Korean Crime Victim Survey(Ⅹ) - An Experimental Study on Survey Methodological Improvements -
- LanguageKorean
- Authors Seonghoon Park, Minyoung Kim, Sunghyun Cho, Heejung Park, Kyungho Cho, Youngwon Kim, Mingue Park, Cheolhyun Park, Myoungjin Lee, Seoksoon Im, Hyojin Eom, Ansik Chang, Hyunga Choi
- ISBN979-11-89908-82-9
- Date December 01, 2020
- Hit513
Purpose of the study
The Korean Crime Victim Survey (KCVS), which is being conducted to estimate
crime victimization rate and to provide more comprehensive information about
the extent and nature of crime damage, has been redesigned in 2009 and collected
data every two years.
Over the past decade or so, the KCVS data has been widely used as a very
useful reference source in academic circles, the public and the community. The
KCVS, which has been widely used through various channels, has been criticized
for the need of improvements since the redesign of the survey in 2009. Despite
its usefulness, there has been consistent criticism on the survey’s capacity to
gather information about crime victimization. The major criticisms on KCVS can
be classified into three types; “modes of survey”, “survey sampling” and “survey
questionnaire”. Modes of survey, sampling and questionnaire are all parts of
survey method, and the growing demand for survey redesign is generally
approached with problems in survey methodology. The purpose of this study
is to deal with the methodological problems of the current KCVS and to ensure
and improve the accuracy of the survey through experimental study on survey
methods.
Research frame of the study
The study was designed with two primary issues.; (1) sample design, (2) survey
mode. Frist, the study on sample design deals with a range of issues on sample
selection and sample replacement, stratification variables, sample design, as well
as longitudinal panel design. The study on survey method deals with issues relating
survey tools, investigation methods, basic questionnaire and crime incident
questionnaire, crime types and classification feasibility.
Results of sampling design study
For the methodological advancement of the survey, variety range of aspects
relating to the sample design was studied. This study especially focused on the
several sample design issues regarding household unit sample selection and
sample replacement, and the study of sample design considering stratification
variables.
The properties of non-response in the Crime Victim Survey were analyzed
through experimental design for improvement of sample selection and sample
replacement. The experience of criminal damage is not easily understood as “rare
events,” and respondents tend to be reluctant to report their criminal damage.
The recent changes in the social atmosphere has added to the difficulty of
household visits. The nature of Crime Victim Survey and changes in survey
environment are likely to increase sample errors by allowing households who
experienced crime damage to avoid response while the alternative household
who inexperienced crime damage to participate in the survey. This study looked
at the occurrence of non response using experimental design surveys, and found
that the victimization rate of original sample households was higher than that
of substitution household. This result shows the correlation between the expedient
sample replacement and the criminal damage rate. Therefore, if the original
sample units were surveyed and the sample substitutes were reduced, it would
be possible to collect relatively reliable data. The damage rate is 3.53% in the
survey of all household members and 5.91% in the one member of the household
survey indicating that the selecting one member of the household survey method
is useful to secure cases of focal variables.
According to the data analysis to find a suitable stratification method for the
Crime Victim Survey, the crime rate per 100,000 population is relatively evenly
distributed by cities and provinces, while the Crime Victim Survey showed a large
deviation. Although crimes have occurred, there are many cases where there is
no damage as they have not been identified in the survey. When it comes to
the sample design for traditional household survey, the victims are less likely
to be included in the sample due to probability of rare events. Therefore,
considerably large samples should be taken into account to accurately estimate
the damage rate. However, the current sample size makes it very difficult to
identify the affected households or members of households, and the variance
of the estimates get larger. To overcome these limitations in part, it may be
considered to stratify the population using variables such as the number of crimes
or the rates of crime, and then to allocate more samples to the layer that are
relatively more vulnerable to crime. Rather than the current stratification method
which simply layers 17 regional units, we explored the new stratification method
of sample according to regional crime rates. It is found that stratification of
sample according regional crime rates are more effective for KCVS. In order to
increase the efficiency of the survey through stratification under the constraints
of the sample size and the given cost of the Crime Victim Survey, the method
of allocating samples based on stratification and the number of crimes based
on the results of the cluster analysis can be considered.
Results of survey mode study
With the rapid transition to a digital society, the survey environment is
constantly changing, and many limitations are pointed out in the Pen-and-Paper
Personal Interveiws (PAPI). In this study, we conducted experimental study
whether the changes in the survey tools affect the survey response. In this study,
the conventional KCVS survey method which the all household members are
surveyed showed no significant differences in the criminal damage rates according
to the survey tools of PAPI(Pen-and-Paper Personal Interview) and
TAPI(Tablet-Assisted Personal Interview). However, one member from household
selection method showed relatively higher criminal damage rates in TAPI based
survey than PAPI based survey. Furthermore, we compared self-report and face
to face interview method and few significant demographic differences were found.
The lower the age and the higher the educational background, it is more likely
to prefer the self-report interview mode than face-to-face interview. While the
PAPI respondents showed a higher preference for self-report method, the TAPI
respondents showed a higher preference for face-to-face interview method. To
sum up the findings of the study, important intuitions were found. Depite its
earlier concerns of mode effects, changes in survey tools didn't generate
significant changes in criminal damage rates. This can be interpreted that there
is not much possibility of error which arises from unfamiliarity or rejection of
using tablet devices. This means that TAPI can be used as an alternative to PAPI
survey and suggests that switching the survey tools from PAPI to TAPI will not
affect the quality of the data.
Results of survey questionnaire study
Since the Crime Victim Survey investigates sensitive issues such as criminal
damage experience, it is possible that the respondents would refuse to respond
and even if they participate, there still is a possibility that they will not respond
honestly. If the response rate is low or the response is false, the damage rate
would be either underestimated or overestimated, failing to accurately reflect the
reality and showing distorted results. The main purpose of redesign the survey
questionnaire part is to redevelop the survey questionnaire to improve survey
accuracy.
The study proposed ‘the new survey questionnaire’ which focused on two major
changes from the ‘existing survey questionnaire’ of the Crime Victim Survey. First,
unlike the previous survey questionnaire which determines the type of crime by
filling the separate crime survey questionnaire section, the new questionnaire
can determine the type of crime damage only by filling the basic questionnaire.
Second, the focus was made on reducing the response burden by halving the
number of questionnaires in the crime survey section. Using the existing and
new survey questionnaire, the number of criminal damage was compared. With
the new survey questionnaire, the criminal damage case per 1,000 person was
74.5 compared to 26.5 in the exiting survey questionnaire, nearly three times
higher. Compared to the previous survey questionnaire, the damage rate for
assault was 2.7 times higher, the sexual harassment damage rate was 3.1 times
higher, continuous harassment damage 5.3 times, and burglary/pickpocket
damage 4.7 times, home invasion(theft) damage 1.3 times , car theft damage 1.3
times, fraud damage 1.2 times, and destruction damage 16.0 times higher than
the previous questionnaire.
In the focus group interview, many respondents suggested that the new survey
questionnaire is easier to respond to than the previous survey questionnaire.
Recommendations for the KCVS improvement
First, for the future crime damage survey, it is necessary to expand the use
of multiple survey tools to take full advantage of each survey tools which enables
to offer a survey in more customized manner considering the situation or
conditions of individual respondents. In sensitive surveys such as crimes or
victimization, neither unilateral self-report method, nor unilateral face-to-face
interview mehod which follows general principles of social investigation cannot
be the only alternative.
Second, what is more important than the survey method is the formation of
a rapport between interviewers and respondents. It is necessary to use flexible
interviewing methods considering respondent’s situation, demands and needs
since the real process of social investigation is another form of social interaction.
In order for the interview process to proceed smoothly and flexibly at the site,
the role of the interviewer is most important. Therefore the efforts should be
taken account to improve the interviewer’s performances.
Third, it is necessary to find ways to maintain the original sample household
as much as possible by finding ways to avoid convenient sample substitutes. Given
the increasing difficulty of household visits in recent years, measures need to
be sought to increase the response rate of samples that are difficult to investigate.
Various action plans should be taken, including increasing the number of re-visits
by interviewers, improving incentives for respondent by providing survey tools
that are preferred, providing detailed informations about Crime Victim survey
such as policy implications, and improving the interviewer’s ability to form
rapport and so on.
Fourth, the respondent‘s lower age limit of 14 needs to be adjusted. According
to the previous survey results, the youth respondents has maintained a small
proportion of 5 percent, and they are not fully aware of their experience in
crime. Considering that separate surveys for youth victimization are being
conducted on an irregular basis, it is necessary to think over changing the age
limit of 14 years old.
Fifth, a very large sample is required to estimate the crime damage rate using
the Crime Victim Survey and to ensure the accuracy of survey. In order to make
efficient use of a given budget, a new sample stratification method and estimation
measures need to be explored in the light of the variables of interest. However,
in order for the new stratification plan to be realized, an approach to raw data
of crime occurrence should be made.
Sixth, the advantages of panel investigation can be actively utilized by
generating profile panel data about victims. The psychological and physical
changes caused by crime damage can be accurately identified and used as basic
data for policies for supporting victims, It can be used as a basis for crime
prevention policy by analyzing longitudinally about the relationship between
vulnerability factors and crime damage and it would be possible to estimate the
hidden crimes or analyze the cause of habitual damage by investigating police
report and repeated crime damage. In reality, however, there are considerable
difficulties in conducting a panel survey on crime damage. Conducting a panel
investigation involves several issues regarding on research ethics that can arise
from identifying private information on damage experiences and continuing to
track and investigate victims, as well as issues related to high budgetary cost
of recruitment and maintenance of panels, including special groups of crime
victims, it is essential to secure resources such as recruiting dedicated experts
and establishing exclusive department to establish crime damage investigation
panels and analyze data.
The Korean Crime Victim Survey (KCVS), which is being conducted to estimate
crime victimization rate and to provide more comprehensive information about
the extent and nature of crime damage, has been redesigned in 2009 and collected
data every two years.
Over the past decade or so, the KCVS data has been widely used as a very
useful reference source in academic circles, the public and the community. The
KCVS, which has been widely used through various channels, has been criticized
for the need of improvements since the redesign of the survey in 2009. Despite
its usefulness, there has been consistent criticism on the survey’s capacity to
gather information about crime victimization. The major criticisms on KCVS can
be classified into three types; “modes of survey”, “survey sampling” and “survey
questionnaire”. Modes of survey, sampling and questionnaire are all parts of
survey method, and the growing demand for survey redesign is generally
approached with problems in survey methodology. The purpose of this study
is to deal with the methodological problems of the current KCVS and to ensure
and improve the accuracy of the survey through experimental study on survey
methods.
Research frame of the study
The study was designed with two primary issues.; (1) sample design, (2) survey
mode. Frist, the study on sample design deals with a range of issues on sample
selection and sample replacement, stratification variables, sample design, as well
as longitudinal panel design. The study on survey method deals with issues relating
survey tools, investigation methods, basic questionnaire and crime incident
questionnaire, crime types and classification feasibility.
Results of sampling design study
For the methodological advancement of the survey, variety range of aspects
relating to the sample design was studied. This study especially focused on the
several sample design issues regarding household unit sample selection and
sample replacement, and the study of sample design considering stratification
variables.
The properties of non-response in the Crime Victim Survey were analyzed
through experimental design for improvement of sample selection and sample
replacement. The experience of criminal damage is not easily understood as “rare
events,” and respondents tend to be reluctant to report their criminal damage.
The recent changes in the social atmosphere has added to the difficulty of
household visits. The nature of Crime Victim Survey and changes in survey
environment are likely to increase sample errors by allowing households who
experienced crime damage to avoid response while the alternative household
who inexperienced crime damage to participate in the survey. This study looked
at the occurrence of non response using experimental design surveys, and found
that the victimization rate of original sample households was higher than that
of substitution household. This result shows the correlation between the expedient
sample replacement and the criminal damage rate. Therefore, if the original
sample units were surveyed and the sample substitutes were reduced, it would
be possible to collect relatively reliable data. The damage rate is 3.53% in the
survey of all household members and 5.91% in the one member of the household
survey indicating that the selecting one member of the household survey method
is useful to secure cases of focal variables.
According to the data analysis to find a suitable stratification method for the
Crime Victim Survey, the crime rate per 100,000 population is relatively evenly
distributed by cities and provinces, while the Crime Victim Survey showed a large
deviation. Although crimes have occurred, there are many cases where there is
no damage as they have not been identified in the survey. When it comes to
the sample design for traditional household survey, the victims are less likely
to be included in the sample due to probability of rare events. Therefore,
considerably large samples should be taken into account to accurately estimate
the damage rate. However, the current sample size makes it very difficult to
identify the affected households or members of households, and the variance
of the estimates get larger. To overcome these limitations in part, it may be
considered to stratify the population using variables such as the number of crimes
or the rates of crime, and then to allocate more samples to the layer that are
relatively more vulnerable to crime. Rather than the current stratification method
which simply layers 17 regional units, we explored the new stratification method
of sample according to regional crime rates. It is found that stratification of
sample according regional crime rates are more effective for KCVS. In order to
increase the efficiency of the survey through stratification under the constraints
of the sample size and the given cost of the Crime Victim Survey, the method
of allocating samples based on stratification and the number of crimes based
on the results of the cluster analysis can be considered.
Results of survey mode study
With the rapid transition to a digital society, the survey environment is
constantly changing, and many limitations are pointed out in the Pen-and-Paper
Personal Interveiws (PAPI). In this study, we conducted experimental study
whether the changes in the survey tools affect the survey response. In this study,
the conventional KCVS survey method which the all household members are
surveyed showed no significant differences in the criminal damage rates according
to the survey tools of PAPI(Pen-and-Paper Personal Interview) and
TAPI(Tablet-Assisted Personal Interview). However, one member from household
selection method showed relatively higher criminal damage rates in TAPI based
survey than PAPI based survey. Furthermore, we compared self-report and face
to face interview method and few significant demographic differences were found.
The lower the age and the higher the educational background, it is more likely
to prefer the self-report interview mode than face-to-face interview. While the
PAPI respondents showed a higher preference for self-report method, the TAPI
respondents showed a higher preference for face-to-face interview method. To
sum up the findings of the study, important intuitions were found. Depite its
earlier concerns of mode effects, changes in survey tools didn't generate
significant changes in criminal damage rates. This can be interpreted that there
is not much possibility of error which arises from unfamiliarity or rejection of
using tablet devices. This means that TAPI can be used as an alternative to PAPI
survey and suggests that switching the survey tools from PAPI to TAPI will not
affect the quality of the data.
Results of survey questionnaire study
Since the Crime Victim Survey investigates sensitive issues such as criminal
damage experience, it is possible that the respondents would refuse to respond
and even if they participate, there still is a possibility that they will not respond
honestly. If the response rate is low or the response is false, the damage rate
would be either underestimated or overestimated, failing to accurately reflect the
reality and showing distorted results. The main purpose of redesign the survey
questionnaire part is to redevelop the survey questionnaire to improve survey
accuracy.
The study proposed ‘the new survey questionnaire’ which focused on two major
changes from the ‘existing survey questionnaire’ of the Crime Victim Survey. First,
unlike the previous survey questionnaire which determines the type of crime by
filling the separate crime survey questionnaire section, the new questionnaire
can determine the type of crime damage only by filling the basic questionnaire.
Second, the focus was made on reducing the response burden by halving the
number of questionnaires in the crime survey section. Using the existing and
new survey questionnaire, the number of criminal damage was compared. With
the new survey questionnaire, the criminal damage case per 1,000 person was
74.5 compared to 26.5 in the exiting survey questionnaire, nearly three times
higher. Compared to the previous survey questionnaire, the damage rate for
assault was 2.7 times higher, the sexual harassment damage rate was 3.1 times
higher, continuous harassment damage 5.3 times, and burglary/pickpocket
damage 4.7 times, home invasion(theft) damage 1.3 times , car theft damage 1.3
times, fraud damage 1.2 times, and destruction damage 16.0 times higher than
the previous questionnaire.
In the focus group interview, many respondents suggested that the new survey
questionnaire is easier to respond to than the previous survey questionnaire.
Recommendations for the KCVS improvement
First, for the future crime damage survey, it is necessary to expand the use
of multiple survey tools to take full advantage of each survey tools which enables
to offer a survey in more customized manner considering the situation or
conditions of individual respondents. In sensitive surveys such as crimes or
victimization, neither unilateral self-report method, nor unilateral face-to-face
interview mehod which follows general principles of social investigation cannot
be the only alternative.
Second, what is more important than the survey method is the formation of
a rapport between interviewers and respondents. It is necessary to use flexible
interviewing methods considering respondent’s situation, demands and needs
since the real process of social investigation is another form of social interaction.
In order for the interview process to proceed smoothly and flexibly at the site,
the role of the interviewer is most important. Therefore the efforts should be
taken account to improve the interviewer’s performances.
Third, it is necessary to find ways to maintain the original sample household
as much as possible by finding ways to avoid convenient sample substitutes. Given
the increasing difficulty of household visits in recent years, measures need to
be sought to increase the response rate of samples that are difficult to investigate.
Various action plans should be taken, including increasing the number of re-visits
by interviewers, improving incentives for respondent by providing survey tools
that are preferred, providing detailed informations about Crime Victim survey
such as policy implications, and improving the interviewer’s ability to form
rapport and so on.
Fourth, the respondent‘s lower age limit of 14 needs to be adjusted. According
to the previous survey results, the youth respondents has maintained a small
proportion of 5 percent, and they are not fully aware of their experience in
crime. Considering that separate surveys for youth victimization are being
conducted on an irregular basis, it is necessary to think over changing the age
limit of 14 years old.
Fifth, a very large sample is required to estimate the crime damage rate using
the Crime Victim Survey and to ensure the accuracy of survey. In order to make
efficient use of a given budget, a new sample stratification method and estimation
measures need to be explored in the light of the variables of interest. However,
in order for the new stratification plan to be realized, an approach to raw data
of crime occurrence should be made.
Sixth, the advantages of panel investigation can be actively utilized by
generating profile panel data about victims. The psychological and physical
changes caused by crime damage can be accurately identified and used as basic
data for policies for supporting victims, It can be used as a basis for crime
prevention policy by analyzing longitudinally about the relationship between
vulnerability factors and crime damage and it would be possible to estimate the
hidden crimes or analyze the cause of habitual damage by investigating police
report and repeated crime damage. In reality, however, there are considerable
difficulties in conducting a panel survey on crime damage. Conducting a panel
investigation involves several issues regarding on research ethics that can arise
from identifying private information on damage experiences and continuing to
track and investigate victims, as well as issues related to high budgetary cost
of recruitment and maintenance of panels, including special groups of crime
victims, it is essential to secure resources such as recruiting dedicated experts
and establishing exclusive department to establish crime damage investigation
panels and analyze data.
File
- 20-B-02_전국범죄피해조사(Ⅹ) 범죄피해조사 방법론 연구_(0310).pdf (11.65MB / Download:328) Download