Abstract
The present paper describes a cross-sectional study aimed at assessing factors contributing to losses from property crime. The literature suggested that GDP per capita, regional education level, and regional urbanization level were possible predictors of losses from property crime rates. The research utilized secondary data acquired from the World Bank. Two two-level multilevel models were assessed based on a sample of 103 countries. The analysis revealed that regional education level was a significant predictor of the dependent variable, while the significance of GDP per capita and regional urbanization level was insufficient. Thus, when designing programs aiming at reducing business losses from property crime, policies for improving the current education level should be considered.
Literature Review
The primary research question of the present paper is the following:
- RQ1: What are the factors affecting country-wide losses from property crime?
The hypothesis of the present research is that country-wide losses from property crime are affected by gross domestic product (GDP) per capita and the mean education and urbanization levels of the region.
Factors affecting losses from property crime are an understudied topic. However, the literature provided on the subject can explain common determinants of country crime rates. Rosenfeld and Messner (2009) conducted a comparative analysis of factors affecting the decline in property crime rates in the US and several European countries. The researchers utilized robust quantitative methods to test the hypothesis that GDP per capita, unemployment, consumer confidence, and the number of imprisonments were correlated with the prevalence of property crime in European countries and the US. After testing several regression models, the researchers accepted their hypotheses, demonstrating that the strength of the law enforcement system and economic factors affected country-wide crime rates.
Similarly, Crawford (2013) claimed that economic determinants and law enforcement spending had a significant effect on property crime prevalence in both developing and developed countries. However, Crawford (2013) utilized only qualitative methods to arrive at a conclusion, which may mean that the results are biased. Ghani (2017) conducted a comparative study using mixed methods to examine different determinants of crime rates, such as education level, poverty, GDP per capita, and unemployment. A special emphasis was put on the fact that urbanization negatively affected crime rates (Ghani, 2017).
Thus, the literature review justified the choice of predictors for the country-wide crime rates. At the same time, it revealed a lack of recent studies on the topic utilizing robust quantitative methods that confirm previous findings. Additionally, there were no studies found assessing the determinants of losses from property crime. Therefore, the present paper will close a significant gap in the literature.
Methods
The present paper is a cross-sectional quantitative study which aims at determining factors affecting country-wide crime levels using secondary data. Cross-sectional studies make inferences about the data gathered during a specific point in time (Baarda, 2010). The method is associated with low costs, assessment of multiple variables, and creation of explanatory models, such as regression equations (Baarda, 2010). The disadvantages of cross-sectional methods include lack of control over the environment, inability to identify cause-effect relationships, and inability to analyze behavior during a period of time (Baarda, 2010).
The utilization of secondary data is also associated with several flaws, such as poor control over variables and lack of knowledge about the methods of data collection (Baarda, 2010). However, the use of secondary data is convenient, as it helps to make the research cheaper and faster. Additionally, secondary data acquired from reputable sources, such as World Bank, is highly reliable. Such a level of data collection accuracy and consistency is difficult to achieve in field research.
The methods were selected to test three hypotheses that emerged from the primary hypothesis mentioned in Section 1 of the present paper:
- H1: Country-wide losses from property crime are negatively correlated with GDP per capita.
- H2: Country-wide losses from property crime are negatively correlated with regional education level.
- H3: Country-wide losses from property crime are positively correlated with regional urbanization level.
Sample
The final dataset was created using secondary data from World Bank (2021a; 2021b; 2021c; 2021d). A total of four complete datasets were used, with 264 lines standing for countries and country groups. Every dataset included observations for all the lines between 1960 and 2020. It was decided to use the average values for the past fifteen years (between 2006 and 2020) to acquire the most recent results. This decision was consistent with the purpose of the present paper to close the gap in the literature that lacked research based on current data.
Considering the facts mentioned above, the initial sample size 264 countries and country groups. However, it was decided to avoid using the information on country groups as they may have caused problems with the reliability of findings. Additionally, data on some countries on different variables were missing, is it was decided to avoid using the information on countries with the lack of data for consistency purposes.
Thus, after the clearing process, the final sample included 103 countries. The countries were classified into six different regions, including Africa, Americas, Asia (n = 20), Europe, Middle East, and island countries. It should be noticed that that region “Americas” does not include the information on Canada and the United States, as the some of the data on these countries was missing. All the island countries such as the Bahamas, Jamaica, and Mauritius, were united into one region due to similarities in the economic structure of these countries.
Measures
Losses from Property Crime
Losses from property crimes were measured using the dataset provided by World Bank (2021a) called Losses due to theft and vandalism (% of annual sales of affected firms). The dataset provided information on the percentage of sales lost by firms due to property crimes. Since the dataset had numerous observations missing for different years and countries, it was decided to use average values for the past 15 years (between 2006 and 2020). As the reason for the lack of observations was unclear, countries with no data on losses from crime property crime were excluded from the analysis. Additionally, as it was mentioned in Section 2A, all the country groups were excluded from the analysis, which left a total of 144 observations for the analysis. The data collection method utilized by World Bank was not specified in the dataset; however, considering the reputation of the source, the data was considered reliable.
GDP per Capita
The information on GDP per capita was also taken from World Bank (2021b) for consistency purposes. There were no countries excluded based on the absence of data about GDP per capita, as World Bank (2021b) provided information on all the 144 countries. Average values for the past 15 years were used to be consistent with the dependent variable. It was also decided to use the value in thousands of USD.
Regional Urbanization Level
The urbanization level was measured as the percentage of the urban population in the country. The data was taken from World Bank (2021c), and average values for the past 15 years were obtained. No countries were excluded base on the absence of data on this variable. The regional urbanization level was calculated by estimating the mean urbanization level for all the countries in the region.
Regional Education Level
Education was measured as the percentage of people above 25 who have completed at least lower secondary education. The measure was selected for convenience purposes, as the data was available from World Bank (2021d). Average values for years between 2006 and 2020 were obtained for consistency purposes. Since the information on the education levels was scarce, 36 countries were excluded from the analysis due to the absence of data. The regional education level was calculated by estimating the mean urbanization level for all the countries in the region.
Procedures
Two two-level multilevel models were analyzed using IBM SPSS (IBM Corp, 2020). The equations for the first model were the following:
- Level1: Yij = β0j + β1jGDPij + εij
- Level2: β0j = ϒ00 + μ0j
The equations for the second were the following:
- Level1: Yij = β0j + β1jGDPij + εij
- Level2: β0j = ϒ00 + ϒ01RULj + ϒ02RULj+ μ0j
β0j = ϒ00
Where:
- Yij – country-level losses from property crime;
- β0j – country-level intercept;
- GDPij – country-level GDP per capita;
- ϒ00 – regional-level intercept;
- RULj – regional urbanization level;
- RELj – regional education level.
The dependent variable was identified as losses from property crime. The literature review revealed that there was a negative correlation between the economic prosperity of the country and crime rates (Crawford, 2013; Ghani, 2017; Rosenfeld & Messner, 2009). All the literature reviewed for the present paper utilized GDP per capita as the primary indicator of economic well-being; thus, GDP per capita was included in the equation. The rationale behind using the urbanization level as an independent variable was Ghani’s (2017) claim that, when clustered together, people are more inclined to commit property crimes. The rationale behind including education level in the analysis was the idea that less educated people have a poor understanding of the consequences of crime (Crawford, 2013). As was mentioned in the previous sections, the issues with missing data were addressed by using average values and excluding the countries with missing data from the analysis.
Results
Descriptive Statistics and Correlations
Descriptive Statistics
Losses from property crime among different countries varied between 0.5 and 15.85 for both datasets. The mean value was 4.52 with a standard deviation (SD) of 3.06. The level of education varied between 6.15 and 99.89. The mean value was 57.13 with an SD of 28.26. The minimum urbanization level was 11.4, while the maximum value was 94.66. The mean urbanization was 54.17 with an SD of 21.21. As for GDP per capita, the minimum value was 0.23, while the maximum value was 54.61. The mean was 7.52 with an SD of 9.01. The descriptive statistics for country-level and regional-level variables are provided in Table 1 below.
Table 1. Descriptive Statistics.
All the countries were distributed into six regions. Frequency distribution table of the frequencies is demonstrated in Table 2 below.
Table 2. Frequency Distribution Table for Regions.
Correlation Analysis
Pearson’s correlation analysis for country-level variables demonstrated statistically significant correlations between all the variables (see Table 2). The analyses revealed strong correlations (>0.5) between GDP per capita and urbanization level. This can be explained that increased city population leads to greater industrialization, which, in turn, increases production and GDP. Additionally, strong correlations were found between GDP per capita and the education level, which can be explained by the fact that increased economic well-being allows the government to spend more on education. The analysis also revealed that there was a negative correlation between urbanization and losses from property crime, which is against one of the hypotheses (H3). This can be explained by the fact that the correlation analysis does not control for GDP per capita.
Table 2. Correlation Matrix.
Model Results
The results for the estimation of the first model revealed that country-level GDP per capita was positively correlated with the country-level losses from property with a significance level of p < 0.001. This finding supports H1. The estimates of fixed effects of the first model are demonstrated in Table 3 below.
Table 3. Model 1 Results.
The results of the analysis of the second model were different from the estimation of the first model. The estimates of fixed effects for the second model are provided in Table 4 below.
Table 4. Model 2 Results.
The estimations of the second model revealed that the only statistically significant predictor of the losses from property crime was the regional education level with a significance level of p < 0.001 (t = -3.52). The model estimation confirmed H2, as the coefficient for the variable was negative. At the same time, H1 and H3 found no support in the estimation results. Even though GDP per capita had a negative coefficient (-0.05), which demonstrates negative correlation, the significance level of the coefficient (0.196) was below the required alpha of 95%. Thus, H1 was rejected based on the two-level multilevel modeling. The coefficient for the regional urbanization level was negative (-0.02) and statistically insignificant (t=-0.75; p = 0.454).
Thus, H3 was also rejected, which implied that there was no significant correlation between the urbanization level and the country-level losses from property crime. The differences in model estimations can be explained be the fact that the first model did not control for other variables and significant correlations between education and GDP per capita.
Discussion
In summary, the present research aimed at explaining the variability in losses from property crime using three predictors, including economic prosperity, regional education level, and regional urbanization level. The results of the analyses were controversial, as estimates of the first model were inconsistent with estimates of the second model. However, such inconsistencies can be explained by correlations between the predictors. Educated people are more likely to make economics of a country to be more intensive, which can lead to increased economic prosperity. Thus, it is natural that if analyzed without controlling for education level, GDP per capita would have a significant effect on losses from property crime. The present paper also demonstrates that due to increased globalization, regional education level is more important than country education level, as people can travel freely between countries, which are close together.
The results of the study are partially inconsistent with previous research. The analysis conducted by Rosenfeld and Messner (2009) generated similar results, as the author concluded that property crimes are more prevalent in less economically advantaged countries. The present paper demonstrates that there are no significant correlations between GDP per capita and losses from property crimes. Moreover, correlations were found between regional urbanization levels and losses from property crime, which is inconsistent with the findings by Ghani (2017). However, regional education level was found to decrease the percentage of sales losses from property crime, which was also mentioned by Crawford (2013). In summary, the present research found sufficient support only for H2.
Limitations and Recommendations for Future Research
The primary limitation of the present research is the absence of data on losses from property crime for high-income European countries. This limitation was the result of using secondary data, which can often be inappropriate for answering the research question. Thus, it is recommended that the results of the present research are primary applied to developing countries. Therefore, future research should aim at conducting similar analyses using primary data to receive consistent results. Additionally, future research is recommended to add other variables to the regression model to achieve the greater predictive ability of the model.
Central Implication of the Study
The present study has a significant implication for government authorities. In particular, when designing programs aiming at reducing business losses from property crime, policies for improving the current education level should be considered. Such policies will allow the general population to understand the consequences of crime (such as imprisonment). Moreover, increased education levels will allow the population to receive higher-paid jobs, which will reduce the need to commit crimes.
References
Baarda, B. (2010). Research: This is it! Groningen, The Netherlands: Noordhoff.
Crawford, A. (2013). Crime prevention policies in comparative perspective. Abingdon-on-Thames: Routledge.
Ghani, Z. A. (2017). A comparative study of urban crime between Malaysia and Nigeria. Journal of Urban Management, 6(1), 19-29.
IBM Corp. (2020). IBM SPSS Statistics for Windows, version 27.0. IBM Corp.
Rosenfeld, R., & Messner, S. F. (2009). The crime drop in comparative perspective: The impact of the economy and imprisonment on American and European burglary rates. The British Journal of Sociology, 60(3), 445-471.
World Bank. (2021a) Losses due to theft and vandalism (% of annual sales of affected firms). Web.
World Bank. (2021b). GDP per capita (constant 2010 US$). Web.
World Bank. (2021c). Urban population (% of total population). Web.
World Bank. (2021d). Education attainment, at least completed lower secondary education, population 25+, total (%), cumulative. Web.