClellan Committee hearings.

Furthermore, the credibility of several individuals who have made these claims is at best

questionable. Organized crime operates with the same degree of secrecy as major intelligence

agencies such as the CIA. Only those who absolutely “need to know” are informed at the time

about particular operations. The average full member of the Outfit would not have known

the information the Giancanas claim to have known, much less Chuck Giancana, who was

11 The precincts are enumerated in a table reporting the voting for the office of Cook County State’s Attorney in the Chicago Daily News (November 9, 1960, p. 26). 12 The Outfit’s size in 1960 is estimated by Art Bilek, chief of the Cook County Sheriff’s Police from 1962 to 1966. The identical figure is reported by Captain William J. Duffy of the Chicago Police Department in 1963 (see Moore (1963)). 13 See the Chicago Tribune, July 4, 1926, on the takeover of Cicero. 14 Capone had 500 men capable of carrying a gun in his employ during Prohibition [see Doherty (1951)]. The combined population of the 1st, 24th, 25th, 28th and 29th wards (see Scammon (1964)) in 1960 was 306,169. 15 Neff (1989) discusses the control of Teamster locals by individual crime families. 16 As discussed below, the Outfit did control or greatly influence the crime families in Rockford and Springfield, Illinois, Milwaukee and Madison, Wisconsin, Kansas City, Missouri and Los Angeles and had the largest presence in Las Vegas, which was an “open city”. Even if union officials in these cities were under Chicago’s thumb, this hardly constitutes the entire country. Also, Chicago did not control Cleveland or St. Louis, as some of the books mentioned above assert. 17 Russo (p. 379) is aware of this fact, which probably leads him to claim that the Outfit influenced other unions to vote for Kennedy. Of course, the leader of a Teamsters union local, especially if he was Irish and admired John Kennedy, might have supported Kennedy, but this hardly constitutes a national effort.

Springer

256 Public Choice (2007) 130:251–266

only a lowly Mob associate.18 Moreover, the Giancanas’ book is not taken seriously by well

informed students of the Chicago Outfit.19 In it the authors claim Sam Giancana was involved

in every major organized crime event in Chicago from his adolescent years onward, even

though most of their assertions are contradicted by known facts or are unsupported by other

evidence.

The same point applies even more strongly to the wife of Murray Humphreys. In the

completely male world of the Cosa Nostra, members do not share information with females,

including wives. In fact, female relatives of gangsters make remarks such as, “I’m a girl.

They never told me anything.”20 Certainly union leaders would have refused to talk busi-

ness with Humphreys if a woman or non-Outfit member were present.21 If Humphreys had

even suggested to his superiors that his wife attend business meetings – much less that she

work with him – they would have decided he was insane and likely killed him and her

also.22

Bob McDonnell is similarly lacking in credibility. A disbarred attorney who was a com-

pulsive drinker and uncontrollable gambler, McDonnell borrowed heavily from Outfit as-

sociated loan shark Sam DeStefano. When he was unable to pay his debts, DeStefano put

McDonnell to work for him, including having him carry two dead bodies from his base-

ment.23 It is difficult to find informed, unbiased individuals, especially retired Chicago

police officers who worked on organized crime, who place faith in statements by Robert

McDonnell.24

3 Methodology and data

Several of the preceding accounts state that Outfit controlled or influenced political districts

and/or labor unions voted unusually heavily for – or against – John Kennedy in 1960. Other

accounts claim that the Outfit exerted little, if any, unusual effort for Kennedy. These hypothe-

ses can be directly tested by examining voting in certain wards or suburbs around Chicago

and by union members nationally or in specific states.

18 Similarly, in the intelligence community CIA or FBI agents are not allowed to discuss their work with relatives. 19 See Brashler (1992) and Reilly (1992). 20 From the author’s conversation with the daughter of at least one prominent Chicago gangster. 21 Retired police officers who specialized in organize crime, including former members of the Chicago Police Department’s Intelligence Unit, scoff at the notion that non-mobster relatives of gangsters would have any information about the Outfit that is not publicly available, such as what is reported in the newspapers. 22 This point is made earlier in Binder (2003, p. 100). I am grateful to Art Bilek for his thoughts on the subject. 23 See Demaris (1969, pp. 50–54). 24 The details of McDonnell’s story about the meeting between Joseph Kennedy and Sam Giancana, which appear in Hersh (1977), are also not plausible. First, Kennedy supposedly solicited Chicago Judge William Tuohy for help in contacting Sam Giancana, who in turn contacted Bob McDonnell. Yet McDonnell admits he did not know Giancana. Tuohy could quite easily have contacted First Ward Democratic politicians, such as John D’Arco or Pat Marcy who were close to Giancana, to more effectively arrange a meeting. Second, McDonnell claims Tuohy wanted him present at the actual meeting (which allows McDonnell to state the meeting did in fact take place). Yet as soon as the parties were introduced, Tuohy and McDonnell left the building. If McDonnell’s presence was not required, why was it necessary that he even attend? Third, it is unlikely that someone as unstable, unreliable and potentially troublesome as McDonnell would be trusted with information this secretive and sensitive.

Springer

Public Choice (2007) 130:251–266 257

Assume that the percentage of voters in state i voting Democratic during the 1960 pres- idential election, pi 60, is related to explanatory variables by the logistic function.

25 The log

odds ratio ODDSRATIOi 60 is a linear function of these variables:

ODDSRATIOi 60 = β0 + k∑

m=1 βm X mi 60 + βk+1UNIONi 60 + βk+2OCi j UNIONi 60 + ui 60,

(1)

where ODDSRATIOi60 = ln[ pi 60/(1 − pi 60)], and the X’s are variables not influenced by organized crime that determine pi 60 and ODDSRATIOi 60. The variable UNIONi 60 measures the percentage of the work force in the state that is unionized (excluding members of the

Teamsters Union). OCi j equals one if the Outfit had unusual influence in the state, as asserted

by Russo or because it controlled (at least partially) organized crime operations there. The

interactive variable OCi j UNIONi 60 allows for unusually heavy Democratic voting by union

members in these states in 1960. Assume a similar relation for the 1956 presidential election,

in which the Outfit exerted no influence:

ODDSRATIOi 56 = α0 + k∑

m=1 αm X mi 56 + αk+1UNIONi 56 + αk+2OCi j UNIONi 56 + ui 56.

(2)

Equations (1) and (2) are estimated jointly as Seemingly Unrelated Regressions (SUR) using

weighted generalized least squares.26

Regarding the independent variables in Equations (1) and (2), National Election Stud-

ies (NES) surveys by the Center for Political Studies at the University of Michigan find

that males, blacks, Southerners, blue collar workers, people with a high school educa-

tion or less and people in the lower 95 percent of the income distribution tended to

vote more heavily Democratic than the average voter. It is likely that in 1960 Catholics

and people of Irish descent also voted unusually heavily Democratic. These variables,

along with indicator variables that equal one if the presidential or vice presidential can-

didate lived in the state, control for the usual factors that affect voting in presidential

elections.

Most of the data used to measure the control variables are from the United States Census.27

The percentage of the electorate that is female, FEMALE, is measured as the estimated

number of women over the age of 21 in the state during 1956 or 1960 divided by the total

population over the age of 21 that year. The variable BLACK is the number of African-

Americans in the state divided by the total population. The qualitative variable SOUTH

equals one for each of the 13 Southern states and zero otherwise.28 The percentage of the

25 See Neter and Wasserman (1974, p. 329). 26 Theil (1971, pp. 294–317) discusses the structure and estimation of Seemingly Unrelated Regressions. In this application, each equation represents a cross-section (year) and observations vary by state, as opposed to the usual case where the statistician has time series data for each equation. Because the standard SUR variance-covariance structure for the error terms is preserved, the method can be used with this data set. The variables, including the vector of ones, are multiplied by the square root of p(1 − p)N , where N is the number of votes cast in the state during the election, to control for hetero-skedasticity inherent in logit models. On the latter point, see Neter and Wasserman (1974, equation (9.53), p. 331) and the related discussion on pp. 131–132. 27 These data are reported in the 1950 or 1960 Censuses or in the Statistical Abstract of the United States. 28 The Southern states are Virginia, West Virginia, North Carolina, South Carolina, Georgia, Florida, Alabama, Mississippi, Arkansas, Louisiana, Texas, Tennessee and Kentucky.

Springer

258 Public Choice (2007) 130:251–266

labor force with blue collar jobs, BLUECOLLAR, equals the estimated number of blue collar

workers divided by the experienced civilian labor force.29 The percentage of the population

whose education terminated with a high school degree or less, EDUCATION, is the estimated

number of people with no more than a high school degree divided by the population 25 years

or older. The variable INCOME is the percentage of the population 14 years and older with

an income above $10,000.30,31

Data on the number of Catholics in each state are obtained for 1956 and 1960 from The Official Catholic Directory. These figures are divided by the state’s estimated total population to calculate the percentage of the population that is Roman Catholic, CATHOLIC. The number

of people of Irish stock, meaning the person or their father was born in the Republic of Ireland,

in each state is from the 1950 and 1960 Censuses and the value for 1956 is obtained by linear

interpolation. The 1956 and 1960 figures are divided by total population to estimate the

percentage of the population of Irish extraction IRISH. The number of AFL-CIO members in

each state, first reported in 1958, is used as an estimate of the figure for 1956. Similar data are

reported for 1960.32 These data exclude Teamsters Union members because the Teamsters

were expelled from the AFL-CIO in 1957. Union membership is divided by the experienced

civilian labor force each year to estimate UNION, the percentage of the work force that is

unionized (but not Teamster affiliated). Data on voting in the 1956 and 1960 elections are

from the Statistical Abstract of the United States (1962).33

Russo’s hypotheses are that union members nationally voted more heavily Democratic

in 1960 than other voters and they voted more heavily Democratic in 1960 than in 1956, i.

e., βk+1 is positive and greater than αk+1. The tests are, however, biased in favor of Russo’s prediction if union members usually voted more heavily Democratic than other voters or

they were relatively unenthusiastic about Adlai Stevenson in 1956 as opposed to Kennedy in

1960.34

If the Outfit was able to deliver union votes for Kennedy anywhere, it would have been in

states where it had organized crime activities, including labor racketeering. The variable OCi 1 equals one for the four states mentioned by Russo – Illinois, Michigan, Missouri and Nevada

– and zero otherwise. Russo’s list is curious because the Outfit had no influence in Michigan

– the Detroit Mob is independent of Chicago – and only minor influence in Nevada, where

they were one of several crime families operating in the “open city” of Las Vegas. Given the

Outfit’s control of the crime families in Milwaukee and Madison, Wisconsin, Rockford and

Springfield, Illinois, Kansas City, Missouri and Los Angeles and its operations in Illinois,

29 Blue collar individuals are those denoted by the Census as clerical and kindred workers, sales workers, craftsmen (including foremen and kindred workers) and operations and kindred workers. 30 This is the divisor used by the Census when reporting percentages of the population in various income groups. 31 In calculating the preceding ratios for 1960 the numerator value is taken directly from the Census, while in every case the 1956 value for the numerator is a linear interpolation of the 1950 and 1960 Census values. The denominators of these ratios for 1956 are also linear interpolations, except for the divisors used to calculate the variables FEMALE and BLACK (which are reported by the government for 1956), while those for 1960 are reported by the Census. 32 The data are from the January 1960 and December 1961 issues of the Monthly Labor Review. 33 Candidates from other parties received less than one percent of the votes cast in the 1956 and 1960 presidential elections and are excluded from the analysis. That is, the variable p is calculated based on just the votes for the candidates of the two major parties. 34 O’Connor (1975, pp. 68 and 145) discusses Stevenson’s lack of political savvy and unpopularity with run of the mill Democrats. McKeever (1989, pp. 218–219) notes that as early as 1952 Stevenson did not garner union support, due to his unwillingness to call for the repeal of the Taft-Hartley Act.

Springer

Public Choice (2007) 130:251–266 259

Northern Indiana, Iowa and in Las Vegas, a better list of states where the Outfit had influence

is: Illinois, Indiana, Iowa, Missouri, Wisconsin, California and Nevada.35 The variable OCi 2 equals one for these states and zero otherwise.

Russo – and Hersh, if McDonnell’s assertion is that the Outfit only influenced unions that

it directly controlled – predicts that βk+2 will be positive while O’Connor predicts that this coefficient will be negative. Tests of these hypotheses are biased if union members regularly

voted more strongly Democratic in some states than in others. A better test, because it controls

for other factors that may affect the coefficients in Equations (1) and (2), is to examine whether

βk+2 equals αk+2. According to Hersh and Russo, βk+2 is greater than αk+2 while O’Connor argues the opposite. This test, and the preceding three, are conducted with the system of

Equations (1) and (2).36

Regarding voting in the Chicago area, Census data are not reported at the ward level.

Therefore, assume a linear relation between ODDSRATIOi 60 and the explanatory variables

for the 50 wards in Chicago and for suburbs of interest in 1960:

ODDSRATIOi 60 = θ0 + δi + θ1 Di 60 + θ2OCi j + ei 60, (3)

where δi is a fixed effect [see Wooldridge (2003, ch. 13)] which varies across wards/suburbs

but not over time and OCi j , with j ≥ 3, equals one if the Democratic ward organization in Chicago, or the suburb in general, was controlled by the Outfit. The variable Di 60 equals one if the Democratic candidate resided in ward or suburb i (as in some local elections examined below), minus one if a non-Democratic candidate lived there and zero

otherwise.

Assume a similar relation for an election at time t which is compared to the 1960 presi- dential election:

ODDSRATIOit = ϕ0 + δi + ϕ1 Dit + ϕ2OCi j + eit, (4)

where ideally organized crime did not exert any effort for or against the Democratic candidate

in this election. That is, ϕ2 equals zero. Subtracting ODDSRATIOit from ODDSRATIOi 60,

moving ODDSRATIOit to the right hand side of the equation and inserting a coefficient

before it yields:

ODDSRATIOi 60 = γ0 + γ1ODDSRATIOit + γ2 Ii + γ3OCi j + εi , (5)

where Ii equals Di 60 minus Dit. Hypotheses about the Outfit’s role in the 1960 election make predictions about the θ2 in Equation (3) and therefore γ3. Brashler and Roemer claim the

Chicago Mob did little or nothing beyond what the Chicago Democratic Machine did in

general, implying γ3 is zero or slightly positive. Giancana and Giancana, Hersh and Russo

assert the Outfit strongly delivered votes for Kennedy, which predicts γ3 is positive while

O’Connor predicts γ3 is be negative. 37

35 Binder (2003, pp. 28 and 77) discusses the Outfit’s influence in other areas. 36 Regressions that use aggregate data, such as this one, may yield biased inferences (see King (1997)) about individual voting behavior. Unfortunately, detailed data on individuals and how they voted is not available for the 1960 presidential election. 37 Hypothesis tests about γ3 control for any tendency by the Outfit controlled areas to normally vote more heavily Democratic than other political districts in the Chicago area because that coefficient measures how those areas voted above and beyond their voting during year t .

Springer

260 Public Choice (2007) 130:251–266

Equation (5) is estimated with data for the 50 wards in Chicago and the two premier Outfit

controlled suburbs, Chicago Heights and Cicero, using weighted least squares to correct for

heteroskedasticity.38 Data on local voting in the 1960 presidential election and other elections

of interest are from the Chicago Tribune and the Chicago Daily News.39 The home addresses of the candidates in any local elections examined below are obtained from newspaper articles

and other sources.

O’Connor (1975) explicitly mentions the 1st and the 28th ward organizations as being

Mob controlled.40 Demaris (1969, pp. 151–185) surveys the links between organized crime

and Chicago politicians through the 1960s. In terms of the strongest Mob influence on politics

in Chicago, he discusses five wards: the 1st, 24th, 25th, 28th and 29th. Based on this, four

indicator variables are used to denote wards/suburbs in which the Outfit had unusual political

influence. The variable OCi 3 equals one for the 1st, 24th, 25th, 28th and 29th wards and zero

otherwise. Folowing O’Connor’s argument that the 45th ward alderman also worked to elect

Nixon, OCi 4 equals one for the same five wards as OCi 3 and the 45th ward. The variable

OCi 5 equals one for the five Outfit controlled wards, Chicago Heights and Cicero. Finally,

OCi 6 equals one for the five Outfit controlled wards, the 45th ward and Chicago Heights and

Cicero.

4 Empirical results

Three of the potential explanatory variables, FEMALE, BLUECOLLAR and EDUCATION,

are collinear with the vector of ones representing the intercept when the data are weighted to

control for heterskedasticity.41 They are therefore, without loss of explanatory power, dropped

from the analysis. When the variable OCi 1 denotes states where the Outfit had influence, the

estimated equations are:

ODDSRATIOi 56 = − .23 + 1.26 BLACKi 56 + .02 SOUTHi 56 − 3.60 INCOMEi 56 (1.83)∗∗(2.26)∗ (.19) (.75)

− .72 CATHOLICi 56 + .23 IRISHi 56 + .44 UNIONi 56 (1.84) (.08) (.52)

+ .06 UNIONi 56OCi 1 + ûi 56 (.09)

R2 = .68

(2)

38 Binder (2003) discusses the Outfit’s long standing control of these suburbs. The weights are discussed in footnote 26 and the equations are estimated using the Robusterrors option in RATS to also control for other types of heteroskecasticity. 39 Because other candidates received very few votes in the elections examined below, voting data are collected only for the candidates from the two major parties. 40 See Tables 1 to 6 and elsewhere in O’Connor (1975). 41 See footnote 26.

Springer

Public Choice (2007) 130:251–266 261

ODDSRATIOi 60 = − .43 + .96 BLACKi 60 + .19 SOUTHi 60 + 1.04 INCOMEi 60 (4.44)∗(2.43)∗ (2.25)∗ (.40)

+ .80 CATHOLICi 60 + 3.03 IRISHi 60 + .91 UNIONi 60 (2.85)∗ (1.12) (1.20)

− .29 UNIONi 60OCi 1 + ûi 60 (.51)

R2 = .58.

(1)

T -statistics (in absolute value) are in parentheses and the R-squareds are reported below each equation. An asterisk (two asterisks) indicates that the coefficient estimate is statistically

significant at the five (ten) percent level in the appropriate one or two tailed test.

The overall fit is quite good, with both R-squareds greater than .55. The results for 1956 indicate that blacks voted more heavily Democratic than the average voter, while Southern

voters and union members did not. In 1960 blacks, Southerners and Catholics, as expected

based on traditional voting patterns and Kennedy’s religion, voted more heavily Democratic

than the average voter, while surprisingly people of Irish extraction did not. Union members

also did not vote unusually heavily Democratic in 1960 or 1956 (all else equal).42

The insignificant coefficient estimate on UNIONi 60 is evidence against the argument that

union members nationally voted heavily for John Kennedy. Similarly, the hypothesis that the

coefficient on UNIONi 60 is greater than the coefficient on UNIONi 56 is not rejected at the five

percent level, which is also evidence against the Russo argument.43 Furthermore, because the

coefficient estimate on the interactive variable in Equation (1) is not significantly different

from zero, there is no evidence that union members in the states highlighted by Russo voted

more or less heavily Democratic in 1960 than other union members. Finally, the hypotheses

that union members in Outfit influenced states voted more/less strongly Democratic in 1960

than in 1956 are also rejected.44 This is evidence against the Hersh, Russo and O’Connor

arguments.

When OCi 2 is used in the regressions, the estimated equations are:

ODDSRATIOi 56 = − .19 + 1.22 BLACKi 56 + .03 SOUTHi 56 − 5.75 INCOMEi 56

(1.46)(2.33)∗ (.24) (1.14) −.82 CATHOLICi 56 + 1.57 IRISHi 56 + .39 UNIONi 56

(2.18) (.54) (.48)

+ .69 UNIONi 56OCi 2 + ûi 56 (1.11)

R2 = .68

(2)

42 The insignificant t statistics on the union variables in both systems of equations is probably due to the inclusion of the Catholic and Irish variables. When the dependent variable is regressed against just the union variable the coefficient estimates on UNION are negative in 1956 and positive in 1960 and both are statistically significant at the five percent level in two tailed tests. 43 The χ 2(1) statistic testing this restriction equals .30 and has a p-value of .583. 44 The χ 2(1) statistic testing the restriction that the coefficients on UNIONi 56OCi 1 and UNIONi 60OCi 1 are equal has a p-value of .492.

Springer

262 Public Choice (2007) 130:251–266

ODDSRATIOi 60 = − .43 + .88 BLACKi 60 + .20 SOUTHi 60 + 1.24 INCOMEi 60 (4.36)∗(2.36)∗ (2.36)∗ (.46)

+ .74 CATHOLICi 60 + 3.45 IRISHi 60 + .90 UNIONi 60 (2.68)∗ (1.26) (1.22)

− .11 UNIONi 60OCi 2 + ûi 60 (.21)

R2 = .58

(1)

Overall, the results are similar to those based on OCi 1. However, when OCi 2 is used

the χ 2(1) statistic testing the restriction that the coefficients on the interactive variables

are equal has a p-value of .074. This is evidence in support of O’Connor’s argument that Outfit influenced unions voted unusually heavily against John Kennedy. Therefore, in the

two systems of equations there is no support for the arguments made by Hersh and Russo

but there is evidence in support of O’Connor’s hypothesis.

When estimating Equation (5), there are several choices for the control variable

ODDSRATIOit. One possibility is to use data from the Chicago mayoral election in April

1955 as O’Connor does.45 However, the Outfit has supported Democratic candidates more

heavily in mayoral elections than presidential ones, especially in 1955, which biases such

tests against the hypotheses by the Giancanas, Hersh and Russo.46

With the exception of 1960, there are no credible claims that the Outfit worked for or

against candidates in a presidential election. Therefore, ODDSRATIOit is measured using

data from the 1956 and 1964 presidential elections. The estimated coefficients on the orga-

nized crime variables are insignificant in all four regressions when ODDSRATIOi 56 is the

control variable.47 The coefficient estimates and related statistics for the regressions which

use ODDSRATIOi 64 as an explanatory variable are reported in Table 1. Because none of

the presidential candidates lived in Chicago, Chicago Heights or Cicero, Ii is a vector of zeros in these cases and is excluded from the analysis. The estimated intercepts, coefficients

on ODDSRATIO and the R-squareds are very similar across the four regressions, with the R-squareds slightly below .50.48 Only when OCi 3 measures the influence of organized crime is there evidence that the Outfit wards behaved unusually in the 1960 presidential election.

45 During this period there were only a few changes in Chicago ward boundaries and some of these were fairly minor. For example, the boundaries were the same from October 1931 to September 1947. The remapping in 1947 consisted of only one major change; the old 20th ward was merged with the 1st ward and a new 20th ward was created in the middle of the South Side. In consequence, the ward boundaries changed only slightly from 1931 to 1961. In September 1961 the old 21st ward was absorbed by the surrounding wards, with a new 21st ward appearing on the Far South Side, and the 45th ward moved from the Near North Side to the Far Northwest Side. 46 For example, when the log odds ratio for each mayoral election in Chicago from 1947 to 1971 is regressed against the same variable measured with data from the next year’s presidential election, the coefficient estimate on OCi 3 is positive and statistically significant at the five percent level in 6 of 7 regressions and at the ten percent level in the remaining test. 47 In the interest of brevity, the results are not reported here. 48 The somewhat low R-squareds, compared to other estimates of Equation (6), are most likely due to the remapping of the Chicago wards in 1961. See footnote 45.

Springer

Public Choice (2007) 130:251–266 263

Table 1 Estimates from Equation (5) for presidential voting in 1960 ODDSRATIOi 60 = γ0 + γ1ODDSRATIOi 64 + γ2 Ii + γ3OCi j + εi Explanatory variable γ̂0 γ̂1 γ̂2 γ̂3 R

2 ̄ p

OCi 3 .15 (2.67) ∗ .39 (10.45)∗ – .37 (2.07)∗ .49 .060

OCi 4 .14 (2.46) ∗ .42 (10.42)∗ – .15 (.83) .47 –

OCi 5 .14 (2.41) ∗ .41 (10.67)∗ – .18 (1.13) .47 –

OCi 6 .14 (2.36) ∗ .42 (10.88)∗ – .07 (.59) .46 –

Although the logistic model is nonlinear, the effect of organized crime on pi 60 can be esti-

mated using a transformation of the logit regression’s fitted values.49 The estimated average

effect of organized crime on pi 60, ̄ p , for the Outfit influenced districts is .06 for the one regression in Table 1 with significant results.

Overall, the evidence in support of the claim the Outfit worked for John Kennedy in

the Chicago area is quite weak because the null hypothesis is rejected in only one of eight

tests. Thus, the evidence from the Chicago area is consistent with the Brashler and Roemer

hypotheses the Mob did little or nothing for Kennedy, but not the allegations by the Giancanas,

Hersh and Russo or O’Connor’s claim that the Outfit worked against Kennedy.

However, analyses of the presidential vote in isolation still ignore local issues which may

cause the preceding results, as weak as they are, to be biased in favor of assertions the Outfit

supported Kennedy. The regular election for Cook County State’s Attorney also took place

in November 1960. During the previous four years the Republican incumbent, Benjamin

Adamowski, was a thorn in the side of the Outfit, conducting raids on gambling joints in

Cicero and strip clubs in Calumet City starting in 1958, and City Hall.50 It was widely believed

that Adamowski, if re-elected, would expose further crime and corruption and then run for

mayor in 1963. The Outfit, therefore, worked hard in support of Adamowski’s opponent,

Daniel Ward, in November 1960. If some of this effort involved straight ticket voting, legal

or fraudulent, e. g., Democratic precinct captains pulling the voting machine levers for illegal

voters as was claimed by the Republican party, then the efforts against Adamowski contributed

votes to Kennedy as a side effect.51

A test of whether the Outfit was more interested in defeating Adamowski than electing

Kennedy is to regress the log odds ratio for the 1960 election of the Cook County State’s

Attorney, ODDSRATIOis60, against ODDSRATIOi 60. If voting for Dan Ward spilled over to

Kennedy, the coefficients on the indicator variable for the Outfit districts in Equations (3)

(with ODDSRATIOis60 on the left hand side) and (4) (with ODDSRATIOi 60 on the left hand

side) will both be positive but the net effect, measured by γ 3 in Equation (5), will also be

positive.

The results are reported in Table 2. The R-squareds in all four equations are quite high, at or near .90, because the elections are concurrent. The estimates of the intercepts and the

slope coefficients on the log odds ratio for the 1960 presidential election and the home district

indicator variable Ii are fairly close to one another, with the bulk of the explanatory power

49 For each ward in question, the fitted values from Equation (6) are used in equation (9.55) of Neter and Wasserman (1974, p. 331) to transform the predicted value of ODDSRATIOi 60 to a prediction for pi 60. The predicted value of pi 60 is then recalculated with the organized crime indicator variable in Equation (6) set to zero and the average difference between the predicted values is computed. 50 An article in the on-line version of the Northwest Indiana News, nwitimes.com, entitled “Feds [sic] Clamped Down on Clubs; Scaring Away Their Patrons” recounts the 1950s in Calumet City, Illinois. 51 The efforts to defeat Adamowski are discussed in greater detail in the preceding draft of this paper.

Springer

264 Public Choice (2007) 130:251–266

Table 2 Estimates from Equation (5) comparing the state’s attorney election to the presidential election in 1960 ODDSRATIOi s60 = γ0 + γ1 ODDSRATIOi 60 + γ2 Ii + γ3OCi j + εi Explanatory variable γ̂0 γ̂1 γ̂2 γ̂3 R

2 ̄ p

OCi 3 −.28 (9.36)∗ .95 (19.26)∗ .04 (1.53)∗∗ .24 (3.18)∗ .90 .038 OCi 4 −.29 (10.35)∗ .96 (22.80)∗ .03 (1.22) .21 (4.62)∗ .90 .036 OCi 5 −.29 (9.73)∗ .98 (21.71)∗ .03 (1.16) .11 (1.32)∗∗ .89 .021 OCi 6 −.30 (10.14)∗ .98 (23.36)∗ .03 (.95) .13 (1.95)∗ .90 .025

coming from the log odds ratio. The coefficient estimates on the organized crime variables

are all positive and significant in one tailed tests at better than the five percent level in three

cases and at the ten percent level in the remaining case, indicating that the Outfit worked

much harder for Dan Ward than for John Kennedy because their objective was to defeat Ben

Adamowski.

The estimates of ̄ p , which measure the Outfit’s effect on the percentage of votes cast for

Dan Ward above and beyond the effect they had on Kennedy’s vote percentage, range from

.021 to .038. Similarly, in the 1955 Chicago mayoral election – where the Outfit greatly feared

the Republican candidate Robert Merriam who ran a strong anti-organized crime campaign

– the estimates of ̄ p , which measure the Outfit’s effect on the percentage of votes cast for

Richard Daley in the wards of interest, range from .067 to .091.52 This indicates the Outfit

was able to deliver strongly in these wards/suburbs when it so desired. It simply was not

interested in delivering for John Kennedy.

Seasoned politicians, such as the Kennedys, would have recognized that Outfit influenced

labor unions voted against John Kennedy and that the Mob’s behavior locally was self-

serving. Therefore, they would have owed the Outfit nothing, even if there was a pre-election

agreement. Thus, there was no “double cross” when the Kennedy administration intensified

the fight against organized crime. In fact, the results do not support the claim that there was

a pre-election deal, because the Outfit had nothing to gain by making an agreement and then

breaking it. Or, if there was an agreement, the Outfit “double crossed” the Kennedys by not

delivering the promised votes.

5 Conclusion

A famous story in the world of horse racing concerns a supposedly fixed race. Various

people had been told about the race, “knew” which horse would win and bet accordingly.

Unfortunately, the horse did not win, prompting one bettor to remark, “Someone forgot to

tell the horse.”53

There are several extreme claims about the role of the Chicago Outfit in the 1960 presiden-

tial election. While these stories lack plausibility when carefully examined and the sources

are not credible, more importantly someone apparently “forgot to tell the voters.” That is,

there is no evidence that labor union members nationally, or in Outfit influenced states, voted

unusually heavily Democratic in the 1960 presidential election, as asserted by Hersh and

52 In the same vein as the regressions reported in Table 2, voting in the 1955 mayoral election is compared to the 1956 presidential election. The results, which are statistically significant in both tests, are discussed in earlier drafts of this paper. See also footnote 46. 53 I am grateful to Ronnie Kelpsas for sharing this analogy with me.

Springer

Public Choice (2007) 130:251–266 265

Russo. In fact, there is evidence that non-Teamster union members in states where the Outfit

operated voted unusually heavily against John Kennedy.

Locally, there is no evidence that Outfit controlled political units voted abnormally heavily

for John Kennedy in 1960. Certainly the Mob’s efforts for Kennedy were nowhere near what

they were capable of, as demonstrated by voting in the 1960 election for Cook County State’s

Attorney or the 1955 Chicago mayoral election. In fact, a study of local politics indicates

that the Outfit’s real objective in 1960 was to defeat Republican State’s Attorney Benjamin

Adamowski. Due to straight ticket voting this gave some votes to Kennedy purely as a side

effect, which is consistent with the Brashler and Roemer arguments.

Based on these results, it is unlikely the Kennedys struck a deal with the Outfit before

the 1960 election. Certainly, John Kennedy did not double cross the Chicago Mob because

they never supported him on election day. Therefore the extreme claims about the Outfit and

events related to the 1960 presidential election by Giancana and Giancana (1992), Hersh

(1997) and Russo (2001) appear to have no basis in fact.

Acknowledgements I am grateful to Chris Berry, Art Bilek, Bill Brashler, Al Egan, Mars Eghigian, Mickey Lombardo, Matt Luzi, Tim Perri, Vince Sacco, Jeff Thurston, Walter Washington and seminar participants at the Harris School of Public Policy at the University of Chicago and the Institute for Government and Policy Analysis at University of Illinois at Chicago for their comments and insight.

References

Binder, J. J. (2003). The Chicago outfit. Chicago, IL: Arcadia Publishing. Brashler, W. (1977). The Don: The life and death of Sam Giancana. New York, NY: Harper and Row. Brashler, W. (1992). The Don did it! Silly mob memories. Chicago Sun-Times (March 15). Brashler, W. (2002). Nostalgia for Chicago’s mob. Chicago Sun-Times (April 7). Brennan, R. (1960a). Pineapple primary recalled in election contest. Chicago Sun-Times (November 9). Brennan, R. (1960b). See Kennedy crackdown on mobsters. Chicago Sun-Times (November 10). The Chicago Daily News. Various editions. (Chicago Daily News: Chicago, IL). The Chicago Tribune. Various editions. (Tribune Publishing Company: Chicago, IL). Demaris, O. (1969). Captive City. New York, NY: Lyle, Stuart, Inc. Doherty, J. (1951).“Al Capone: King of the hoodlums. Chicago Sun (April 1). Feds Clamped Down on Clubs, Scaring Away Their Patrons. nwitimes.com (http://www.thetimesonline

.com/articles/2003/08/10). Giancana, S., & Giancana, C. (1992). Double Cross. New York, NY: Warner Books. Hersh, S. (1997). The dark side of camelot. (Little, Brown and Company: New York, NY). King, G. (1997). A solution to the ecological regression problem. Princeton, NJ: Princeton University Press. McKeever, P. (1989). Adlai Stevenson: His life and legacy. New York, NY: William Morrow and Company,

Inc. McClellan, J. H. (1962). Crime without punishment. New York, NY: Duell, Sloan and Pearce. Moore, W. (1963). Wilson tells probers gang executions here total 976. Chicago Tribune (October 12). Neff, J. (1989). Mobbed up. New York, NY: Atlantic Monthly Press. Neter, J., & Wasserman, W. (1974). Applied linear statistical models. Homewood, IL: Richard D. Irwin, Inc. O’Connor, L. (1975). Clout: Mayor daley and his city. Chicago, IL: Henry Regnery Company. Reilly, B. (1992). It’s the reader who gets double crossed in this book. Illinois Police and Sheriff ’s News (n.

d.). Roemer, W. F., Jr. (1989). Roemer: Man against the mob. New York, NY: Ivy Books. Russo, G. (2001). The outfit. New York, NY: Bloomsbury. Scammon, R. M. (1964). America votes (vol. 5). Pittsburgh, PA: University of Pittsburgh Press. Schoenberg, R. J. (1992). Mr. Capone. New York, NY: William Morrow and Company, Inc. Theil, H. (1971). Principles of econometrics. Santa Barbara, CA: John Wiley and Sons. United States Department of Commerce. Statistical Abstract of the United States. Washington, DC: U. S.

Government Printing Office.

Springer

266 Public Choice (2007) 130:251–266

United States Department of Labor, Bureau of Labor Statistics. (1958). Union membership, 1958. Monthly Labor Review (January): 1–9.

United States Department of Labor, Bureau of Labor Statistics. (1960). Membership of American Trade Unions, 1960. Monthly Labor Review (December): 1299–1308.

Wooldridge, J. M. (2003). Introductory econometrics. Mason, OH: Southwestern.

Springer

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.