Robert T. Holden
Indiana University
 

It has often been claimed that aircraft hijacking is a "contagious" phenomenon, that the motivation to hijack aircraft spreads from one individual to another as a result of media coverage of hijacking incidents. This article develops a mathematical model of contagion and applies it to aircraft hijackings in the United States between 1968 and 1972. Analyses show that successful hijackings in the United States did generate additional hijacking attempts of the same type (either transportation or extortion). There were no contagion effects of unsuccessful hijacking attempts in the United States or any effects on U.S. hijacking attempts of such attempts outside the United States.

In this article, I will first detail various arguments about the contagiousness of hijacking. I will then review the history of hijacking with a special focus on incidents that occurred in the United States during the period 1968-72 and on the possibility that contagion was involved. A general mathematical model of contagion will be developed and applied to data on hijacking attempts in the United States during the peak period.

SOCIAL CONTAGION AND AIRCRAFT HIJACKING

The assertion that aircraft hijacking is a contagious phenomenon, that one hijacking lea& to another, appears in a wide variety of literature. Rich (1972) and Whelton (1972) write of a "skyjack virus" transmitted through the media. Phillips (1973) discusses the importance of imitation in explaining hijackings, and Chauncey (1975) and Minor (1975) consider a "hijacking as fad" hypothesis. Arey (1972) attributes several specific incidents to imitation of certain other incidents. Similar assertions are made in more general literature on "terrorism" and collective violence (e. g., Hacker 1981; Livingstone 1982; Pitcher, Hamblin, and Miller 1978) and in Hawkes's (19 7 1) discussion of a stochastic model of contagion. The cross-national diffusion of terrorist hijacking incidents is discussed by Heyman and Mickolus (1981). In another variation on the idea of contagion, Hamblin, Jacobsen, and Miller (1973) discuss hijacking in terms of the diffusion and modification of a basic invention, claiming in effect that each hijacker is attempting to outdo previous hijackers by inventing a  better hijacking. The assertion of contagiousness of hijacking also appears in newspapers (e.g., New York Times [January 28, 1972]) and magazines (e.g., U.S. News and World Report [February 17, 19681, p. 68).

The claim that publicity stimulates subsequent hijackings receives credence from studies of the effects of media coverage on other types of violent acts. Berkowitz and Macaulay (1971) report that violent crimes increased in the United States following certain wellpublicized mass murders and following the assassination of President Kennedy, Phillips (1978, 1979, 1980) and Bollen and Phillips (1982) report that various types of suicides and murder-suicides increase following other well-publicized suicides and murder-suicides, and Phillips (1983) reports that homicides increase following well-publicized championship prizefights. Mazur (1982) reports that bomb threats against nuclear energy facilities are influenced by media coverage of the nuclear energy controversy. Claims of contagiousness have also been made about larger-scale incidents of violence, including racial disturbances (Spilerman 1970), disorders in schools (Ritterband and Silberstein 1973), political violence (Hamblin et at. 1973), military coups (Li and Thompson 1975; Midlarsky 1970), and incidents of international terrorism (Hamilton and Hamilton 1983).

Berkowitz (1984) reviews much of the literature on the effects of the depiction of violence and discusses a number of ways in which such depiction may lead to violence by the viewer. The National Institute of Mental Health (1982) reviews a number of studies demonstrating that children (but not necessarily adults) imitate aggressive behavior immediately after they have seen it on television. Of particular interest is that many of the studies have shown that imitation is more likely if aggressive acts are seen to be rewarded and less likely if such acts are seen to be punished.

Two Possible Links

The use of terms such as "contagion" and "skyjack virus" is clearly metaphoric and does not explain the link between the stimulus (publicity) and the motivation of individuals to commit hijackings. I will be considering two possible links.

Mental illness and the desire for publicity. -The hypothesis that publicity stimulates hijackings is often combined with the assertion that hijackers are mentally ill. Aggarwala (1971), Hubbard (1971), Kaplan (1981), and various other authors suggest that hijackers are individuals who seek publicity because of psychological instability or low self-esteem, so publicity is in itself the motivation for the hijackings. Psychological problems are thus seen as prerequisite to the influence of media coverage.

That view is just one variation of a general view that aircraft hijackers are mentally ill.

Goal seeking.-It is not necessary to presuppose that hijackers are mentally ill or that publicity is their motive in order to predict stimulating effects of media coverage. it is possible that aircraft hijackers are behaving rationally in attempts to reach specific goals or to solve certain types of personal or political problems. For example, a substantial number of hijackers of planes from the United States to Cuba have been of Cuban birth, and hijacking is clearly a solution to the problem of returning to a home that cannot easily be reached by conventional means. Media coverage might simply serve as advertising of hijacking as a possible means of solving their problems, thus stimulating the attempts.

Operational Definition of Contagion

Up to this point the term "contagion" has been defined very loosely, and a more precise definition will be required. Hereafter, hijacking will be said to be contagious if the rate of hijacking attempts increases, however temporarily, following other hijacking attempts. This is a very undemanding definition, but it is one that can readily be applied to available data.' Although they are often linked in other literature, the hypotheses that contagion occurs, that media coverage provides the link between incidents, and that publicity is the hijackers' motive are logically independent under the present definition.

I will bring data on hijacking attempts in the United States during the peak period to bear on various hypotheses about contagion. However, in order to develop sensible hypotheses it is necessary that I first examine the nature and history of the phenomenon more closely.

TWO TYPES OF HIJACKINGS AND A BRIEF HISTORY

Typology

The importance of maintaining a distinction between two types of hijackings will be shown below.' These types can readily be distinguished on the basis of the hijackers' demands.

Hijackings for transportation. -In the majority of hijacking incidents hijackers have demanded only to be transported to a particular destination. The transportation hijacking category is very broad and includes, for example, escaping refugees (common after World War II and after the Cuban revolution), escaping criminals (including several Americans), and many others with more ambiguous motives. The majority of such attempts in the United States have been aimed at diverting planes to Cuba, although a number of other destinations have also been sought (see table 1).

Extortion hijackings.-In a sense, all hijackings involve extortion, in that the hijackers make demands that are backed up by the threat of harm to the plane, its passengers, or its crew. However, in many cases, hijackers have demanded something other than or in addition to transportation. Most extortion attempts in Europe and the Middle East have had political objectives. In the United States, the majority of extortion attempts have been in pursuit of money for personal gain, although several have involved attempts to free prisoners. The extortion category includes incidents involving both extortion (i.e., demands other than for transportation) and diversion to a particular destination because the primary motive in these cases is presumed to be other than transportation.

History

Hijacking frequencies and dates given below are based on Federal Aviation Administration (FAA) documents (1983)-' other references cited give detailed descriptions of specific incidents. Additional information on the history of aircraft hijacking can be found in scholarly papers (e.g., Aggarwala 197 1; Evans 1969; McWhinney 1.97 1; Turner 1969) and the popular press (Arey 1972; Hubbard 1971; Phillips 1973; Whelton 1972).

Aircraft hijackings prior to the late 1950s bear little relation to the later incidents in the United States.' However, the incidents that began occurring in the late 1950s are relevant. From late 1958 through 1969, aircraft hijacking was predominantly a phenomenon of the Western Hemisphere, centered on Cuba, and many of the hijackings of U.S. planes to Cuba are best understood in that larger context. Of the 177 worldwide hijacking attempts between 1958 and 1969, 80% originated in the Western Hemisphere and 77% either originated in Cuba or were efforts to divert planes to Cuba.

The incidents involving Cuba happened in several distinct phases. The first wave of hijackings began in 1958, after Fidel Castro had taken control of Cuba.' Those hijackings were primarily attempts by antiCastro individuals to divert Cuban planes to the United States. A number of Cuban boats were also hijacked to the United States." Immediately after the Bay of Pigs invasion in 1961, the direction of the hijackings reversed and there was a brief wave of diversions of U.S. planes to Cuba, many of these carried out by Cuban exiles.' The earliest hijackings of U.S. planes to Cuba might thus be regarded as an extension of the incidents that occurred during the Cuban revolution.

In the mid-1960s there were only a handful of hijacking attempts, including several attempts to and from Cuba and some isolated hijackings in Hawaii, However, as figure I shows, the hijacking rate in the United States increased dramatically in 1968 and remained high through 1972, A similar increase occurred in hijacking attempts outside the United States. There were two peaks in the rate of U.S. hijacking activity during that period, one early in 1969 and one in 1972 (see fig. 2). The first peak consisted primarily of hijackings by individuals seeking transportation to Cuba, whereas the second consisted primarily of extortion attempts.

Transportation hijackings, 1968-72. The first hijacking attempt in 1968 occurred on February 9, when a U.S. Marine attempted unsuccessfully to hijack a military charter flight out of South Vietnam. The first attempt in the United States followed shortly thereafter, on February 17, when a private plane was hijacked from Florida to Cuba. Four days later the first hijacking of a scheduled -Kirline flight to Cuba was attempted, committed by an escaping murderer. A flood of similar incidents followed. In 1968 there was a total of 19 hijackings of domestic U.S. flights to Cuba plus a hijacking of a U.S. - registered plane from the Bahamas to Cuba. There were also unsuccessful attempts to hijack flights from the United States to Mexico and to South Vietnam.

The hijacking pattern of 1968 continued in 1969, the biggest year ever for hijacking in the United States. Except for two incidents late in the year (see below), all the hijackers sought Cuba. The hijackings to Cuba during that period were largely routine. Airliners carried approach plans for the Havana airport and crews were instructed not to resist hijackers. There were also standard diplomatic procedures for obtaining the return of planes and passengers (Phillips 1973). As a result, there were no particularly dramatic hijackings to Cuba whose impact might be traceable in the subsequent history of hijacking.

The unique hijacking of 1968-69 occurred in October 1969, when an absent-without-leave Marine named Raphael Minichiello boarded a Los Angeles-San Francisco flight and hijacked it to Rome (Phillips 1973). That incident set a record for greatest distance traveled by a hijacker and received front-page publicity for several days in the United States and overseas. Ten days later there was a hijacking attempt by a 14-year-old boy in Cincinnati who, his mother claimed, had been influenced by the publicity (New York Times [November 11, 19691). The Minichiello incident seems to have influenced subsequent hijackers' choices of destinations: every hijacker who preceded Minichiello in 1969 had sought Cuba, but two who followed soon after chose destinations in Europe. However, despite the anecdotal evidence of imitation, there was no great increase in the hijacking rate following the Minichiello incident (see fig. 2).

Transportation hijacking attempts continued in the United States in the latter part of the 1968-72 period, but at a lower rate. By the end of 1971 there were far more hijackings for extortion than for transportation. Of 111 total transportation hijacking attempts between 1968 and 1972, 90 were attempts to reach Cuba (see table 1). Of these, 24 were by native Cubans and I I were by persons of other or unknown Latin American backgrounds (FAA 1983). Thus, many of the hijackers were merely seeking transportation home (see e.g., Time [March 2 2, 19681, p. 34). Because travel to Cuba was banned by the State Department, hijacking was one of the few means of transportation available (Loy 1969). 10 Another 2 3 of those who hijacked planes to Cuba were black Americans, who commited the hijackings during a period of militant civil rights activity in which leftist ideology was prominent. The white, non-Latin individuals, on whom much of the psychiatric theorizing by Hubbard (1971) and others is based, composed little more than a third of all transportation hijackers.

Although far more hijackings occurred in the United States than in any other country, hijacking was a worldwide phenomenon. As figure 3 shows, the peak years for transportation hijacking outside the United States were 1969 and 1970. Between 1968 and 1972 there were almost as many hijackings to Cuba from outside the United States as there were from within it. The majority of those incidents took place in Latin America, although there were also two attempted hijackings to Cuba from Japan and one from England. In addition to the foreign hijackings to Cuba, there were 22 attempts to hijack planes from the Soviet Union and Eastern Europe and a number of transportation hijackings in the Nliddle East, as well as many isolated incidents around the world.

Some of the foreign hijackings were publicized in the United States, although many others were not. For example, one highly publicized incident was the hijacking of a Japanese plane to North Korea, in March 1970, by nine students wielding Samurai swords (Phillips 1973). Although an increase in the U.S. hijacking rate might be expected following that incident, figure indicates that none occurred.

Extortion hijackings, 1968-72.-The first extortion attempt during the peak period occurred outside the United States. Following the Arab Israeli war of 1967 and the diversion to Algeria of a plane carrying Moise Tshombe, the Popular Front for the Liberation of Palestine (PFLP) hijacked an Israeli airliner to Algiers on July 23, 1968 (Phillips 1973). With the cooperation of the Algerian government, the hijackers demanded that Israel free a number of Palestinian guerrillas in return for the release of the passengers. That incident was followed by a series of related incidents, some of which are described below.

The first extortion hijacking in the United States, the first modern hijacking for money, occurred in June 1970. A man named Arthur Barkley held a plane hostage at Dulles Airport while demanding a ransom of $100 million from the government. He was eventually over-powered, and his plans for escape were not clear. The Barkley incident was an isolated one in that it followed the earlier foreign extortion hijackings only after a long delay and does not appear to have generated any additional extortion attempts in the United States.

For almost a year after Barkley's attempt, extortion hijackings occurred exclusively outside the United States. The most spectacular hijacking event ever was the simultaneous hijacking by the PFLP of two American planes and a Swiss plane, along with an unsuccessful attempt to hijack an Israeli plane, on September 6, 1970. Three days later a British plane was also hijacked. Three of the planes were taken to a desert air strip in Jordan, the fourth to Cairo, and all were later blown up (Arey 1972; Phillips 1973).

The PFLP hijackings apparently had repercussions overseas because, in the months that followed, Iranian, Costa Rican, and Indian planes were hijacked by groups demanding political concessions. However, it was not until the end of May 1971 that another extortion hijacking occurred in the United States. That hijacker successfully held a plane for ransom and then ordered it to fly to Nassau, Bahamas, where he was arrested. A similar unsuccessful attempt was made the following month, by an extortionist who planned to escape to North Vietnam.

It appears that up until that time there was little relationship between US. and foreign extortion hijackings. The first three U.S. extortion hijackings were the first modern attempts to hold planes for money, and those incidents did not closely follow on any foreign incidents. However, in July 1971, just a month after the third U.S. incident, a plane was hijacked for money outside the United States. That occurrence suggests that there may have been diffusion of ransom attempts from the United States to foreign countries. The next extortion hijacking in the United States, in September 197 1, was an attempt to gain the release of several Black Panthers from prison. That incident happened during the same period as several hijackings of Jordanian planes by Palestinians.

The most famous of all U.S. hijackings took place on November 24, 1971. A man who boarded a flight in Portland, Oregon, under the name of D. B. Cooper hijacked the plane, demanding a ransom of $200,000 and two parachutes. His demands were met and he later bailed out with the money. The idea of a parachute escape may not have originated with Cooper, however; a hijacker who held a Canadian plane for ransom just 12 days earlier had attempted unsuccessfully to bail out (Phillips 1973).

A wave of extortion attempts followed Cooper's (see fig. 2). In the following month there were two attempts, including one by a hijacker who demanded a parachute in an apparent attempt to imitate Cooper. Of the 31 U.S. hijacking attempts in 1972, 19 involved extortion and 15 were by hijackers who demanded parachutes.

Hijacking for money also became common overseas (see fig. 3). Three months after the Cooper hijacking, the PFLP hijacked a West German plane, released the passengers, and held the plane for a $5 million ransom. At least eight other foreign planes were hijacked for ransom in 1972, including flights from Indonesia, Ecuador, Brazil, and Australia, hijacked by individuals who apparently planned Cooper-style parachute escapes. A number of other foreign hijackers followed the PFLP's earlier pattern of demanding the release of criminals or political prisoners.

The hijacking rate in the United States began to decline in late 1972 and never again reached the high level of the period 1968-72. In the 10-year period 1973-82, there Was an average of only 9.3 hijacking attempts in the United States per year, compared with 29 attempts per year for 1968-72. Foreign hijackings also decreased after 1972, though not as sharply as U.S. hijackings.

HYPOTHESES

The literature on contagion of violence and the historical discussion above suggest several factors on which contagion effects might depend. I will express them in terms of five hypotheses, which will be tested quantitatively. The first is the basic hypothesis that contagion occurred. The history of hijacking yields a great deal of anecdotal support for this hypothesis, but some counterexamples have also been noted.

HYPOTHESIS 1: The rate of aircraft hijacking attempts in the United States increased following other hijacking attempts.

The view that hijacking was stimulated by media publicity leads to a refined version of the first hypothesis.

HYPOTHESIS 2: The rate of aircraft hijacking attempts in the United States increased following publicized hijacking attempts, but not following unpublicized attempts.

The third hypothesis is suggested by previous findings that behavior is more likely to be imitated if it is seen to be rewarded. The assumption is made that a "successful" hijacking was a "rewarded" hijacking.

HYPOTHESIS 3: Successful hijacking attempts had a greater stimulating effect than unsuccessful attempts.

The motivations for transportation and extortion hijacking attempts may be very different, and the history of hijacking indicates that the peak periods for the two types of hijackings were separated by three years. Thus, it seems reasonable to distinguish between the two types of incidents and to test the following quantitatively.

HYPOTHESIS 4: Transportation hijackings were stimulated only by prior transportation hijackings, and extortion hijackings only by prior extortion hijackings.

Many foreign hijackings were reported in the U.S. media, and therel is a possibility that cross-national diffusion occurred. Alternatively, foreign hijackings were often different from U.S. hijackings and may not have been relevant to potential hijackers in the United States. The historical evidence for imitation of foreign hijackers is inconsistent, so it is important to distinguish foreign hijackings from U.S. hijackings and to test the following prediction.

HYPOTHESIS 5: The stimulating effect on the U.S. hijacking rate was greater for hijackings from the United States than for foreign hijackings.

MATHEMATICAL MODEL

Dynamic models will be estimated for hijacking attempts disaggregated in various ways. 'rests of the various hypotheses can be obtained by comparing the estimated contagion effects of prior attempts of various types. However, it is first necessary to develop a mathematical model of contagion and appropriate statistical estimation and testing procedures.

The model developed below is a model of hijacking as a stationary stochastic process. Unlike previous models of social contagion that regard the phenomenon as transitory, with a well-defined beginning and a rate of occurrences that is either increasing or forms a bell-shaped curve over time, the present model assumes that the hijacking rate is basically constant over time but has occasional local peaks as epidemics occur and then die out. The model is thus consistent with the fact that hijacking neither began in 1968 nor ended in 1972. That the (temporally) local epidemics die out does not require any assumptions such as that there is a finite population of potential hijackers (as in Hamblin et al. 1973) or that social controls have been imposed (as in Hamilton and Hamilton 1983). The temporary nature of the epidemics follows from the stationarity of the process.

Linear Excitation Model

This section provides only a brief outline of the contagion model and its interpretation. Details of the model and of the estimation and testing procedures can be found elsewhere (Holden 1985).

The sequence of airplane hijackings over time will be regarded as a point process, a sequence of points occurring randomly on the time line (Cox and Isham 1980). The events will be modeled with a discrete time version of, the linear excitation model that Hawkes (1971) developed for orderly point processes in continuous time."

Time will be assumed to take the value t in 1, 0, 1, 2, The number of events (hijacking attempts) at time t will be denoted by N, and is assumed to have a Poisson distribution with expectation Xt = v + 6, where v is a constant (unexplained) baseline hijacking rate and 6t is an increment owing to the history of hijacking up through time t -- 1. Under the linear self-exciting model (adapted from Hawkes 1971), 6t = 1-n = W~Nt - ~, so that an event at time t - n contributes an amount Wn to the hijacking rate at time t. The effects of multiple -prior events are assumed to be additive. Because bt involves only Nt -,, Nt - 2, ..., the rate of new incidents at time t is influenced only by events that occurred during at least one time period in the past.

The lag structure ~W.J will be assumed to have the form

<insert formulas>

for some positive integer P, where 0 :5 0 < 1. The parameter u can be either positive or negative but is subject to certain restrictions (see Holden 1985). If a- > 0, the occurrence of a hijacking increases the intensity of the process above its previous level (i.e., contagion occurs). However, it is also mathematically possible to allow u < 0, so each event tends to reduce the subsequent rate of hijacking attempts.

The lag structure is such that the contagiousness of an event (or its inhibiting effect, if a < Q) increases steadily for p time periods, then declines. The specific form of the contagiousness function We simplifies many of the computations necessary in the statistical analysis (as described in Holden 1985).

A very important property of the model is that as the lag n approaches infinity the lag coefficient We, approaches zero, so the model incorporates the assumption that every incident tends to be forgotten eventually. Except for the condition v > 0, it would be possible for hijackings to die out completely even when contagion occurs. The positive lower bound for X, can be given the interpretation that hijacking is generated by exogenous factors at the constant rate v.

The contagion model is easily generalized to include effects of other types of events by writing X, = v + 80, + ... + 8,,,, where 8k1 is the contribution of the history of the kth input series of events to the hijacking rate at time to Each 8k, will be assumed to have a lag structure of the form described above. One of the inputs might be the predicted series itself, so a contagion effect is posited in addition to effects of other inputs. However, in some cases the history of the predicted series may be disaggregated into various component series, and the predicted series would then be omitted as an input (e.g., when past successful and unsuccessful hijacking attempts are used as separate predictors of the subsequent hijacking rate)

Summary Parameters

'Be parameterization of the effect of the kth input series in terms Of Ok, Ilk, and gas,, is one of mathematical convenience only. For the purpose of interpretation, several computed quantities are far more useful. The first is the long-term effect Ilk ~ I Wake, which can be interpreted as the expected number of additional hijacking attempts generated directly by each event in the Ruth input series (or the expected number prevented if at is negative).

The second summary quantity is the mean of the lag distribution, Tit- , nWk,. When the input event has a stimulating effect, tk gives the expected lag at which the generated output events occur. The mean lag may be misleading when the contagiousness function Wkn has a long tail. In that case, the median lag, Ek may be a better indicator of the delay between an input event and the output events it generates. The various summary quantities thus describe the magnitude of the stimulating or inhibiting effect of an event and the delay at which the effect occurs.

Data

In order to test the hypotheses, time series consisting of daily counts of hijacking attempts of various types between 1968 and 1972 were constucted from records obtained from the FAA (1983) (see h. 5). 1 considered it important to use data that were not aggregated over longer periods because I expected contagion effects to be measured best in days instead of in weeks or months. The FAA records contained information on the following variables of interest: I date of incident; (2) flight plan, including boarding location of hijacker(s); (3) hijacker's desired destination; (4) nature of any extortion demands; and (5) outcome (successful or unsuccessful hijacking). The FAA information about extortion demands was sometimes supplemented with information from other sources (e.g., Phillips 1973).

All incidents were first coded as either U.S. or foreign hijackings. The U.S. hijacking attempts included all incidents in which the hijacker(s) boarded the plane within any of the 50 states or Puerto Rico, since it was assumed that individuals in those areas were exposed to much of the same mass media content. There was no distinction made between scheduled airline flights and general aviation flights, and the registry of the plane was also considered immaterial. Any incident in which the hijacker boarded outside the United States and Puerto Rico was considered a foreign hijacking attempt.

The FAA recorded attempts to hijack U.S.-registered aircraft as "successful," "unsuccessful," or "incomplete," and attempts on foreign aircraft as "successful" or "unsuccessful.""' Except for the fact that the latter two categories were combined for U.S. aircraft, the FAA code was used without modification. The appropriateness of the code is discussed further below.

For the purpose of testing the dependence of contagion effects on publicity, media publicity was measured by coverage of hijacking incidents on the evening news programs of the three major U.S. television networks. Data on television coverage were obtained from two sources. For the period of August 1968 through the end of 1972 data were obtained from the Vanderbilt Television News Archive Indexes and Abstracts. If the abstracts showed any mention of a particular hijacking attempt on the evening news broadcasts of ABC, CBS, or NBC during a period from zero to three days after the event, the event was coded as "reported." Although some incidents were reported a day or more after their occurrence, no attempt was made to record the exact dates of the reports or the amounts of coverage. The Vanderbilt archives were not in operation during the first seven months of 1968. For that period, data were obtained from indexes of the "CBS Evening News" and the same coding procedure was used.

The television news data showed that about 90% of the hijacking attempts in the United States were reported on network evening news programs, and those not reported were almost exclusively attempts to hijack general aviation flights (i.e., private and commuter planes). As a result of the small number of unreported U.S. hijacking attempts, no distinction was made in the analyses between reported and unreported U.S. hijackings. However, only about one-third of all foreign hijacking attempts were reported (see table 2), so it was worthwhile to distinguish reported from unreported foreign incidents.

RESULTS

I first constructed a model to predict the overall rate of hijacking attempts in the United States. Input events included previous U.S. hijacking attempts (disaggregated by outcome) and previous foreign hijacking attempts (disaggregated by outcome and by U.S. television coverage). The results are shown in table 3.

Separate analyses were first performed for each type of input event. That is, each input series was treated as the sole predictor of U.S. hijacking attempts. The results of the single-input analyses are shown in the first three columns of table 3. The X' statistic in the third column is the likelihood ratio statistic of the null hypothesis that the counts of hijacking attempts in the United States were independently and identically distributed Poisson variates. The best predictor of hijacking attempts was previous successful hijackings from the United States. The long term effect O was estimated as 0.506 new attempts generated by each successful attempt. The mean delay t was 35.5 days, and the median E was 32.0. However, that effect did not quite reach statistical significance at the .05 level, and none of the effects of the other types of input events came close to significance.

I also estimated the effects of the various types of input event jointly, and the results are shown in the right-hand side of the table. The X 2 statistics were obtained by successively deleting input series from the model and computing the change in the overall likelihood ratio statistic. Because the various types of input event are listed in the table in the order of their theoretical importance, the order of deletion of inputs was from the bottom to the top of the list. The results of the joint analysis were very similar to the results of the single-input analyses.

Contagion Effects on Transportation Attempts

In the remaining analyses, all hijacking attempts were categorized as either transportation or extortion attempts. I first performed analyses using transportation hijacking attempts in the United States as the output series and various types of transportation hijackings as input events. Those results are shown in the top section of table 4.

When successful U.S. transportation hijackings were the only predictors of later U.S. transportation attempts, the long-term effect was estimated to be O = 0.758, and the mean time lag was t = 80 .0 days. However, because the estimated contagiousness function had a very long tail, the median lag E = 60.6 days is a better measure of the delay before additional attempts occurred. The X2 statistic of 13.972 was significant at well beyond the .05 level. The unexplained rate of U.S. transportation attempts for that model was v = .0284, so the contagion effect of previous successful transportation hijackings explained 53% of the overall level of .0608 U.S. transportation hijacking attempts per day.

The estimated single-input effect of unsuccessful U.S. transportation hijacking attempts was negative, suggesting that unsuccessful attempts tended to inhibit subsequent attempts. However, neither that effect nor the effects of any type of foreign transportation attempt were statistically significant.

Joint effects of all the types of transportation hijacking attempts were also estimated and tested in the manner described above (with inputs deleted from the bottom of the list to the top of the full model). As shown in the top right-hand part of table 4, the results of the joint estimation were very similar to the results obtained by estimating the effects separately. Thus, the only types of transportation hijacking that had a significant effect on subsequent transportation attempts in the United States were successful transportation hijackings that originated in the, -United States.

Analyses of contagion effects of extortion attempts on transportation attempts in the United States are shown in the lower half of table 4. Successful U.S. transportation hijackings were also included in the joint analysis as a control series because of the significant effect found in the previous analysis. Because reporting of foreign extortion hijackings was highly correlated with their outcomes (see table 2), it was not possible to disaggregate foreign extortions to the same extent that foreign transportation hijackings were disaggregated. Therefore, the foreign extortion attempts were disaggregated by outcome but not by television coverage. Also, the U.S. transportation hijackings did not overlap U.S. extortion hijackings temporally to an extent that allowed distinctions to be made between outcomes of U.S. extortion attempts in the analyses of transportation attempts.

When the effects were estimated separately, both U.S. extortion attempts and unsuccessful foreign extortion attempts showed apparent inhibiting effects on U.S. transportation attempts. There was an apparent reduction of 1.377 transportation attempts for each extortion attempt in the U.S. and a reduction of 2.833 U.S. transportation attempts for every unsuccessful foreign extortion attempt. However, when the effects of the three extortion input series were analyzed jointly with those of prior successful U.S. transportation hijackings, the effects vanished. Thus. none of the three types of extortion hijackings had significant effects on transportation hijacking attempts in the United States.

Contagion Effects on Extortion Attempts

Analyses of contagion effects on U.S. extortion hijacking attempts are shown in table 5. The top of the table shows analyses of the effects of prior extortion hijackings. When effects were estimated separately, both successful and unsuccessful U.S. extortion attempts showed highly significant contagion effects. The long-term effect of a successful extortion was it = 2 .014 new attempts, with a median lag of E 44.8 days. The corresponding effect of an unsuccessful attempt was O 1.412 and

O = 59.5.

The U.S. media coverage of foreign hijackings was highly correlated with their outcomes (successful foreign extortions were much more likely to be reported in the United States than unsuccessful ones, as shown in table 2), so effects could not be estimated for foreign attempts disaggregated by both outcome and media coverage. Because the disaggregation by outcome yielded more significant effects, only those results are shown. The analyses of the single-input effects of the foreign extortions showed a nonsignificant effect of successful incidents but a highly significant effect of unsuccessful incidents, with O = 1 .213 and a median lag of ~ = 39 .0 days.

The right-hand side of the top of table 5 shows the contagion effects estimated when all four types of previous extortion attempt were entered simultaneously as predictors. The value of O was reduced for each type of input event. Although both unsuccessful U.S. extortions and unsuccessful foreign extortions had produced very high X2S when used as the sole inputs, they contributed an increment of only 6.1 to the XI (with 6 df) in the joint analysis with successful U.S. extortions. Thus, when input series were deleted from the full model starting from the bottom of the list (i.e., unsuccessful foreign extortions were deleted first), the only significant effect that remained was that of successful U.S. extortions.

Contagion effects of transportation hijackings on extortion hijackings are shown in the lower portion of table 5. Successful U.S. extortion hijackings were retained in the joint analysis as a control series. Because of the lack of temporal overlap in the transportation and extortion hijackings, it was not possible to disaggregate transportation hijackings to the extent that was done in table 4.

The single-input effect of U.S. transportation hijacking attempts was negative but nonsignificant. The single-input effect of foreign transportation hijacking attempts was also negative, with O = -0.423 and ~ = 20.5 days, and was significant at beyond the .05 level. However, when the contagion effects were estimated simultaneously and the effect of successful U.S. extortions was controlled, the effect of foreign transportation attempts became positive and nonsignificant.

Thus, the only significant effect on U.S. extortion attempts that remained was that of successful U.S. extortions. For that single-input model, the unexplained rate of extortion attempts was only .002, so the contagion effect of previous successful extortion hijackings explained about 85% of the .014 attempts per day.

DISCUSSION

Because I found certain types of stimulating effects of prior hijacking attempts, I concluded that there were contagion effects on the aircraft hijacking rate in the United States between 1968 and 1972 and that the first hypothesis was confirmed. However, I must add the restriction that only successful hijackings that occurred in the United States had effects; unsuccessful attempts in the United States had no effects, nor did successful or unsuccessful hijacking attempts outside the United States.

The results of the analyses also provide tests of the other four hypotheses. I will discuss them in the reverse of the order in which they are listed above.

Tests of Hypotheses

Location of prior hijacking. -The analyses of both transportation and extortion hijacking attempts supported the hypothesis that previous hijacking attempts in the United States had a greater impact than previous foreign hijacking attempts. Although statistically significant contagion effects were found for certain types of U.S. hijacking attempts, no effects were found for any type of foreign hijacking (when the effects of U.S, hijackings were controlled). Thus, the results suggest that U.S. hijackers were generally uninfluenced by incidents outside the United States.

Perhaps the most interesting result of the analyses of the effects of foreign hijackings was the vanishing of the significant effect of unsuccess. ful foreign extortion attempts on U.S. extortion attempts when successful U.S. extortions were controlled. That result is entirely consistent with the anecdotal evidence. Although some U.S. extortion hijackers (e.g., D. B. cooper) may have imitated foreign hijackers, the evidence seems much stronger that foreign extortions were stimulated by hijackings in the United States. The first modern hijackings for money occurred in the United States, and it was only after a number of such incidents that foreign hijackers made similar attempts. It was also in the United States that parachute escapes first caught on, and that tactic was later copied by foreign hijackers (the Canadian incident that preceded D. B. Cooper's attempt was not widely publicized, and most hijackers were probably aware of only the U.S. extortions). Thus, the significant single-input effect of unsuccessful foreign extortion attempts on U.S. extortion attempts was spurious, since the causal connection ran in the other direction.

The discrepancy between the effects of successful and unsuccessful foreign extortion attempts can also be explained by referring to the history of hijacking. The successful foreign extortions during the 1968-72 period were generally politically motivated (e.g., those by Palestinian guerrillas) And preceded the period of extortions for money. Those incidents do not s to have stimulated similar incidents in the United States. The foreign attempts to extort money that followed D. B. Cooper were largely coterminous with the unsuccessful foreign extortions. Thus, there was a spurious) effect only of the unsuccessful incidents.

Type of prior hijacking.-It was confirmed that transportation hijack ings generated only transportation hijackings and extortion hijackings only extortion hijackings by the finding of contagion effects for certain hijacking attempts of the same type but none for hijackings of the opposite type. Although there were apparent inhibiting effects of the opposite type of hijacking when the effects of each input series were estimated separately, those effects were due to the lack of temporal overlap of transportation and extortion attempts and vanished when the full models which controlled for the effect of the history of the same type of hijacking) were estimated.

This result perhaps offers support to the view of hijacking as goal directed behavior. The majority of the transportation hijackings in the United States were to Cuba, and those hijackers were predominantly Latinos (often Cuban exiles) or black Americans. For them, hijacking was a means of returning home or traveling to a presumably friendlier political climate. Although there were exceptions, extortion hijackers were predominantly white and appear to have been using hijacking to achieve the more commonplace goal of getting rich, The analyses suggest ,hat hijackings of a particular type tended primarily to influence individuals with the appropriate type of motivation.

Outcome of prior hijacking.-One of the most important hypotheses theoretically was the third, which predicted that successful hijackings had a greater stimulating effect than unsuccessful hijackings. I found no effects on the U.S. hijacking rate for foreign hijackings or for U.S. hijackings of the opposite type, regardless of their outcomes. Significant contagion effects, however, were found for successful U.S. hijackings of the same type (either transportation or extortion) but not for unsuccessful hijackings. Thus, the hypothesis was supported, provided it is interpreted to refer only to hijackings of the same type and those from the United States.

This finding suggests more rationality in the hijackers than is often credited. It is consistent with the view that the hijackers were genuinely attempting to solve problems because they should have been more willing to imitate a successful act than an unsuccessful one. But the result would not have been expected if hijackers were merely publicity seekers, since most hijackings received publicity regardless of their outcomes. The result is even more at variance with Hubbard's (1971) theory that the hijackers did not really want to go to Cuba but were seeking oblivion through an act symbolic of suicide. Hubbard's views suggest that the effects of unsuccessful hijackings should have been greater than those of successful hijackings.

A caveat must be added about the meaning of a "successful" hijacking. Instead of finding a better life in Cuba, hijackers who reached that destination were usually imprisoned (Phillips 1973); thus, a successful transportation hijacking was not necessarily a successful solution to the hijacker's problems. That is equally true of extortions hijackings. Only five of the extortion attempts in the United States that followed D. B. Cooper were regarded by the FAA as successful, but even a successful hijacking was not necessarily a successful crime. Two of the bail-out attempts following Cooper's were coded as successful, but both hijackers were captured within days. The other "successes" involved two groups of hijackers who extorted ransoms and then escaped to Algeria and one group that escaped to Cuba. In all three cases, the money was returned by the foreign government, although the hijackers were not. Thus, none of the extortionists who preceded or followed Cooper actually got away with the ransom. Moreover, recent reports (New York Times [February 13, 19801) that Cooper's loot was found in the wilds of Washington State indicate that his crime was not as successful as was originally believed either.

The difference between the effects of clearly unsuccessful hijackings and apparently successful (but, in reality, unsuccessful) hijackings suggest that viewers' perceptions of the outcomes were based primarily on I the initial image presented instead of on the eventual outcome of the hijacking. Incidents in which an immediate capture appeared as part of the initial story on the hijacking were not imitated. Incidents in which capture was not immediate left the impression of success and were imitated, even if they were ultimately unsuccessful.

Media coverage and contagion. -Although it seems self-evident that media publicity is necessary for contagion to occur, one of my aims here was to demonstrate the connection statistically. The data proved inadequate, however, because coverage of hijackings in the United States was so extensive that it was of little use to perform separate analyses of the few incidents that (lid not receive national coverage.

Additional problems arose with the data on foreign hijackings. The limitations of the data prevented me from analyzing the effects of foreign extortion hijackings by U.S. media coverage. The only analyses of foreign hijackings disaggregated by U.S. news coverage (in tables 3 and 4) showed no significant effects of any type of foreign hijacking. Since there were no effects of the foreign incidents whether they were reported or not, it cannot be concluded that coverage made any difference.

An Alternative to Contagion

Consistent with the operational definition of contagion, any positive dependence (of a particular functional form) between counts of hijacking attempts in different time intervals has been attributed to contagion. However, there typically exist stochastic processes that, although not truly contagious, can generate observations that are statistically indistinguishable from those generated by a contagious process, that is, processes that generate "spurious contagion" (Taibleson 1974). More specifically, there may be some underlying stochastic process {vt} such that Nt is a Crosson variate with mean vt, where vt, is not a function of the history of {Vt} The process {NJ would then be a discrete-time doubly-stochastic Poisson process (Cox and Isham 1980). Elsewhere I derive the necessary form for a process {vt} yielding observations indistinguishable from those of the contagion process used in this article (Holden 1985).

The possibility that the hijackings were generated by exogenous events may be particularly likely, given the nature of the contagiousness funtion, that were estimated. The significant effects of successful hijacking on attempts of the same type were spread over fairly long periods of time for both types of hijackings. Such long-tailed effects might be more conistent with spurious contagion than with causal connections between eents. That is, exogenous factors may have raised the hijacking rate over certain long periods, yielding dependence between counts of events at time points within those periods, Of course, contagion and exogenous
causation are not mutually exclusive, and it is possible that both types of effects occurred.

The findings of the present research should not be dismissed because of the mere logical possibility that a completely exogenous rate process {vt}exists. However, it would be worthwhile in the future empirically to test hypotheses that specific exogenous factors produced spurious contagion. There are numerous factors that one may speculate to have influenced the hijacking rate between 1968 and 1972. For example, the peak period for transportation hijacking in the United States corresponded roughly to the peak period of U.S. involvement in Vietnam and also to a period of militant civil rights activity, and one might hypothesize that the hijacking rate was a function of those events. I have shown how the methodology of the present research can be extended for estimation of effects of input events or input time series other than hijackings so that the methodology. is sufficiently powerful to allow investigation of such exogenous factors (Holden [1985]),

Summary

I found that each successful transportation hijacking in the United States generated an average of .7 5 8 additional attempts, with a median delay of 60.6 days, and that that effect accounted for about 53% of the total rate of U.S. transportation hijacking attempts. Each successful extortion hijacking in the United States generated an average of 2.014 additional attempts, with a median delay of 44.8 days. That effect explained about 85% of the total rate of U.S. extortion hijacking attempts.

Even though it was not possible to show statistically that media coverage was responsible for the stimulating effects, the results tend to support the common belief that hijacking spread as a result of publicity. That only successful hijackings had stimulating effects is consistent with previous studies showing that violent acts are more likely to be imitated if they are seen to be rewarded. The finding is more consistent with the assumption that most hijackers were rational than that they were irrational. The general pattern of the findings lends support to the view that hijackers were seeking goals other than publicity.

Finally, I have raised the possibility that exogenous factors may have created spurious contagion among hijacking attempts. I plan to do future research on that possibility.

Attempts to divert aircraft from their scheduled flight plans through force or threats of force are familiar occurrences in the United States and throughout the world. Such incidents are referred to by a variety of terms, including "aircraft hijacking," "aerial hijacking," "air piracy," and "skyjacking."

Although there have been hijackings from 1931 to the present, their peak period was 1968-72 (see fig. 1). During that period there were 326 hijacking attempts worldwide, or one every 5.6 days. These included 137 attempts by individuals who boarded flights in the United States, or one such attempt every 13.3 days. Newspapers, television, and other mass media constantly carried stories about aircraft hijackings, and it was often suggested that the motivation to hijack planes spread from individual to individual as a result of the media coverage.