Econometrics report factors affecting crime rates in g20 countries from 2010 – 2020

TABLE OF CONTENTS 2 ABSTRACT 1 INTRODUCTION 1 SECTION I: LITERATURE REVIEW 3 1.1. Economic Growth (GDP growth) 3 1.2. Inflation Rate 4 1.3. Population 4 1.4. Income Inequality 5 1.5. Urbanization 5 1.6. Governance 6 SECTION II: RESEARCH METHODOLOGY 7 2.1. Methodology 7 2.2. Model specification 7 2.3. Data description 8 2.3.1. Sources of data 8 2.3.2. Descriptive statistics and interpretation for each variable 9 2.3.3. Correlation matrix between variables 10 SECTION III: RESEARCH RESULTS AND IMPLIACATION 12 3.1. Choosing the Estimated Model 12 3.1.1. Breusch Pagan test 12 3.1.2. Hausman test 13 3.2. Diagnosing the problems of the model 14 3.2.1. Testing for multicollinearity 14 3.2.2. Testing for heteroskedasticity 14 3.2.3. Testing for serial correlation 15 3.2.4. Testing for crosssection correlation 15 3.3. Fixing the model 16 3.4. Hypothesis testing 16 3.4.1. Test the overall significance of the observed multiple regression 17 3.4.2. Test the individual significance: 18 3.5. Result analysis and implications 19 CONCLUSION 21 REFERENCES 22 APPENDIX 25

FOREIGN TRADE OF UNIVERSITY

FACULTY OF INTERNATIONAL ECONOMICS

Ha Noi, June 2023

TABLE OF CONTENTS TABLE OF CONTENTS

ABSTRACT

INTRODUCTION

SECTION I: LITERATURE REVIEW

1.1 Economic Growth (GDP growth) 3

3.2 Diagnosing the problems of the model 14

3.2.1 Testing for multicollinearity 14 3.2.2 Testing for heteroskedasticity 14 3.2.3 Testing for serial correlation 15 3.2.4 Testing for cross-section correlation 15

Understanding the determinants of crime is essential for governments andpolicymakers in developing effective crime prevention strategies The literaturereview explores the complexities surrounding crime rates and various economic,demographic, and societal factors While economic growth demonstrates a weakcorrelation with crime rates, inflation is found to have a positive association,potentially incentivizing criminal behavior Moreover, population size, particularly

in urban areas, is consistently linked to higher crime rates However, the impact ofincome inequality and governance on crime rates remains inconclusive

The regression analysis results reveal several significant findings GDPgrowth, population size, urbanization, and inflation exhibit statistically significantrelationships with crime rates However, income inequality and governance do notdemonstrate significant effects These findings contribute to our understanding ofthe complex relationships between economic and demographic factors and crimerates in G20 countries

Keyword: Crime rates, G20, growths

Crime rates continue to be a pressing concern for governments worldwide, asthey pose significant threats to societal well-being, economic development, and publicsafety Understanding the factors that influence crime rates is crucial for formulatingeffective policies and strategies to combat crime and maintain social order Thisresearch paper delves into the exploration of the factors affecting crime rates in G20countries from 2010 to 2020, employing an econometric approach to estimate therelationships between various independent variables and the dependent variable, crimerate

The independent variables include GDP growth, inflation, population size,income inequality, urbanization, and governance By examining the relationshipsbetween these factors and crime rates, this research aims to contribute to the existingbody of literature on the topic, shed light on the complexities surrounding crime rates,and provide insights that can inform evidence-based crime prevention strategies Theindependent variables include GDP growth, inflation, population size, incomeinequality, urbanization, and governance

The literature review reveals that the relationship between economic factors andcrime rates is not straightforward Economic growth, as measured by GDP growth,demonstrates a weak correlation with crime rates, with some studies suggesting thatperiods of rapid economic growth may contribute to an increase in crime rates Thisphenomenon may be due to technological advancements and easier access tocommunication methods, which can facilitate criminal activities Furthermore, amaterialistic way of life associated with economic prosperity may lead to a decline inmoral values, potentially influencing crime rates

Inflation, another economic variable, has been found to positively associatedcrime rates Studies indicate that inflation weakens purchasing power, reducing thequality of life and incentivizing individuals to engage in criminal behavior tosupplement their resources Conversely, a decline in inflation during periods ofeconomic growth has been associated with a decrease in crime rates

Population size, particularly in urban areas, has consistently been linked to highercrime rates Larger cities tend to experience higher crime rates compared to smallercities, with the relationship potentially following a linear pattern The growth of urbanareas facilitates social contacts, both positive and negative, depending on the formal

structure of the population Consequently, urbanization is regarded as a contributingfactor to increased crime rates.

Income inequality has been a subject of debate concerning its impact on crimerates Some studies suggest that economic inequality fosters an environment conducive

to criminal behavior, while others argue that the relationship between incomeinequality and crime rates depends on cultural contexts and perceptions of inequality

as unjust

Governance plays a crucial role in addressing crime and maintaining social order.Strong state institutions and effective governance mechanisms are associated withlower crime rates, while poor governance, corruption, and weak state institutions arelinked to higher levels of crime Empirical studies have identified indicators of poorgovernance, such as the rule of law and legitimacy, as significant independent factorsinfluencing crime rates across different countries

To examine the relationships between crime rates and the aforementionedindependent variables, an econometric model was constructed using Stata software.The analysis utilized a dataset comprising 363 observations from G20 countries overthe period from 2010 to 2020 The regression analysis, employing the Driscoll-Kraaystandard errors method, generated coefficients and standard errors for eachindependent variable The statistical significance of these coefficients was assessedusing t-statistics and p-values, while 95% confidence intervals provided a range withinwhich the true population coefficients are likely to fall

About the structure of the research report: The essay is made with three mainsections:

Section I: Literature Review

Section II: Research Methodology

Section III: Research Results and Implication

SECTION I: LITERATURE REVIEW

As old as humanity, crime has grown to be a significant issue for governments.Therefore, stopping illegal activities is a key goal for the government Because aproper understanding of these is required to prevent crime, the causes of crime havealso been extensively investigated According to the economic model tradition, themajority of the important variables influencing crime are thought to be economic,including people's level of education, income, income inequality, unemployment, rate

of urbanization, age of the population, gender ratio, labor force participation rate,number of laws, security expenditures, and criminal records

The impact of socioeconomic and demographic factors on crime rates has beenexamined in numerous empirical studies Different time periods and nations have beentaken into consideration, along with a variety of methodological approaches Thiscould help to explain why there isn't much agreement on the effects of these factors inthe literature—there are even some results that are at odds with one another

1.1. Economic Growth (GDP growth)

Detotto and Otranto (2010) stated that criminal activity acted as a tax on theentire economy, lowering the competitiveness of businesses, discouraginginvestments, and reallocating resources, leading to uncertainty and inefficiency Thesefactors all had a negative impact on economic performance Li et al (2018) analysisincluded using SOM on crime in Japan from 1926 to 2013 to cluster and facilitatepractical comparison between different historical periods with GDP growth rate andcriminal features They found that there were only weak correlations between GDPgrowth rate and crime rates, but ultimately, they stated that the relationship betweencrime and economic development had been considered complicated Fajnzylber et al.(2002) suggested that the widespread adoption of technological advancements andeasy access to more sophisticated communication methods during periods of rapideconomic growth could result in an increase in crime rates Mauro and Carmeci (2007)empirically explore the link between crime, unemployment, and economic growthusing Italian regional data Using a standard overlapping exogenous growth model,they found transitional negative effects of unemployment and crime on income growthand permanent income level effects

1.2. Inflation Rate

The rate of inflation is also a significant factor in determining crime However,despite its importance, it is frequently left out of studies of crime Studies by Teles(2004), and Tang and Lean (2007) that looked into inflation as a potential factor incrime are among the exceptions As inflation weaken purchasing power, individualsmay purchase lesser goods with a given income This could potentially lower theirquality of life and encourage them to turn to crime as a means of obtaining additionalresources As Teles (2004) noted, it was evident that income encourages criminalinvolvement because it affected preferences for crime and because inflation reducedreal income Similarly, Deadman and MacDonald (2002) discovered in a studyconducted in the USA that a decline in inflation was associated with a drop in crimerates during a period of sustained economic growth

In Ralston (1999) research in the USA, he provided evidence in favor of thetheory by demonstrating a causal link between inflation and crime rates A similarpositive relationship between inflation and crime rates as well as a long-term co-integration between them was discovered by Tang and Lean (2007), who also looked

at the USA Inflation, according to Seals and Nunley's 2007 study, is a significantcontributor to crime In particular, they discovered a correlation between inflation andcrime rates in the 1960s, 1970s, and 1990s According to Tang's 2009 study ofMalaysia, crime was significantly influenced by inflation from 1970 to 2006 in thecountry A co-integration test conducted as part of the same study found a long-termconnection between inflation and crime Similarly, inflation and poverty have a long-term relationship with crime, according to Gillani et al (2009), who conducted asimilar analysis of Pakistan

1.3. Population

The population also plays a role in determining the crime rate, several studies haddiscovered small, favorable correlations between population size and crime rate Blau(1977) also proposed a connection between population size and crime, based on thestraightforward premise that social associations required opportunities for socialcontacts He postulated that there was a good chance that the population size and crimerelationship was linear According to Chamlin and Cochran (2004), population sizehad no impact on any of the equations relating to crime rates But regardless of thefunctional form looked at, population size had a big impact on how many violent andproperty crimes there were Population growth facilitates all types of social contacts,

from the most beneficial to the most harmful, according to the formal structure of thepopulation, which asserts categorically (Mayhew and Levinger 1976; Blau 1977).Although the relationship between population size and crime rate might seem to beinsignificant (Nolan, 2004), it may actually be statistically significant The factorsinfluencing crime rates in the EU-15 countries from 2000 to 2007 were examined byLauridsen et al (2013) The study concentrated on the rate of inflation, educationalattainment, earnings, and employment The findings revealed that while inflation rate,employment potential, and the urban population had positive effects on crime.

1.4. Income Inequality

Stven Stack (1984) delves into the interaction between inequality and variablesthat are believed to contribute to a perception of inequality as unjust The coreargument posits that the extent to which inequality affects crime rates is contingentupon a contextual element: a culture that fosters radical egalitarianism and condemnsinequality The study analyzes data on property crime across 62 nations

In a recent study, Eran Itskovich (2023) presented a novel explanation for apositive correlation between economic inequality and criminal activity, grounded inthe social resistance framework The hypothesis suggests that economic inequalitycauses individuals to feel disconnected from societal institutions and values, leadingthem to resist these structures through criminal activity Through survey data fromIsrael, they tested this theory on two distinct types of crime and applied structuralequation modeling to validate our results Their findings provide preliminary evidencethat economic inequality fosters a fertile environment for criminal behavior bypromoting resistance to fundamental societal values and institutions

1.5. Urbanization

The correlation between crime and city size is a well-established fact that hasbeen recognized by social observers for quite some time Criminologists haveextensively discussed the tendency for urban areas to experience higher levels ofcriminal activity Denis A Ladbrook (1988) conducted a study using cross-sectionalJapanese data from 1970 to explore why conventional crime rates are more prevalent

in urban areas than in rural areas Through his research, Ladbrook identified threesociological explanations for this phenomenon Firstly, he attributed the higher rates ofurban crime to the degree of urbanization and population density Secondly, he notedthat urban populations experience greater rates of migration and population growth,

which can contribute to the increase in crime Lastly, Ladbrook pointed out that thedemographic structures of urban and rural areas differ, with urban areas having ahigher proportion of young people who may be more likely to engage in criminalbehavior These findings suggest that urban environments may present uniquechallenges when it comes to reducing crime rates Ajaz Ahmad Malik (2016) statedthat urbanization can be advantageous as it allows for economies of scale, leading tothe growth and development of industries from an economic perspective However,from a social standpoint, urbanization has been linked to an increase in crime rates inlarge cities and urban areas While urbanization is not solely responsible for the rise incrime, there are various other factors closely associated with it that contribute to thistrend.

SECTION II: RESEARCH METHODOLOGY

2.1 Methodology

In this research, we use the quantitative approach to analyze the factors affectingCrime rates The OLS estimation method is applied to estimate the effects ofindependent variables on a dependent variable Our group processed the raw data andran the test using Stata/MP 15.0 software

We implement several data researching and collecting steps and make it into adata table in Excel used for analysis, hence the total number of observations is 363

2.2 Model specification

We suggest the following population regression model:

crimerate = β 0 + β 1 gdp + β 2 pop + β 3 urban + β 4 inflation + β 5

The variables list and the expected signs are summarized in the table 1 below

Table 1 Explanation of variables

Type Variables Name Description Unit

Expected sign of regression coefficient

Theeconomics

inflation

Inflation,consumerpricesindex

The inflation

equality Gini Index

Theeconomicsinequality orincomeinequality

legitimacy StateLegitimacy

Index

Thepopulation’sconfidencein

The Development Data Group of the World Bank organization, guided byprofessional standards in the collection, compilation and dissemination of data,provides us with quality and reliable sources of information Therefore, we chose tocollect data on the World Bank official website: Crime rates, Gross Domestic Growth,Population ages 15-64, Urban population growth, Inflation, Consumer price and GiniIndex across 34 selected G20 countries in 2010-2020 specifically

Another online tool that provides accessibility to systematic data is Fragile States

taken from Fragile States Index official website

2.3.2 Descriptive statistics and interpretation for each variable

We collected the data from all 34 countries that have the accessibility to data onG20 Statistics Moreover, 11 years from 2010 - 2020 are obtained for the sake of thisdataset In total, we have 363 observations, so our sample data represents thepopulation

Before analyzing the collected data, we will bring in a general description of themodel and the parameters by using the command sum in STATA The commandreveals the Observations (Obs), Mean, Standard Deviation (Std Dev.) as well asMinimum (Min) and Maximum (Max) values of the variables

The statistics of the regression model’s variables are summarized in the table 2below

Table 2 Summary statistics of the regression model’s variables

The mean rate of crime rates is 4.117448%, the standard deviation is 7.900481,the minimum amount is 0.24% and the maximum is 36.4167%.

Gross Domestic Production Growth

The average Gross Domestic Production Growth rate takes a value of about1.837162% with the standard deviation being 3.485995 The minimum rate is -11.32544% and the maximum is 24.37045%

Population ages 15-64

With the rate of population ages 15-64, it is 66.58341 on average, the standarddeviation is 2.769828 The minimum rate is 58.50111% and the maximum is73.2711%

Urban population growth

The mean of urban population growth rate is 0.6805662% This index has astandard deviation of 0.8603099, the minimum number is approximately -2.282468%and the maximum is around 3.255366%

Inflation, consumer prices index

This variable has a mean of 2.100161%, and the standard deviation is about2.182555 The minimum amount of inflation, consumer prices index is -2.096998%and the maximum one is 15.5344%

Gini Index

In terms of Gini Index, the average value is 0.3465439 with a standard deviation

of 0.0782781 The minimum and maximum amounts are 0.232% and 0.664%respectively

State Legitimacy

The average point of the population’s confidence in governance is about3.596143, distributed among 34 selected countries This point has the standarddeviation of 2.186204; the minimum number is approximately 0.5 point and themaximum is around 8.8 point

2.3.3 Correlation matrix between variables

The correlation coefficients among variables are summarized in the table 3

Table 3 Correlation coefficients among variables using command

crimerate gdp pop urban inflation equality legitimacy

gdp has a relatively low correlation coefficient of (-0.0527), and the minus sign

indicates a negative impact it has on crimerate, matched with the expected sign (-).

population has also a pretty low correlation coefficient of (+0.0739), and the

plus sign means that there is a positive effect between population and crimerate,

matched with the expected sign (+)

urbanization has a pretty high correlation coefficient of (+0.2918), and the plus

sign indicates positive relationship with crimerate, matched with the expected sign

(+)

inflation has a high correlation coefficient of (+0.4986), and it has a positive

impact on crimerate, matched with the expected sign (+).

equality has a very high correlation coefficient of (+0.8725), and the plus sign

indicates a positive effect on crimerate, matched with the expected sign (+).

legitimacy has a high correlation coefficient of (+0.3994), and it has a positive

impact on crimerate, unmatched with the expected sign (-).

The independent variables such as equality, urbanization, inflation, legitimacy

also have a high correlation coefficient with crimerate Whereas the correlation

coefficient that gdp regards is the smallest, which suggests an insignificant effect the growth rate of GDP has on crimerate.

The table indicates that there is no perfect multicollinearity among ourregressors, that is the correlation between two variables different from ±1 Thus, oursample data satisfy the assumption that there is no perfect multicollinearity

SECTION III: RESEARCH RESULTS AND IMPLIACATION

3.1 Choosing the Estimated Model

We use Stata 15 in order to examine and find out the most suitable choice for ourdataset among three models: Pooled OLS model (POLS) Fixed Effects model (FE) andRandom Effects model (RE) This is done by following the steps below

3.1.1 Breusch - Pagan test

We apply Breusch - Pagan test to choose between FE/RE or POLS for thesubstantial distinction over units

H 0 : no significant difference across units (no panel effect) [no existence of ai]

H 1 : significant difference across units (panel effect) [there is existence of ai]

xtreg crimerate gdp pop urban inflation equality legitimacy, re

xttest0

Breusch and Pagan Lagrangian multiplier test for random effects

crimerate[code,t] = Xb + u[code] + e[code,t]

Source: Stata’s result

From Stata’s result, it is clear that with a 5% level of significance, Prob > chibar2

= 0.0000 < 0.05, so we reject H0 In other words, we would rather use Random Effects

or Fixed Effects Model than POLS

xtreg crimerate gdp pop urban inflation equality legitimacy, fe

b = consistent under Ho and Ha; obtained from xtreg

B = inconsistent under Ha, efficient under Ho; obtained from xtreg

Test: Ho: difference in coefficients not systematic

chi2(6) = (b-B)'[(V_b-V_B)^(-1)](b-B)

= 87.48

Prob>chi2 = 0.0000

Source: Stata’s results

As a result of the Hausman test, with 5% level of significance, we have: Prob >chi2 = 0.0000 < 0.05 Therefore, we reject H0 or we can conclude that the FixedEffects model (FE) is chosen to estimate the data

3.2 Diagnosing the problems of the model

3.2.1 Testing for multicollinearity

Statistical consequences of multicollinearity include difficulties in testingindividual regression coefficients due to inflated standard errors Thus, you may beunable to declare an X variable significant even though (by itself) it has a strongrelationship with Y We use the command VIF in order to test whether the model hasmulticollinearity or not

reg crimerate gdp pop urban inflation equality legitimacy

Source: Stata’s results

Since Mean VIF = 1.31 < 10, we can conclude that the model has nomulticollinearity

3.2.2 Testing for heteroskedasticity

One of the assumptions made about residuals/errors in OLS regression is that theerrors have the same but unknown variance This is known as constant variance orhomoscedasticity When this assumption is violated, the problem is known asheteroscedasticity It has been shown that models involving a wide range of values aremore prone to heteroskedasticity because the differences between the smallest andlargest values are so significant

H 0: the model has homoscedasticity

H 1: the model has heteroscedasticity

xtreg crimerate gdp pop urban inflation equality legitimacy, fe

With 5% level of significance, we have: Prob>chi2 = 0.0000 < 0.05, so we reject

H0 or we can conclude that the Random Effects Model suffers from heteroskedasticity

3.2.3 Testing for serial correlation

xtserial crimerate gdp pop urban inflation equality legitimacy

Wooldridge test for autocorrelation in panel data

H0: no first-order autocorrelation

F( 1, 32) = 37.078

Prob > F = 0.0000

With a 5% level of significance, we have Prob>F = 0.0000< 0.05 Therefore, the

FE model suffers from auto-correlation

3.2.4 Testing for cross-section correlation

xtreg crimerate gdp pop urban inflation equality legitimacy, fe

xtcsd, pesaran abs

Pesaran's test of cross-sectional independence = 8.161, Pr = 0.0000

Average absolute value of the off-diagonal elements = 0.405

With a 5% level of significance, we have Pr = 0.0000< 0.05 Therefore, the FEmodel suffers from cross-section correlation

3.3 Fixing the model

From previous tests, we can see that the model is suffering from serialcorrelation, heteroskedasticity, and cross-sectional correlation Therefore, to fix them,

we will run our Fixed-effects model with Driscoll-Kraay standard errors We have theresult below:

xtscc crimerate gdp pop urban inflation equality legitimacy, fe

Regression with

Method: Fixed-effects

Source: Stata’s results

3.4 Hypothesis testing

In this section, tests will be conducted with a 5% level of significance (α = 0.05)

to examine if the model is statistically significant using the overall significance testand to examine if all independent variables do affect the dependent variable using thejoint significance test

According to the aforementioned results in the table above, the GDP growth rate,population ages 15-64, urbanization, and inflation rate are the significant variable forthe crime rates of G20 countries from 2010 to 2020 with the P-value of 0.001, 0.005,0.003, 0.021 correspondingly, whereas the income inequality rate and state legitimacyindex are individually insignificant in the model as the P-values are much higher thanthe significance level However, regarding previous research, income inequality rateand state legitimacy index are important factors that contribute significantly to thelevel of crimerates, and we have no multicollinearity in this model, so those factorsshould be added into the model

3.4.1 Test the overall significance of the observed multiple regression

The F-distribution is used to test hypotheses involving several independentvariables Hypotheses are as follows:

   The coefficient of determination R2 is 0.0889, which means that approximately8.89% of the variance of the dependent variable is explained by the variance of the

independent ones and the rest is the chance of other factors R2 indicates the goodness

of fit, how well a regression model fits a dataset The value of R2 is not considered to

be a high value and just rather fit, however, still significantly different from 0 andindicates statistically significant explanatory power as crime rates is also affected by avariety of other unexplainable factors with many observations in different time sets Inthis report, the main focus is to figure out if the independent variables affect thedependent variable, and the interpretation of a regression coefficient that is statisticallysignificant does not change based on the R-squared value

3.4.2 Test the individual significance: