27 May 2004 Off wth ther heads: Terrorsm and electoral support for captal punshment n Australa * Snclar Davdson a, Lsa Farrell b, Clare Felvus a and Tm R. L. Fry a a School of Economcs and Fnance Royal Melbourne Insttute of Technology GPO Box 2476V Melbourne, Vctora 3001 b Department of Economcs The Unversty of Melbourne Parkvlle, Vctora 3010, Australa Abstract: Terrorst attacks such as the attacks on the World Trade Centre n September 2001 have generated new nterest n the debate on captal punshment. It has been suggested that support for the death penalty could be hgher n the wake of terrorst actvty. Usng data from the Australan Electon Study we nvestgate voters atttudes towards captal punshment. Paradoxcally, overall support for the death penalty at the 2001 Federal electon was lower than at prevous electons. In ths paper we utlse a treatment effects models to model the determnants of those atttudes and to nvestgate the mpact of terrorsm on support for the death penalty at the 2001 Federal electon. In partcular, we address the queston of whether voters who felt terrorsm was an mportant ssue had hgher levels of support for the death penalty than voters who dd not feel that terrorsm was mportant. J.E.L. Classfcaton: D72, C35 * The research n ths paper was supported by Australan Research Councl grant DP0449846: Economc reform and Australan electoral decson makng. The data used s from the Australan Electon Study. We wsh to state that those who carred out the orgnal analyss and collecton of the data bear no responsblty for our analyss or nterpretaton of the data.
1. Introducton The last person executed n Australa was hanged for murder n 1967. Between 1901 and 1967 only 114 ndvduals were executed n Australa. Of those only 21 had been executed snce 1940. Pror to Federaton n 1901, up to 80 ndvduals were executed each year. The state of New South Wales was the last jursdcton to mantan the death penalty (for treason and pracy) but abolshed captal punshment n totalty n 1985. In any event the last person hanged n that state was executed n 1940. The Commonwealth of Australa had abolshed captal punshment n ts terrtores n 1973, but no person had ever been executed n ether the Northern Terrtory or the Australan Captal Terrtory. 1 Ths type of analyss mght suggest that Australans are opposed to captal punshment. That s not the case evdence from opnon polls (McAllster, Mackerras & Brown Boldston 1997) shows that general publc n support. Thus, t appears that Australan poltcal eltes are opposed to captal punshment but that the general populaton are not. Nether of the major Australan poltcal partes supports the rentroducton of captal punshment. Recently both the Prme Mnster and the Leader of the Opposton ndcated that they dd not support the rentroducton of captal punshment. The Leader of the Opposton went further to argue that he dd not support the death penalty n Australa nor for Australans overseas. The recent publc nterest n the death penalty has been generated by terrorst actvty such as the attacks on the World Trade Centre n 2001. There are, at the tme of wrtng, two Australans held at Guantanamo Bay n Cuba awatng tral. The Australan government has ndcated that the US s free to try them but has also made t clear that executon would be unacceptable. In contrast, when the Bal bombers were sentenced to death the Australan government welcomed the decsons and declned to object or protest ndcatng that the judcal system should be allowed to take ts course. A large part of the lterature on captal punshment tends to concentrate on ssues relatng the deterrence value of captal punshment. Ths s an emprcal ssue that has generated substantal controversy. In ths paper, however, we are nterested n dentfyng the factors that determne the level of support that an ndvdual has for the 1 A bref hstory of Australan captal punshment can be found n Potas and Walker (1987). 2
death penalty. In partcular, we are nterested n nvestgatng whether an ndvduals concern about terrorsm s related to ther level of support for the death penalty. The Australan Federal Electon was held on 10 November 2001. The events of 11 September 2001 had put a spotlght on the threat of global terrorsm. There was a percepton of an ncreasng number of llegal mmgrants and asylum seekers beng smuggled nto Australa. In August 2001, the Federal government had used mltary specal forces to prevent a shp carryng Afghan refugees from enterng Australan terrtoral waters. Indeed, both McAllster (2003a; 2003b) and Edwards (2002) argued that the 2001 electon was about natonal securty and that the events of 11 September 2001 may have substantally contrbuted to the electoral outcome that returned the ncumbent Federal government. We use survey data concernng the 2001 electon to nvestgate the hypothess that ndvduals who stated that terrorsm was an mportant ssue had dfferent levels of support for captal punshment compared to those who dd not. The plan of the rest of ths paper s as follows. In secton 2 we dscuss the lkely determnants of support for the death penalty. In partcular, we hypothesse a potental lnk between support for the death penalty and concern about terrorsm. Secton 3 descrbes the survey data that we use and provdes a descrptve analyss of that data. The treatment effects model, whch s adopted to model support for the death penalty and concern about terrorsm, s outlned n secton 4. Fnally, secton 5 contans some concludng remarks. 2. Support for Captal Punshment 2.1 Background The level of support for the death penalty that an ndvdual has wll be nfluenced by a number of factors. Pre-emnent n these factors are lkely to be the socoeconomc characterstcs of the ndvdual. The socoeconomc characterstcs that we hypothesse wll nfluence support for the death penalty are ncome, educaton, gender and age. For example, hgh ncome s lkely to be assocated wth hgher support for captal punshment. That s, wealther ndvduals feel threatened by poorer and low soco-economc ndvduals and so demand captal punshment as a mechansm to mantan the status quo (Baumer, Messner & Rosenfeld 2003). 3
Educaton may also play a role n determnng the level of support for the death penalty wth hgh levels of educaton potentally beng assocated wth lower levels of support for the death penalty. A study by Whtehead and Blankenshp (2000) usng survey data from Tennessee ndcates that gender s another mportant factor. After controllng for other determnants they fnd that males are more supportve of captal punshment than females. Ths gender effect s not unformly supported. Stack (2000) usng GSS data fnds no sgnfcant gender effect. In terms of an age effect t s not clear whether the relatonshp s ncreasng, decreasng or non-lnear where older and younger ndvduals wll have lower levels of support for the death penalty. Other characterstcs that have appeared n the lterature nclude race, wth whtes more lkely to support captal punshment than blacks. We suspect ths s due to hstorcal and contemporary features of captal punshment n the US. 2 Banner (2002), for example, ndcates that those states where slavery was practced mantaned captal punshment for longer and for more crmes than dd non-slavery states. In addton, they had dfferental punshments on the statute books. At present, there s the percepton that black defendants are more lkely to be executed than whte defendants. When examnng Australan data, however, we do not expect to observe ths relatonshp. Australa has no hstory of slavery. The country was ntally establshed as a penal colony and convcted crmnals would have done the types of work normally reserved for slaves. Moreover, the racal composton of Australa s, compared to the US, farly homogenous. Thus we do not consder the mpact of ethc background n ths study. The three monothestc relgons all support captal punshment n ther wrtten scrptures but not necessarly n practce. The US evdence ndcates that evangelcal Protestant Chrstans tend to support captal punshment whereas Roman Catholcs and moderate Chrstans do not. Jews tend to not support captal punshment. Alard and Wang (2001) hghlght the paradox of Buddhsm (probably) beng opposed to captal punshment, yet t beng common n countres wth sgnfcant Buddhst 2 We also note that ths partcular racal dvde s lkely to be US-specfc. Most whte Amercans orgnate from Europe whch generally does not practce captal punshment. In contrast, most black Amercans orgnate from Afrca whch generally does practce captal punshment. 4
populatons 3. Thus an ndvduals relgous afflaton may be related to ther level of support for the death penalty. Poltcal deology wll play a role. Indvduals who consder themselves to be more conservatve are more lkely to support captal punshment than those who consder themselves more lberal or progressve. That s ndvduals who place themselves at the extreme rght are more lkely to support captal punshment. Related to ths s the potental mpact of vews on compulsory mltary servce, confdence n the armed forces, vews on sentencng n the legal system and membershp of chartable organzatons. Our fnal determnant of support for captal punshment s related to arguments concernng the mpact of the crme rate 4. In partcular, the crme rate for those crmes, potentally punshable by captal punshment. Gross and Ellsworth (2001) ndcate that we cannot expect a smple drect relatonshp between the two. Atttudes towards captal punshment are lkely to be a functon of the perceved homcde rate and, perhaps, the ntensty of that percepton. Thus a person who klls ther estranged spouse may not generate any demand for executon, whereas a mass murderer mght. In ths context vews on terrorsm may play a role as the ntroducton of the death penalty for terrorsts has been canvassed as an opton. Therefore, we hypothesse that support for captal punshment could be related to the mportance that an ndvdual places on the ssue of terrorsm. 2.2 Captal Punshment and Terrorsm. Whlst t s possble to hypothesse a lnk between support for captal punshment and an ndvduals vews on terrorsm, t should be recognsed that vews on the ssue of terrorsm themselves could be determned by a number of factors. Indeed, n the uncertan clmate of late 2001 t s probable that the formaton on vews on captal punshment and terrorsm were jontly determned. In ths secton we dscuss the 3 Alard and Wang (2001) argue that Buddhsts may not support the death penalty phlosophcally but recognse the states rght to enforce law and order. Seventeen of the 24 Buddhsts n our sample dd support captal punshment. Ths s nconsstent the envronmental acceptance argument set out n Alard and Wang (2001). 4 Homcde rates do not dffer dramatcally across regons n Australa and n the context of our (cross secton) survey data are unlkely to have an mpact. 5
factors lkely to mpact upon the mportance that an ndvdual places upon the ssue of terrorsm. Some of the socoeconomc factors that we suspect nfluence ndvduals vews on terrorsm are the same as those that we suspect nfluence ndvduals support for the death penalty. These factors are ncome, educaton, gender and age. We do not, however, expect relgous afflaton to be an mportant determnant n ths case. In addton to socoeconomc characterstcs, other lkely determnants of ndvduals vews on terrorsm nclude poltcal deology and conservatsm, confdence n the mltary, and confdence n the legal system. Whlst confdence n the legal system s also antcpated to play a role n determnng the level of support for the death penalty, t wll have a dfferent mpact on concern about terrorsm. Increased confdence n the legal system wll lead to greater support for the death penalty, but lower levels of concern about terrorsm. Terrorsm s often vewed as a threat to natons and as such we hypothesse that the level of attachment that an ndvdual has to the culture and nsttutons of the country wll nfluence ther concern about terrorsm. Jones (1997) nvestgated ssues of natonal dentty n Australa. He ntroduced and valdated three contnuous scale varables to represent natonal dentty. These varables relate to Australan natvsm (country of brth, long resdence and beng Chrstan), affectve cvc culture (respect for Australan laws and nsttutons and feelng Australan) and nstrumental cvc culture (ctzenshp and Englsh-language competence). He further developed a typology based upon the frst two of these varables that Charnock (2001) found sgnfcant n determnng support for Australa becomng a republc. We hypothesse that the natvsm and nstrumental cvc culture dmensons of natonal dentty, n partcular, wll mpact concern about terrorsm. We expect that ndvduals who score hghly on these dmensons wll be more concerned about terrorsm. 3. Data 3.1 The Australan Electoral Study (AES) Ths paper uses survey data pertanng to the 2001 Federal Electon to clarfy the mpact of varous nfluences upon voter behavour economcs or ssues. Our 6
analyss of the 2001 electon s based on the Australan Electoral Study (AES) 5 (Bean, Gow & McAllster 2002). 6 The AES s conducted after every Federal Electon and surveyng for the 2001 electon occurred between 12 November 2001 and 5 Aprl 2002. The sample s drawn from the electoral roll whch, gven Australa s mandatory votng regme, s kept up-to-date and s relable. In total there are 2010 cases (ndvdual voters) and 379 varables per case n the fnal data set. It should be noted that, unlke the US, Australa does not operate a fxed term electoral cycle. The lfe of any Federal parlament s three years but, generally, the tmng of an electon s at the dscreton of the Prme Mnster. We have data for the 1987, 1990, 1993, 1996, 1998 and 2001 Federal Electons. 7 Although the emphass n ths paper s on the determnants of support for captal punshment as recorded n the 2001 survey, we begn by lookng at the levels of support for the death penalty found n the sx surveys. Fgure One ndcates the level of support for the death penalty n Australa. Fgure 1: Support for the Death Penalty. 75 65 Percent (%) 55 45 35 25 15 1987 1990 1993 1996 1998 2001 Year Total support Strongly support Support Source: AES 1987, 1990, 1993, 1996, 1998 & 2001. 5 The orgnators of the AES data bear no responsblty for our analyss or nterpretaton of the data. 6 A full descrpton can be found at http://assda.anu.edu.au/codebooks/aes2001/ttle.html 7 In 1987 respondents were asked whether they agreed wth the statement, Brng back the death penalty, whereas n all other years the statement read, The death penalty should be rentroduced for murder. 7
We have broken up the data by those who Strongly Agree and Agree wth rentroducng the death penalty. We also show a total of the two. Qute clearly over ths perod a majorty of Australans supported the rentroducton of captal punshment. What s of nterest s that support for the death penalty has fallen n the 2001 electon to 56.5 percent. Our subsequent analyss s based upon the determnants of support for captal punshment n 2001 and, n partcular, any relatonshp between concern about terrorsm and support for captal punshment. The 2001 survey s the only one that asked respondents about terrorsm. Thus the analyss n ths paper cannot address the queston why the level of support fell n 2001 relatve to that at prevous electon tmes. 3.2 The 2001 Australan Electon Study The appendx to ths paper detals the constructon of the varables that represent the factors dscussed above. As s typcal wth survey data most of these varables are dscrete n nature. The exceptons are support for the death penalty, famly ncome, and support for compulsory mltary servce. These three varables are beng treated as contnuous. The other non-dscrete varables are own-left rght poston and the natvsm varables, whch are all standardsed (Z score) varables. 8 We begn by presentng descrptve statstcs for these varables n Table 1. 8 Standardsed varables were constructed on the assocated complete non-mssng sample and so may not have means of zero and standard devatons of one n the estmaton sample. That s, mssng data s dropped rather than mputed n our analyss. 8
Table 1: Descrptve Statstcs Varable Mean Standard Devaton Mnmum Maxmum Support for the death penalty 3.360 1.396 1 5 Terrorsm extremely mportant ssue 0.470 0.499 0 1 Male 0.477 0.500 0 1 Age 45.510 15.087 17 89 Income 5.647 3.142 1 11 Tertary educaton 0.253 0.435 0 1 Catholc 0.276 0.447 0 1 Anglcan 0.243 0.429 0 1 Untng 0.081 0.274 0 1 Orthodox 0.031 0.173 0 1 Presbyteran 0.038 0.191 0 1 Other Chrstan 0.069 0.253 0 1 Other relgon 0.041 0.198 0 1 Charty membershp 0.231 0.422 0 1 Gun ownershp 0.108 0.310 0 1 Support for compulsory mltary servce 2.850 1.284 1 5 Own left-rght -0.047 0.980-2.677 2.370 Strongly agree law breakers stffer sentences 0.298 0.457 0 1 Agree law breakers stffer sentences 0.411 0.492 0 1 Great deal of confdence n legal system 0.046 0.210 0 1 Qute a lot of confdence n legal system 0.323 0.468 0 1 Not very much confdence n legal system 0.523 0.500 0 1 Great deal of confdence n armed forces 0.244 0.430 0 1 Qute a lot of confdence n armed forces 0.609 0.488 0 1 Natvsm -0.121 0.945-2.045 1.784 Instrumental culture -0.028 0.987-4.046 0.869 We see that support for the death penalty s such that the estmaton sample appears neutral on ts re-ntroducton. Some 47% of the sample stated that terrorsm was an extremely mportant ssue n the electon. 9
4. Estmatng the Relatonshps 4.1 The Treatments Effects Model A smple regresson model that relates support for the rentroducton of the death penalty for murder by ndvdual, y, to the factors n our model (e.g. gun ownershp), x, s: ' y = x β + δt + u, (4.1) where T s the dummy varable ndcatng whether or not the ndvdual specfed that terrorsm was an extremely mportant ssue; T = 1 f ndvdual beleved terrorsm to be an extremely mportant ssue. The problem les n correctly estmatng δ, the coeffcent that measures the mpact of concern about terrorsm on support for the rentroducton of the death penalty. There may be a correlaton between an ndvdual s tendency to be concerned about terrorsm, and the extent to whch they support/oppose the death penalty. For example, an ndvdual who s n favour of the death penalty may also be very concerned about terrorsm because they wsh the perpetrators of terrorsm be brought to justce. Ths s a problem of self-selecton because t depends upon the ndvdual as to whether or not they are extremely concerned about terrorsm. Ths self-selecton wll result n a based estmate of δ (Greene 2000, pp. 933-934). A better way to model would be to use a treatment effects regresson model (Wooldrdge 2002). In ths model the death penalty equaton gven n (4.1) above s augmented wth a second equaton that captures the determnaton of the factors assocated wth an ndvdual s concern about terrorsm. The second component nvolves modellng extreme concern about terrorsm usng a bnary Probt formulaton. The propensty to be concerned about terrorsm, ' * T, s gven by: T * = z γ + v (4.2) An ndvdual s extremely concerned about terrorsm ( T = 1) f large. Thus, * T s suffcently T 1 = 0 f f T T * * 0 < 0 10
A key aspect of ths model s that the stochastc components n (4.1) and (4.2) wll be correlated wth correlaton equal to ρ. The treatment effects model can be estmated by maxmum lkelhood (ML) methods under the assumpton of jont normalty of u and v 9. In estmatng the model t s hypothessed that support for the death penalty (for murder) may be determned the factors dscussed n secton 2.1 above and that concern about terrorsm s hypothessed to depend upon the factors n secton 2.2 above. The next secton dscusses the results of estmatng the model. 4.2 Results We begn by dscussng the overall ft of the model n ths secton. Secton 4.3 then dscusses selected results on the ndvdual factors found sgnfcant n the model and ther mpacts upon the two outcomes n the model (support for captal punshment and mportance of terrorsm). Table 2 presents the maxmum lkelhood estmates of the parameters of the treatment effects model. 9 We use Stata, Verson 8 to estmate the model by Maxmum Lkelhood. 11
Table 2: The estmated treatment effects model. Coeffcent T-rato Death Penalty Equaton Constant 1.9452 5.66 Male 0.2554 3.40 Age 0.0094 0.77 Age squared -0.0002-1.80 Income 0.0003 0.02 Tertary educaton -0.1980-2.08 Catholc -0.0309-0.32 Anglcan 0.1604 1.58 Untng 0.1894 1.37 Orthodox 0.0395 0.20 Presbyteran 0.2010 1.08 Other Chrstan -0.0413-0.29 Other relgon 0.2040 1.16 Charty membershp -0.1661-1.86 Gun ownershp 0.2976 2.79 Support for compulsory mltary servce 0.1779 4.95 Own left-rght 0.0772 1.73 Strongly agree law breakers stffer sentences 1.1117 11.63 Agree law breakers stffer sentences 0.5844 7.00 Great deal of confdence n legal system -0.4805-2.34 Qute a lot of confdence n legal system -0.4011-3.03 Not very much confdence n legal system -0.2353-1.94 Terrorsm an extremely mportant ssue 1.2447 4.47 Wald Test 660.41 22 d.f. Terrorsm an extremely mportant ssue (Y/N)-Probt Constant 0.0169 0.07 Male -0.2243-2.75 Age 0.0032 1.12 Income -0.0496-3.49 Tertary educaton -0.2228-2.13 Charty membershp -0.2160-2.16 Support for compulsory mltary servce 0.1252 3.53 Own left-rght 0.2169 4.80 Great deal of confdence n armed forces 0.2831 2.14 Qute a lot of confdence n armed forces 0.0644 0.57 Great deal of confdence n legal system -0.5922-2.40 Qute a lot of confdence n legal system -0.2753-1.87 Not very much confdence n legal system -0.1318-0.97 Natvsm 0.2120 4.47 Instrumental culture 0.1070 2.44 ρ -0.5414 4.56 σ 1.1752 21.1 Log- Lkelhood -2293.656 B.I.C. 4866.141 Sample sze 1105 12
The model s well determned wth reasonable levels of statstcal sgnfcance assocated wth the ncluded varables. The varables n the death penalty equaton are jontly sgnfcant (the Wald test wth 22 degrees of freedom beng the large sample verson of an overall F-test n a least squares regresson). Moreover, the probt equaton for the mportance of terrorsm also performs well. Table 3 presents the classfcaton accuracy of ths component of the treatment effects model. Table 3: Classfcaton Accuracy of the Probt equaton for Terrorsm Actual Predcton 0 1 Tota1 % Correct 0 411 175 586 70.14 1 185 334 519 64.36 Total 596 509 1105 67.42 The value of the Franses test statstc of 10.709 ndcates that the model has sgnfcantly better classfcaton accuracy than random assgnment (Franses 2000). Ths shows that the model does well n predctng the mportance of terrorsm. Ceters parbus, the estmated model tells us that an ndvdual who vews terrorsm as an extremely mportant ssue wll have a hgher level of support for captal punshment. The estmated mpact s +1.245 that s equal to an mpact of 24.9% (=1.245/5). As a fnal test of the relablty of the estmate of ths treatment effect we estmated the model n two parts. Frst we estmated the probt model for mportance of terrorsm and generated the predcted probabltes. These predcted probabltes were then used n place of the ndcator varable for terrorsm n a regresson model for support for the death penalty. Wooldrdge (2002) shows that ths gves a robust estmate of the treatment effect. Ths estmaton confrmed the magntude of the treatment effect 10. 4.3 Dscusson We now turn to dscuss the mpacts of some of the factors on the level of support for the death penalty and also on the mportance of terrorsm. Specfcally, we wll consder the mpact of changes n the level of confdence n the legal system, famly 10 Full results avalable on request. 13
ncome and the left-rght postonng of an ndvdual on support for the death penalty 11 and the mportance of terrorsm. Fgure 2 depcts the mpact of changng levels of confdence n the legal system. Both the level of support for the death penalty and concern about terrorsm fall off wth ncreasng confdence n the legal system. It s predcted that those who have no confdence at all n the legal system are most lkely to be concerned about terrorsm. Indeed, only those who have no confdence n the legal system are predcted to be extremely concerned about terrorsm. Such ndvduals also have the hghest level of support for the death penalty. Fgure 2: The effect of confdence n the legal system on concern about terrorsm and support for the death penalty Predcted probablty terrorsm extremely mportant 0.55 0.50 0.45 0.40 0.35 0.30 0.25 none not very much qute a lot a great deal level of confdence 4.50 4.00 3.50 3.00 2.50 Predcted support for the death penalty Terrorsm (predcted) Death penalty (predcted) Fgure 3 depcts the mpact of famly ncome 12. Ceters parbus, famles wth ncomes below Aus$30,000 are extremely concerned about terrorsm. There s a slght postve relatonshp between ncome and support for death penalty, wth a large downward movement at about the same ncome band. Those wth ncomes below $30,000 support death penalty and those above oppose t. 11 Support for the death penalty s represented on a fve-pont scale where 1 = strongly dsagree and 5= strongly agree. 12 The ncome categores are n $10, 000 bands. Income category 1 = less than $10,000, 2 = $10,001 to $20,000 11= greater than $100,000 14
Fgure 3: The effect of ncome on concern about terrorsm and support for the death penalty. 0.60 4.50 Predcted probablty terrorsm extremely mportant 0.55 0.50 0.45 0.40 0.35 0.30 1 2 3 4 5 6 7 8 9 10 11 Income category 4.00 3.50 3.00 2.50 Predcted support for death penalty Terrorsm (predcted) Death penalty (predcted) Fgure 4 presents the relatonshp between an ndvduals left-rght poston and ther concern about terrorsm and also ther level of support for the death penalty. We see that as an ndvdual moves towards the rght they are more lkely to be concerned about terrorsm. Once an ndvdual s rght of centre they are predcted to be extremely concerned about terrorsm. In terms of support for the death penalty, those to the left of ths pont oppose the death penalty and those at ths pont or further to the rght support death penalty. 15
Fgure 4: The effect of poston on the left-rght poltcal spectrum on concern about terrorsm and support for the death penalty Predcted probablty terrorsm extremely mportant 0.7 0.6 0.5 0.4 0.3 0.2-2.68-2.17-1.67-1.16-0.66-0.15 0.35 0.86 1.36 1.87 2.37 Standardsed own left-rght poston Terrorsm (predcted) Death penalty (predcted) 4.50 4.00 3.50 3.00 2.50 Predcted support for death penalty Our fnal fgures, fgures 5 and 6, llustrate the predctons of the treatment effects model for stylsed ndvduals. We stylsed ndvduals for each of the poltcal partes and for the whole estmaton sample, and then compared model predctons wth actual outcomes. Usng the stylsed Lberal supporter as an example, the observed probablty that terrorsm s an extremely mportant ssue for ths person s equal to the proporton of ndvduals wthn the sub-sample of Lberal supporters who beleved terrorsm was an extremely mportant ssue. For the same stylsed ndvdual, the predcted probablty was estmated by settng the explanatory factors n the model to the mean values calculated over the sub-sample of Lberal supporters only. The same process was repeated for the remanng partes, wth the results shown n Fgure 5 below. Fgure 6 compares the observed and actual levels of support for the death penalty for stylsed ndvduals. 16
Fgure 5: Model predctons of concern about terrorsm by party. Probablty that terrorsm extremely mportant 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 Sample Lberal Labor Natonal Democrats Greens One Naton Party Identfcaton Observed Predcted Fgure 6: Model predctons of support for the death penalty by party. 5.00 4.50 Support for death penalty 4.00 3.50 3.00 2.50 2.00 1.50 1.00 Sample Lberal Labor Natonal Democrats Greens One Naton Party Identfcaton Observed Predcted These fgures confrm that the model provdes a good ft of the data overall at sample means. They also ndcate that the model correctly predcts that Lberal, Natonal and One Naton supporters are extremely concerned about terrorsm. The death penalty component of the model does not perform as well as that for mportance of terrorsm but stll performs well. 17
5. Concluson In ths paper we use survey data concernng the 2001 electon to nvestgate the hypothess that ndvduals who stated that terrorsm was an mportant ssue had dfferent levels of support for captal punshment than those who dd not. The survey data pertan to ndvduals and ther vews on a range of ssues at the tme of the Australan Federal Electon of 10 November 2001. At ths tme, the events of 11 September 2001 had put a spotlght on the threat of global terrorsm. Usng a treatment effects model we fnd evdence that, ceters parbus, ndvduals who thought that terrorsm was an extremely mportant ssue had a 24.9% hgher level of support for the re-ntroducton of captal punshment. 18
6. References Alard, L. and Wang, H.M. (2001). Mercy and punshment: Buddhsm and the death penalty. Socal Justce, 28(1), 231-247. Banner, S. (2002). The death penalty: An Amercan hstory. Cambrdge: Harvard Unversty Press. Baumer, E., Messner, S. and Rosenfeld, R. (2003). Explanng spatal varaton n support for captal punshment: A multlevel analyss. The Amercan Journal of Socology, 108(4), 844-875. Bean, C., Gow, D. and McAllster, I. (2002). Australan Electon Study, 2001 [computer fle]. Canberra: Socal Scence Data Archves, The Australan Natonal Unversty. Charnock, D. (2001). "Natonal dentty, partsanshp and populst protest as factors n the 1999 Australan republc referendum." Australan Journal of Poltcal Scence, 36(2), 271-291. Edwards, L. (2002). How to argue wth an economst: Reopenng poltcal debate n Australa. New York: Cambrdge Unversty Press. Franses, P.H. (2000). A test for the ht rate n bnary response models. Internatonal Journal of Market Research, 42(2), 239-245. Greene, W.H. (2000). Econometrc analyss, 4th Edton. New Jersey: Prentce Hall. Gross, S. and Ellsworth, P. (2001). Second thoughts: Amercans vews on the death penalty at the turn of the century. In: Garvey, S.P, ed. Captal punshment and the Amercan future. Duke Unversty Press. Avalable from: http://papers.ssrn.com/paper.taf?abstract_d=264018. Jones, F.L. (1997). "Ethnc dversty and natonal dentty." Australan and New Zealand Journal of Socology, 33(3), 285-305. McAllster, I. (2003a). "The federal electon n Australa, November 2001." Electoral Studes, 22(2), 381-387. McAllster, I. (2003b). Border protecton, the 2001 Australan Electon and the Coalton vctory. Australan Journal of Poltcal Scence, 38(3), 445-463. McAllster, I., Mackerras, M. and Brown Boldston, C. (1997). Australan poltcal facts, 2 nd Edton. South Melbourne: Macmllan. Potas, I. and Walker, J. (1987). Captal punshment. No. 3 n Trends and ssues n crme and crmnal justce. Australan Insttute of Crmnology, February 1987. Avalable from: http://www.ac.gov.au/publcatons/tand/t03.pdf Stack, S. (2000). Support for the death penalty: A gender-specfc model. Sex Roles, 43(3), 163-179. Whtehead, J. and Blankenshp, M. (2000). The gender gap n captal punshment atttudes: An analyss of support and opposton. Amercan Journal of Crmnal Justce, 25(1), 1-13. Wooldrdge, J.M. (2002). Econometrc analyss of cross secton and panel data. Cambrdge: MIT Press. 19
Appendx: Varable Defntons Ths appendx detals the defntons of the varables used n the analyss and ther relatonshp wth the orgnal 2001 AES varables. The dependent varable: Deathp: A fve-pont scale measurng agreement/dsagreement wth the statement, the death penalty should be rentroduced for murder. 1= strongly dsagree, 5= strongly dsagree. [E4deathp]. The treatment varable: Terror: A dummy varable that ndcates that terrorsm was an extremely mportant ssue n a respondent s votng decson. 1= extremely mportant ; 0= qute mportant, or not very mportant. [D1terror] Demographcs: Male: Ths varable ndcates gender. 1= male; 0= female. (I1) Age: Ths varable s the age (n years) of the respondent. (I2) Agesq: Agesq s the square of age. Income: The codng of Income, an 11-value varable that s treated as beng contnuous, s shown below. It s the response to the queston: What s the gross annual ncome, before tax or other deductons, for you and your famly lvng wth you from all sources? Please nclude any pensons and allowances, and ncome from nterest or dvdends. (I17) Value Income Band 1 Less than $10,000 per year 2 $10,001 to $20,000 per year 3 $20,001 to $30,000 per year 4 $30,001 to $40,000 per year 5 $40,001 to $50,000 per year 6 $50,001 to $60,000 per year 7 $60,001 to $70,000 per year 8 $70,001 to $80,000 per year 9 $80,001 to $90,000 per year 10 $90,001 to $100,000 per year 11 Greater than $100,000 per year 20
Tertary: Ths varable ndcates that the respondent has tertary qualfcatons. 1= Postgraduate degree or Postgraduate dploma, or Bachelor degree (ncludng honours ; 0= No qualfcatons snce leavng school, Undergraduate dploma, Assocate dploma, Trade qualfcaton, or Non-trade qualfcaton. (H3) Relgon dummes: A set of dummy varables representng the respondents relgon was created from the extended relgon varable, x5. The recodng was as follows: Cath: Ths varable ndcates that the respondent s a member of the Catholc Church. 1= Catholc-Roman, or Catholc-not Roman ; 0= otherwse (wth the excepton of mssng data). Anglcan: Ths varable ndcates that the respondent s a member of the Anglcan Church. 1= Church of England; 0= otherwse (wth the excluson of mssng data). Untng: Ths varable ndcates that the respondent s a member of the Untng Church. 1= Untng or Methodst; 0= otherwse (wth the excluson of mssng data). Orthdx: Ths varable ndcates that the respondent s a member of the Orthodox Church. 1= Orthodox; 0= otherwse (wth the excluson of mssng data). Presby: Ths varable ndcates that the respondent s a member of the Presbyteran Church. 1= Presbyteran; 0= otherwse (wth the excluson of mssng data). Othchrst: Ths varable ndcates that the respondent s a member of another Chrstan Church. 1= Baptst, Brethren, Churches of Chrst, Jehovah s Wtness, Latter Day Sants, Lutheran, Salvaton Army, Seventh Say Adventst, Other Protestant, or Other Chrstan ; 0= otherwse (wth the excluson of mssng data). Othrel: Ths varable ndcates that the respondent s relgon s not Chrstan. 1= Buddhst, Hebrew/Jewsh, Muslm, Other Non-Chrstan, Hndu, Other (Not Specfed). Charty: A dummy varable for membershp of a chartable organsaton: 1= actve member or nactve member ; 0= don t belong. (I14P4) 21
Gun: Indcates that a respondent, or someone n hs or her household owns a frearm. 1= yes, 0= no. (I9) Poltcal poston, poltcal ssues and nsttutons: Zb10own: A standardsed varable (Z score) for the respondent s own left-rght poston on the left-rght poltcal spectrum. Low values = left; hgh values= rght. (B10own) Ordnary: A fve-pont scale reflectng how well Federal poltcans understand ordnary Australans. 1= Federal poltcans don t know what ordnary people thnk ; 5= Federal poltcans know what ordnary people thnk. (C10) Lawb_sa: A dummy varable for those who strongly agree wth the statement, People who break the law should be gven stffer sentences. 1= strongly agree ; 0= agree, nether agree nor dsagree, dsagree, or strongly dsagree. (E4lawbrk) Lawb_a: A dummy varable for those who agree wth the statement, People who break the law should be gven stffer sentences. 1= agree ; 0= strongly agree, nether agree nor dsagree, dsagree, or strongly dsagree. (E4lawbrk) Conscrpt: A varable that reflects agreement wth the statement, Australa should have compulsory mltary servce. 1= strongly dsagree; 5= strongly agree. (E4mltr) Confdence n Australan organsatons: Armed_g: Ths dummy varable ndcates those who have a great deal of confdence n the armed forces. 1= A great deal of confdence ; 0= Qute a lot of confdence, Not very much confdence, or None at all. (G4P1) Armed_q: Ths dummy varable ndcates those who have a qute a lot of confdence n the armed forces. 1= Qute a lot of confdence ; 0= A great deal of confdence, Not very much confdence, or None at all. (G4P1) Legal_g: Ths dummy varable ndcates those who have a great deal of confdence n the legal system. 1= A great deal of confdence ; 0= Qute a lot of confdence, Not very much confdence, or None at all. (G4P2) 22
Legal_q: Ths dummy varable ndcates those who have a qute a lot of confdence n the legal system. 1= Qute a lot of confdence ; 0= A great deal of confdence, Not very much confdence, or None at all. (G4P2) Legal_nv: Ths dummy varable ndcates those who do not have very much confdence n the legal system. 1= Not very much confdence ; 0= A great deal of confdence, Qute a lot of confdence, or None at all. (G4P2) Natvsm varables Znatv1: Standardsed constructed varable representng natvsm. natv = G6P1 (born Australa) + G6P3 (Lve Australa) + G6P5 (beng Chrstan) Ths varable was standardsed and then multpled by 1 so that low values = not at all mportant, and hgh values= very mportant. Zaffcult1: Standardsed constructed varable representng affectve cvc culture. Affcult = G6P6 (Respect Laws) + G6P7 (Feelng Australan) Ths varable was standardsed and then multpled by 1 so that low values = not at all mportant, and hgh values= very mportant. Znstcult1: Standardsed constructed varable representng nstrumental cvc culture. nstcult = G6P2 (Australan ctzen) + G6P4 (Speak Englsh) Ths varable was standardsed and then multpled by 1 so that low values = not at all mportant, and hgh values= very mportant. 23