Effects of Interview Mode on Evaluation of the Police Volker Hüfken Department of Sociology, University Düsseldorf 1. Introduction Attitudes towards the police in Europe and the US, be it of victims or of the population in general, is rather positive. For example Hindelang (1974,1981) reports for the USA that around 60% of the respondents have a positive view of the police s work. Stephan (1976) reports similar findings in the 80ies for Western Germany, and the latest studies from the Mid-Nineties for reunified Germany also come to that conclusion (Dörmann 1996, Forschungsgruppe Kommunale Kriminalprävention in Baden-Württemberg 1996). It s true for Germany as well that victims and those groups have a positive attitude towards the police who feel subjectively more victimized. According to these findings, the police are held in higher esteem by older people and members of the lower classes than by younger or middle class people respectively (Schwind et al. 2001). Only from the Nineties onwards are there more frequent studies on the evaluation of the police s work for Germany. No victim study that uses different modes of interview in order to estimate their influence on evaluation of the police s work has been carried out for Germany so far. If certain groups, for example victims of a specific criminal offence, are interviewed in victimization studies instead of the usual interview of a cross-section of society, the question arises in how far the mode-effect influences the findings. Lamnek (1986) could show in his study that the evaluation of police behavior by the victims varies in relation to the gravity of the criminal offence. Victims of comparatively petty offences like fraud and offences against or damage of property feel they are treated worse by the police than victims of bodily harm, rape or burglary. One explanation for the influence of the type of criminal offence could be that in case of material damage or loss the insurance companies can compensate, at least to an extent, the victim for the offence. In the case of burglary, especially in private residences, the shelter in which the 1
victim thought himself safe is violated. That s why Kaiser (1997) points out: The more significant the victim feels the criminal breach to be, the deeper is the wish for punishment, to which the wish for poetic justice and for prevention of further criminal offences are added. Especially from the perspective of the two groups of offences, burglary and persons involved in road accidents, effects of selection cannot be excluded. In the group persons involved in road accidents not only the victim but also the person who caused the accident was interviewed. One has to take into account, though, that the mere fact that people who had been involved in an accident participated in the interview in the first place, may well have influenced their answers. Their statements probably turned out more positive because they wanted to exclude negative consequences for their lawsuit. 2. Method The mail survey took place in October 98. The telephone survey was carried out between January and March 1999. For the survey, all cases from the police register from the 1. June 98 till the time of the survey were chosen. In the present survey, 1800 numbers were selected for the telephone survey, of which 1283 people were interviewed after a maximum of 5 contacts. This is a rate of 71 %. 3496 addresses were used for the mail survey, of which 1204 people answered (response-rate about 34 %). In comparison to the usual response-rate in the general population, which is between 40 and 60 %, the rate of 71 % can be regarded as high. The response-rate of the mail survey is similar to that of surveys in which no reminder action has taken place. Nevertheless, rates of up to 70 % are possible in general population surveys (Hippler 1985, Reuband and Blasius 1996, Reuband 1999,2001). Small response-rates wouldn t be a problem if there wasn t a difference between both groups. If the groups differ systematically, however, regarding one or more relevant variables, one assumes that these differences exist even more between those who answer directly and those who answer later, after several reminder actions (Binder et al. 1979, Kish 1965, Reuband 2001, Zimmer 1956). That s why people with little interest in the survey or with little education are underrepresented in mail surveys. If, in addition to that, education correlates with the evaluation of the police, there can occur a bias for the in each case relevant variables. Regarding the satisfaction with the police, we assume the quality of data to be 2
better in the mail survey (Hindelang et al. 1981; Kury 1993). People will probably be more prepared to express criticism of the addressee in written form. In the following, the question is raised how the mode of survey influences the structure of the sample. And is the police evaluated differently in general and with regard to a specific event according to the interview mode? Which effects are to be expected from both kinds of criminal offences? 3. Results Table 1: Socio-demographic and socio-economic variables by Mode (in %) phone (n=1275) mail (n=1201) total (n=2478) Cramérs V sex n.s female 38 35 36 male 62 65 64 age n.s 14 17 years - 1-18 25 years 13 12 13 26 35 years 20 23 22 36 45 years 23 22 23 46 55 years 17 17 17 56 65 years 15 15 15 66 75 years 8 7 8 76 years and older 2 3 3 education.06* no degree 2 2 2 Haupt- /Volksschule 47 42 45 Realschule 26 27 26 Fachhochschulreife 16 17 16 Abitur 9 13 11 Employment.10*** self ployed 14 19 16 ployee 47 45 46 pupil 8 8 10 housewife 11 8 8 pensioner 18 15 17 unemployed 1 2 1 others 1 2 7 nationality n.s german 97 95 96 other Nationality 3 5 4 *p<.05, **p<.01, ***p<.001 3
Table 1 shows, as was to be expected, little effects of educational and professional groups. The formally less educated, Hauptschulabschluss, are less represented in the mail survey than in the telephone survey. And those with a formally higher education are more represented in the mail survey than in the telephone survey. Because the self-employed are hard to contact, they are more likely to take part in mail than in telephone surveys. The quota of house-wives and retired people is, however, higher in the telephone survey. Table 2: Socio-demographic, socio-economic variables and delictgroups by Mode (in %) Persons involved in road accident phone mail Cramérs V phone burglary (n=757) (n=756) (n=757) (n=756) sex.00 n.s.09** female 33 34 46 36 male 67 66 54 64 mail Cramérs V age.09 n.s.11 n.s 14 17 years - 1 - - 18 25 years 20 18 3 3 26 35 years 25 26 13 20 36 45 years 23 19 24 26 46 55 years 14 15 22 20 56 65 years 12 13 20 18 66 75 years 5 6 14 9 76 years and older 1 2 4 4 education.06 n.s.07* no degree 2 1 2 2 Haupt- /Volksschule 43 38 53 47 Realschule 28 28 24 25 Fachhochschulreife 18 19 13 14 Abitur 11 14 8 12 employment.05 n.s.20*** self-ployed 9 10 21 35 ployee 55 52 35 34 pupil 13 13 2 1 housewife 9 8 15 9 pensioner 12 13 25 17 unemployed 1 2 1 3 others 2 3 1 2 nationality.00 n.s.09* german 96 96 98 94 other Nationality 4 4 2 6 *p<.05, **p<.01, ***p<.001 4
We can see that both groups of victims of criminal offences are composed differently according to social features. We don t have to take that fact into account for the comparison of mode. It will be needed later, however, when we compare both groups of victims regarding the effect of the modes on evaluation of the police. As shown in table 2, the effect of mode is weak in both victim groups. You can see the effect caused by education in both groups, though. Especially the victims of burglary are influenced by the mode effect with regard to the social composition. The quota of women is higher in the telephone than in the mail survey. The quota of the self-employed is conspicuously higher in the mail survey. If there are differences to a higher degree (10 % or more), the people interviewed on the phone are the ones who show higher satisfaction rates (see table 3). 66 % state that they have a positive opinion on the police in North-Rhine Westphalia (value on scale 1+2), in the mail survey, 55 % are of that opinion. Police in the interviewee s place of residence are evaluated in a similarly positive manner, 68 % express satisfaction on the phone, 57 % by mail. The effect of mode is less significant in the case of questions regarding the evaluation of police behavior in connection to the incident on record. As opposed to a general evaluation of police in NRW, the interviewee has an actual experience which he uses to evaluate the police. With the exception of few items, the telephone survey shows the higher results, independent of the respective victim group. If you compare the eta-coefficients for all items in both victim groups, you find that the effects of mode are a lot stronger with the victims of burglary than with the persons involved in road accidents. How can we explain the mode effects? Do we have to expect, depending on the interview situation and the victim experience, that people tend to answer socially desired? Or does the different structure of the socioeconomic variables in the sub-groups still play a role here for the answer pattern? 5
Table 3: Evaluation and Delictgroups by Mode persons involved in road burglary accidents phone mail % eta beta phone mail % Diff Diff (n=757) (n=756) (n=513)(n=430) Opinion Police NRW Very positiv/positiv 65 55-10.07**.08** 69 57-12 Opinion Police at the place of residence.07*.08** Very positiv/positiv 64 55-9 Opinion on the appearance, Place: record at the place.06*.07** Very well cared-for/well... 86 82-4 Opinion on the appearance, Place: Office.03 n.s.01 n.s Very well cared-for/well... 77 76-1 Opinion of the Information, Place: first contact by phone.20*.20 n.s Very helpful/helpful 89 61-28 Opinion of the Information, Place: Record at the place.02 n.s.03 n.s Very helpful/helpful 74 71-3 Opinion of the Information, Place: Office.02 n.s.03 n.s Very helpful/helpful 60 67 +7 Comprehensibility of the documents.10***.10*** Very clear/clear 81 84 +3 Satisfied with the arrangement of the indication.02 n.s.02 n.s Very satisfied/satisfied 74 75 +1 *p<.05, **p<.01, ***p<.001 73 63-10 91 82-9 85 80-5 80 61-19 78 68-10 85 61-24 90 86-4 75 69-6 eta beta.11***.08*.09**.07*.07*.05 n.s.05 n.s.04 n.s.16*.04 n.s.06 n.s.00 n.s.19**.16*.09 n.s.12*.05 n.s.01 n.s In order to examine this question, stratified multiple Classification Analyses are applied with all items you can see in table 3 as dependent variable. As Independent variables are used: Gender, Age, Education, Job and Mode. Comparing the eta-coefficients with the betacoefficients, it is striking that especially in the group of the victims of burglary the effect strength, which is clearly visible throughout all variables, is smaller. In the group of persons involved in road accidents, however, effects stay on the same level. 6
The effect of mode seems to bear on the items on Evaluation of the police in almost the same strength independent of victimization. The experience of the victim according to the kind of offence influences much more clearly the choice of survey. 4.Conclusions As has also been shown in the general literature on methods, one can show for special populations of interviewees the effect of mode. As I showed in my report, people gave higher rates of satisfaction in the telephone survey than in the mail survey. In addition to that, the readiness of interviewees, depending upon the experience as a victim, to take part in a telephone or a mail survey varies. Both effects can cumulate and therefore lead to clear effects of mode. References: Binder, J., Sieber, M and J. Angst (1979): Verzerrungen bei postalischen Befragungen: das Problem der Nichtbeantworter. Zeitschirft für experimentelle und angewandte Psychologie, 26: 53 71. Dörmann, U. (1996): Wie sicher fühlen sich die Deutschen? BKA-Forschungsreihe 3, Wiesbaden. Forschungsgruppe Kommunale Kriminalprävention in Baden-Württemberg (1996): Opfererfahrungen, Kriminalitätsfurcht und Vorstellungen zur Delinquenzprävention. Ergebnisse von Bevölkerungsbefragungen im Rahmen des Begleitforschungsprojekts Kommunale Kriminalprävention in Baden-Württemberg in: Trenczek T. und H. Pfeiffer (eds.), Kommunale Kriminalprävention, Wiesbaden: 118 139. Hindelang, M. J. (1974): Public Opinion regarding crime, criminal justice, and related topics. Crime and Delinquency Literature: 501 523. Hindelang, M.J., Hirschi, T. and J. G. Weis (1981): Measuring Delinquency, Sage 123, Beverly Hill, California. Hippler, H.-J. (1985): Schriftliche Befragung bei allgemeinen Bevölkerungsstichproben. Untersuchungen zur Dillmanschen Total Design Method. ZUMA-Nachrichten 16: 39 56. Kaiser, G. (1997): Kriminologie, Heidelberg. Kish, L. (1965): Survey Sampling, New York. 7
Kury, H. (1993): Der Einfluss der Art der Datenerhebung auf die Ergebnisse von Umfragen erläutert am Beispiel einer Opferstudie, in: G. Kaiser and H. Kury (eds.), Criminological Research in the 1990 s, Freiburg i. Br.: 321 410. Lamnek, S. and W. Kiefl (1986): Soziologie des Opfers, München. Schwind, H.-D., Fetchenhauer, D., Ahlborn, W. and R. Weiß (2001): Kriminalitätsphänomene im Langzeitvergleich am Beispiel einer deutschen Großstadt. Bochum 1975-1986-1998, Neuwied. Stephan, E. (1976): Die Stuttgarter Opferbefragung, BKA-Forschungsreihe 3, Wiesbaden. Reuband, K.-H. and J. Blasius (1996): Face-to-face, telefonische und postalische Befragungen. Ausschöpfungsquoten und Antwortmuster in einer Großstadtstudie. Kölner Zeitschrift für Soziologie und Sozialpsychologie 48: 296 318. Reuband, K.-H. (1999): Postalische Befragungen in den neuen Bundesländern. Durchführungsbedingungen, Ausschöpfungsquoten und soziale Zusammensetzung der Befragten in einer Großstadtstudie. ZA-Information 45: 71-99. Reuband, K.-H. (2001): Möglichkeiten und Probleme des Einsatzes postalischer Befragungen, in: Kölner Zeitschrift für Soziologie und Sozialpsychologie 53: 338 364. Zimmer, H. (1956): Validity of extrapolating non-response bias from mail-questionnaire follow-ups, in: Journal of applied Psychology 40: 117-121. 8