according to the professor’s comments, you have only 3 hours to revise this paper, but no matter you changed how much, just gave me on time

Effect of Minimum Legal Drinking Age (MLDA) on crimes and alcohol consumption

Student Name:

Student Number:

Abstract

The aim of this study was to investigate the impact of Minimum Legal Drinking Age (MLDA) on the proportion of population drinking and crimes. In this case, data from National Health Interview Sample Adults File since 1997 to 2007 was adopted for the realization of the aims of the study. The data was analyzed using multiple linear regression and a series of graphical presentations. From the findings, it was noted that Minimum Legal Drinking Age (MLDA) had a negative effect on the proportion of alcohol consumption and a positive effect on the number of arrests from various crimes around the US.

Introduction

Minimum Legal Drinking Age (MLDA) is associated with different events around the US and the rest of the world. In this case, there has been concerns to understand the implications Minimum Legal Drinking Age (MLDA) presents on crime rates and alcoholic consumption across the US and the rest of the world.

*Research questions*

1. How much does the Minimum Legal Drinking Age (MLDA) reduce the proportion of the population that drinks?

2. How much does the MLDA reduce crime?

Data

The data used for this study was obtained from drug abuse and crime investigations around the US. It was generated from National Health Interview Sample Adults File since 1997 to 2007 on alcohol consumption and number of arrests due to crimes. Each of the variables presented from the data was either discrete or dummy in nature as noted from the analytical sections. The data was downloaded in a csv format from online sources with proper information on both alcoholic consumption, number of arrests and their associated factors over the past years. This data was later imported to STATA, to ease statistical analysis using the respective commands adopted for the creation of the presented analysis in a bid to answer the proposed research questions. Most of these changes were observed from different individuals over time as noted from the developed findings.

The data entailed five variables namely; the number of arrests, age, post21, birthday and alcohol consumption as a percentage. The variables are summarized in the table below;

Table 1: Variable definitions

Variable | Discrete, continous or dummy | Dependent or independent | Definition |

Number of arrests | Discrete | Dependent | The actual number of arrests made per 100,000 persons |

Alcohol consumption | Continous | Dependent | The degree of alcohol consumption measured in percentage |

Age | Discrete | Independent | The age of the respondent |

Post21 | Dummy | Independent | Whether the person had attained a legal age of 21 to buy and consume alcoholic drinks |

Birthday | Dummy | Independent | Whether the person had celebrated his or her 21st birthday |

Empirical methods

In the realization of the intended analysis from the data provided, both inferential and descriptive statistics were adopted. The main inferential statistics adopted for this case were the Independent variable (IV) estimate and Regression Discontinuity. In most cases, Regression Discontinuity measures are used in validation of the treatment effects across some balanced treatments that work in the same way as randomization. The need for this approach is majorly to obtain sharp discontinuity results during development of regression models at the point in which there exists a sharp discontinuity in the area in which a treatment is held and there is the need for crossing the discontinuity outcome since nothing is expected from the treatment. This is attributed as some change in magnitude between the dependent observations across the discontinuity to the existing treatment effect around that specific point.

It is clear that the RD approach works best in visualizing regression lines with discontinuity aimed at presenting the respective MLDA treatment effects based on both consumption of alcohol and arrests as they may occur. In making the desired figures, there is the need for informative and observable cases through predetermination of scales across the map for figuring the specified bin sizes and the range of the response variable across different analyses.

Figure 1 Band Size Graphs Overview

In the realization of a proper band size for running the respective variables, there is need for proper choice of the range and the bins. For instance, a bin size with a band moving from 19 to 23 years is appropriate since the median age that is assumed at 21 years was properly covered as illustrated from the figures. On the first figure, the range moves from 21 to 21.8 years as the second graph having a range from 19 to 23 years, the third graph showing a bin moving from 18 to 24 years as the last graph indicating a bin ranging from 18 to 26 years. As noted from the figures, bins ranging from 18 to 23 to 24 years presented drinking of alcohol as normally distributed as anticipated from the aims of the study. This demonstrates a proper linear distribution between drinking of alcohol and the individual’s age.

Figure 2 Bin Size Graphs Overview

From the graphs, the bins were changed from 10, 20, 30 and 80 days as noted with 12 data points created across the existing intervals over the period. The dots presented from the graphs present average points for the explanatory variables across the x-axis intervals noted as defined by the bin width. Using the bin width creates desired information for the figure that proves noisy in observing any presence of bias based on existence of narrow information that may distort the inferences.

In the case of the bins, we determine the dependent variable from the interval magnitudes associated with the respective outcome variables. This translates to estimating alcohol consumption as the dependent variable, defined based on the percentage drinking rates with its estimations ranging from 0.4 to 0.8 in observed value. This creates room to employ the dependent variable to assume variables ranging from 0.45 to 0.75 in order to make the observations quite focused on the anticipated threshold. Making the estimations for the next variable defined as total arrests ranges from 1500 to 1700 for every 10,000 individuals annually. As the outcome variable is defined in terms of the total arrests and there is the need for combining the estimations with other branches to a single figure to reduce chances of over-lapping from the regression lines and creation of confusion in the process, the arrests were set to lie between 0 to 2000. This was done to make them lie within the regression line.

The response variable for the analysis was set from 0 to 400 as the number of arrests for every 100,000 individuals within a year fell into the specified category with each arrest showing smaller magnitudes as compared to the total arrests. These changes are noted as the arrests pile along the bottom of the regression line, as influenced by MLDA. Despite the branches causing a number of observation ranges, there was the need to choose branches having smaller magnitudes as it enabled visual comparison to obtain variations in magnitude from the respective branches. In using RD in MLDA measurement effect in respect to both arrest rates and consumption of alcohol, two models were developed defined by the equations;

From the developed models, Arrest Ratesi is defined as the yearly arrests for every 10,000 persons from the provided dataset. Agei is the age to the 21st birthday, having negative values when the ages are closer to the birthday dates. Birthdayi variable is dichotomous, defining whether an individual is right on their 21st birthday with value of 1 and 0 when otherwise. Furthermore, Post21i is a dummy variable describing the possibility of a person passing their 21st birthday and allowed by law to purchase alcoholic drinks with a value of 1 when the Age is more than 21 an 0 when otherwise. This indicates whether the individual is of legal age to purchase the drinks from the respective sellers across different points within the region.

The second model defines alcohol consumption as the dependent variable. This variable (alcohol consumption) is termed as the rate of alcohol uptake as the other independent variables within the model similar to the ones adopted on the number of arrests’ model. This implies that there is no change in the independent variables as defined from the developed model with respect to the change in the dependent variable from the variation in the provided independent variables across different times. The essence of this model was to ascertain whether the independent variables have an effect on either the number of arrests of alcohol consumption as defined over time.

Moreover, it can be noted that the two regression equations employ quadratic polynomials to calculate the respective specifications. Across the two equations, is considered a quadratic function on Age variable, resulting in regression equation parabolas that define changes on both alcohol consumption and the number of arrests at the respective periods as noted from the dataset during and after the 21st birthday of the respondents involved in this study.

Results

On the developed regression model, the model for both number of arrests and alcohol consumption presented R-square of 0.514. This implied that the model explained 51.4% of the regression of both number of arrests and alcoholic consumption. In this model, it was clear that age to 21 had a negative effect on the dependent variables (β=-72.78), with a statistical significant effect on the respective model (p<0.01). Nonetheless, Post 21 presented a positive effect on both alcohol consumption and number of arrests (β=97.219). The parameter also proved statistically significant at 5% level of significance (β=97.219). Nonetheless, the interaction between Age and Post 21 presented a negative effect on both the number of arrests and alcoholic consumption (β=-20.019). The parameter, however, proved statistically insignificant at 5% level of significance (p>0.1). Square of the ages depicted a negative effect on both the number of arrests and alcoholic consumption (β=-24.002), with a statistical significant effect on the respective model. The interaction between square of age and post 21 presented a positive effect on both the number of arrests and alcoholic consumption (β=34.856), with a statistical significant effect on the specified model (p<0.01). Whether the persons had celebrated the 21st birthday also presented a positive effect on number of arrests and alcoholic consumption (β=627.932). The parameter proved statistically significant at 5% level of significance hence desired for the prediction of the number of arrests and proportion of individuals consuming alcohol over the years.

On the second model, it was noted that age to 21 had a positive effect to drive under influence (β=24.987). The parameter proved statistically significant at 1% level of significance (p<0.01). Furthermore, post 21 also presented a positive effect on driving under influence (β=52.758). The parameter also proved statistically significant at 5% level of significance as presented from the model (p<0.01). Age and post 21 however, presented a negative effect on driving under influence (β=-21.927). The parameter also proved statistically significant at 1% level of significance. Square of the age presented a negative effect on drive under influence (β=-2.507), showing a significant effect on driving under influence.

The square of age and post 21 presented a positive effect on driving under influence (β=2.977). The parameter proved statistically insignificant at 10% level of significance. Whether an individual has attained their 21st birthday also presented a positive effect on driving under influence (β=129.210). The parameter proved statistically significant at 1% level of significance as noted from the respective outcomes associated with the study. This model presented R-squared value of 0.943, an indication that 94.3% of the relationship between driving under influence and other covariates explained by the model. This presented a higher level of explanation between the response and the predictor variables.

On the third model, robbery model’s R-squared=0.719, an indication that 71.9% of the regression between number of robberies, age, post 21, age* post 21, age squared, age squared*post 21 and birthday was explained by the model. This presented a higher predictive power between the response and the predictor variables. Furthermore, age to 21 had a negative effect on the model (β=-2.901), with a significant effect on number of robberies (p<0.01). Nonetheless, post 21 presented a positive effect on the number of robberies reported (β=1.732). The parameter proved statistically significant at 5% level of significance as demonstrated from the respective model findings. The interaction between age and post 21 presented a negative effect on the number of robberies noted over the respective individuals within the model (β=-2.205). The parameter also proved statistically significant at 1% level of significance as presented within the model (p<0.01). Square of age presented a positive effect on the number of robberies (β=1.232) with a significant effect on the robberies as noted from the respective findings. Age squared and post 21 interaction presented a negative effect on the number of robberies as presented from the findings (β=-0.422) with an insignificant effect on the number of robberies. Whether an individual had attained their 21st birthday presented a positive (β=5.918) significant effect on the number of robberies (p<0.1).

On the fourth model, it was noted that liquor law violation model had R-squared value of 0.990, showing that the model explained 99.0% of the regression. This gave a very high explanation between the predictor and the response variables over time. Age presented a negative effect on the liquor law violations (β=-33.608) with a significant effect on the respective liquor law violations (p<0.01). Post 21 presented a negative effect on the liquor law violations (β=-64.973). The parameter proved statistically significant at 1% level of significance (p<0.01), an indication that the parameter was desired for prediction of the liquor law violations over time. Moreover, Age and post 21 presented a positive (β=26.544) significant (p<0.01) effect on the liquor law violations over time. Furthermore, the square of age presented a negative (β=7.736) significant (p<0.01) effect on the liquor law violations as demonstrated from the figure. This implied that a rise in the square of age was associated with a decline in the liquor law violations over time as noted from the analysis. Age squared and post 21 presented a positive (β=9.548) significant effect (p<0.01) on the liquor law violations over time as presented from the model outcomes. Furthermore, it was noted that birthday had a positive (β=5.134) significant effect (p<0.01) effect on the liquor law violations over time as noted from the findings.

Simple assault model presented age to 21 as having a negative effect on the simple assaults (β=1.247) with an insignificant effect on the number of simple assaults. It was noted that post 21 had a positive significant effect on the respective model (p<0.01). Age squared and post 21 and birthday also presented a positive significant effect on the number of assaults noted from the statistical outcomes.

Table 2 Regression Table for Branch Causes

(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |

VARIABLES | All | Drive under influence | Robbery | Liquor Lawviolation | Simple assault | Aggravatedassault | Vagrancy & Disorder | Drunk risk |

Age to 21 | -72.782*** | 24.987*** | -2.901*** | -33.608*** | -1.247 | 1.446 | -1.505*** | 4.224 |

(9.431) | (2.807) | (0.694) | (1.189) | (1.111) | (1.181) | (0.580) | (2.705) | |

Post 21 | 97.219*** | 52.758*** | 1.732*** | -64.973*** | 5.247*** | 4.045*** | 2.012*** | 33.690*** |

(5.786) | (1.722) | (0.426) | (0.729) | (0.682) | (0.724) | (0.356) | (1.659) | |

Age & Post 21 | -20.019 | -21.927*** | -2.205** | 26.544*** | -1.091 | -2.823* | -0.830 | -19.916*** |

(13.351) | (3.974) | (0.983) | (1.683) | (1.573) | (1.672) | (0.821) | (3.829) | |

Squared Age | -24.002*** | -2.507* | 1.232*** | -7.736*** | -0.115 | -0.060 | 0.509* | -0.078 |

(4.560) | (1.357) | (0.336) | (0.575) | (0.537) | (0.571) | (0.280) | (1.308) | |

Age Squared & Post 21 | 34.856*** | 2.977 | -0.422 | 9.548*** | -0.145 | 0.599 | -0.109 | 4.538** |

(6.460) | (1.923) | (0.475) | (0.814) | (0.761) | (0.809) | (0.397) | (1.853) | |

Birthday | 627.932*** | 129.246*** | 5.198* | 5.134 | 38.005*** | 23.699*** | 14.915*** | 193.944*** |

(36.958) | (11.000) | (2.720) | (4.659) | (4.356) | (4.627) | (2.273) | (10.601) | |

Constant | 1,542.960*** | 194.210*** | 24.546*** | 81.164*** | 54.862*** | 65.266*** | 15.116*** | 104.358*** |

(4.090) | (1.217) | (0.301) | (0.516) | (0.482) | (0.512) | (0.251) | (1.173) | |

Observations | 1,460 | 1,460 | 1,460 | 1,460 | 1,460 | 1,460 | 1,460 | 1,460 |

R-squared | 0.514 | 0.943 | 0.719 | 0.990 | 0.148 | 0.249 | 0.340 | 0.688 |

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Conclusion

According to the analysis, it was noted that the Minimum Legal Drinking Age (MLDA) had a negative effect on the proportion of drinks consumed by the respective consumers from the study. This indicated that individuals outside the legal drinking age were not required to consume alcohol as noted from the study. The pattern depicts that most of alcoholic consumers are beyond the ages of 21 years. Nonetheless, it was also noted that Minimum Legal Drinking Age (MLDA) had a negative effect on the number arrests made from the study. This showed that ages above 21 years were prone to more arrests from different crimes as compared to those below the legal age of 21 years. These differences noted that Minimum Legal Drinking Age (MLDA) had a negative effect on the proportion of alcohol consumption and a positive effect on the number of arrests from various crimes around the US.

.

6

.

6

2

.

6

4

.

6

6

.

6

8

.

7

D

r

i

n

k

A

l

c

o

h

o

l

2121.221.421.621.8

Age

Graph1: bin width = 30 days

.

4

5

.

5

.

5

5

.

6

.

6

5

.

7

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph2: bin width = 30 days

.

2

.

4

.

6

.

8

1

D

r

i

n

k

A

l

c

o

h

o

l

18202224

Age

Graph3: bin width = 30 days

.

2

.

4

.

6

.

8

1

D

r

i

n

k

A

l

c

o

h

o

l

1820222426

Age

Graph4: bin width = 30 days

.

4

.

5

.

6

.

7

.

8

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph1: bin width = 10 days

.

4

5

.

5

.

5

5

.

6

.

6

5

.

7

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph2: bin width = 20 days

.

4

5

.

5

.

5

5

.

6

.

6

5

.

7

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph3: bin width = 30 days

.

5

.

5

5

.

6

.

6

5

.

7

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph4: bin width = 80 days

Effect of Minimum Legal Drinking Age (MLDA) on crimes and alcohol consumption

Student Name:

Student Number:

Abstract

The aim of this study was to investigate the impact of Minimum Legal Drinking Age (MLDA) on the proportion of population drinking and crimes. In this case, data from National Health Interview Sample Adults File since 1997 to 2007 was adopted for the realization of the aims of the study. The data was analyzed using multiple linear regression and a series of graphical presentations. From the findings, it was noted that Minimum Legal Drinking Age (MLDA) had a negative effect on the proportion of alcohol consumption and a positive effect on the number of arrests from various crimes around the US.

Introduction

Minimum Legal Drinking Age (MLDA) is associated with different events around the US and the rest of the world. In this case, there has been concerns to understand the implications Minimum Legal Drinking Age (MLDA) presents on crime rates and alcoholic consumption across the US and the rest of the world.

*Research questions*

1. How much does the Minimum Legal Drinking Age (MLDA) reduce the proportion of the population that drinks?

2. How much does the MLDA reduce crime?

Data

The data used for this study was obtained from drug abuse and crime investigations around the US. It was generated from National Health Interview Sample Adults File since 1997 to 2007 on alcohol consumption and number of arrests due to crimes. Each of the variables presented from the data was either discrete or dummy in nature as noted from the analytical sections. The data was downloaded in a csv format from online sources with proper information on both alcoholic consumption, number of arrests and their associated factors over the past years. This data was later imported to STATA, to ease statistical analysis using the respective commands adopted for the creation of the presented analysis in a bid to answer the proposed research questions. Most of these changes were observed from different individuals over time as noted from the developed findings.

The data entailed five variables namely; the number of arrests, age, post21, birthday and alcohol consumption as a percentage. The variables are summarized in the table below;

Table 1: Variable definitions

Variable | Discrete, continous or dummy | Dependent or independent | Definition |

Number of arrests | Discrete | Dependent | The actual number of arrests made per 100,000 persons |

Alcohol consumption | Continous | Dependent | The degree of alcohol consumption measured in percentage |

Age | Discrete | Independent | The age of the respondent |

Post21 | Dummy | Independent | Whether the person had attained a legal age of 21 to buy and consume alcoholic drinks |

Birthday | Dummy | Independent | Whether the person had celebrated his or her 21st birthday |

Empirical methods

In the realization of the intended analysis from the data provided, both inferential and descriptive statistics were adopted. The main inferential statistics adopted for this case were the Independent variable (IV) estimate and Regression Discontinuity. In most cases, Regression Discontinuity measures are used in validation of the treatment effects across some balanced treatments that work in the same way as randomization. The need for this approach is majorly to obtain sharp discontinuity results during development of regression models at the point in which there exists a sharp discontinuity in the area in which a treatment is held and there is the need for crossing the discontinuity outcome since nothing is expected from the treatment. This is attributed as some change in magnitude between the dependent observations across the discontinuity to the existing treatment effect around that specific point.

It is clear that the RD approach works best in visualizing regression lines with discontinuity aimed at presenting the respective MLDA treatment effects based on both consumption of alcohol and arrests as they may occur. In making the desired figures, there is the need for informative and observable cases through predetermination of scales across the map for figuring the specified bin sizes and the range of the response variable across different analyses.

Figure 1 Band Size Graphs Overview

In the realization of a proper band size for running the respective variables, there is need for proper choice of the range and the bins. For instance, a bin size with a band moving from 19 to 23 years is appropriate since the median age that is assumed at 21 years was properly covered as illustrated from the figures. On the first figure, the range moves from 21 to 21.8 years as the second graph having a range from 19 to 23 years, the third graph showing a bin moving from 18 to 24 years as the last graph indicating a bin ranging from 18 to 26 years. As noted from the figures, bins ranging from 18 to 23 to 24 years presented drinking of alcohol as normally distributed as anticipated from the aims of the study. This demonstrates a proper linear distribution between drinking of alcohol and the individual’s age.

Figure 2 Bin Size Graphs Overview

From the graphs, the bins were changed from 10, 20, 30 and 80 days as noted with 12 data points created across the existing intervals over the period. The dots presented from the graphs present average points for the explanatory variables across the x-axis intervals noted as defined by the bin width. Using the bin width creates desired information for the figure that proves noisy in observing any presence of bias based on existence of narrow information that may distort the inferences.

In the case of the bins, we determine the dependent variable from the interval magnitudes associated with the respective outcome variables. This translates to estimating alcohol consumption as the dependent variable, defined based on the percentage drinking rates with its estimations ranging from 0.4 to 0.8 in observed value. This creates room to employ the dependent variable to assume variables ranging from 0.45 to 0.75 in order to make the observations quite focused on the anticipated threshold. Making the estimations for the next variable defined as total arrests ranges from 1500 to 1700 for every 10,000 individuals annually. As the outcome variable is defined in terms of the total arrests and there is the need for combining the estimations with other branches to a single figure to reduce chances of over-lapping from the regression lines and creation of confusion in the process, the arrests were set to lie between 0 to 2000. This was done to make them lie within the regression line.

The response variable for the analysis was set from 0 to 400 as the number of arrests for every 100,000 individuals within a year fell into the specified category with each arrest showing smaller magnitudes as compared to the total arrests. These changes are noted as the arrests pile along the bottom of the regression line, as influenced by MLDA. Despite the branches causing a number of observation ranges, there was the need to choose branches having smaller magnitudes as it enabled visual comparison to obtain variations in magnitude from the respective branches. In using RD in MLDA measurement effect in respect to both arrest rates and consumption of alcohol, two models were developed defined by the equations;

From the developed models, Arrest Ratesi is defined as the yearly arrests for every 10,000 persons from the provided dataset. Agei is the age to the 21st birthday, having negative values when the ages are closer to the birthday dates. Birthdayi variable is dichotomous, defining whether an individual is right on their 21st birthday with value of 1 and 0 when otherwise. Furthermore, Post21i is a dummy variable describing the possibility of a person passing their 21st birthday and allowed by law to purchase alcoholic drinks with a value of 1 when the Age is more than 21 an 0 when otherwise. This indicates whether the individual is of legal age to purchase the drinks from the respective sellers across different points within the region.

The second model defines alcohol consumption as the dependent variable. This variable (alcohol consumption) is termed as the rate of alcohol uptake as the other independent variables within the model similar to the ones adopted on the number of arrests’ model. This implies that there is no change in the independent variables as defined from the developed model with respect to the change in the dependent variable from the variation in the provided independent variables across different times. The essence of this model was to ascertain whether the independent variables have an effect on either the number of arrests of alcohol consumption as defined over time.

Moreover, it can be noted that the two regression equations employ quadratic polynomials to calculate the respective specifications. Across the two equations, is considered a quadratic function on Age variable, resulting in regression equation parabolas that define changes on both alcohol consumption and the number of arrests at the respective periods as noted from the dataset during and after the 21st birthday of the respondents involved in this study.

Results

On the developed regression model, the model for both number of arrests and alcohol consumption presented R-square of 0.514. This implied that the model explained 51.4% of the regression of both number of arrests and alcoholic consumption. In this model, it was clear that age to 21 had a negative effect on the dependent variables (β=-72.78), with a statistical significant effect on the respective model (p<0.01). Nonetheless, Post 21 presented a positive effect on both alcohol consumption and number of arrests (β=97.219). The parameter also proved statistically significant at 5% level of significance (β=97.219). Nonetheless, the interaction between Age and Post 21 presented a negative effect on both the number of arrests and alcoholic consumption (β=-20.019). The parameter, however, proved statistically insignificant at 5% level of significance (p>0.1). Square of the ages depicted a negative effect on both the number of arrests and alcoholic consumption (β=-24.002), with a statistical significant effect on the respective model. The interaction between square of age and post 21 presented a positive effect on both the number of arrests and alcoholic consumption (β=34.856), with a statistical significant effect on the specified model (p<0.01). Whether the persons had celebrated the 21st birthday also presented a positive effect on number of arrests and alcoholic consumption (β=627.932). The parameter proved statistically significant at 5% level of significance hence desired for the prediction of the number of arrests and proportion of individuals consuming alcohol over the years.

On the second model, it was noted that age to 21 had a positive effect to drive under influence (β=24.987). The parameter proved statistically significant at 1% level of significance (p<0.01). Furthermore, post 21 also presented a positive effect on driving under influence (β=52.758). The parameter also proved statistically significant at 5% level of significance as presented from the model (p<0.01). Age and post 21 however, presented a negative effect on driving under influence (β=-21.927). The parameter also proved statistically significant at 1% level of significance. Square of the age presented a negative effect on drive under influence (β=-2.507), showing a significant effect on driving under influence.

The square of age and post 21 presented a positive effect on driving under influence (β=2.977). The parameter proved statistically insignificant at 10% level of significance. Whether an individual has attained their 21st birthday also presented a positive effect on driving under influence (β=129.210). The parameter proved statistically significant at 1% level of significance as noted from the respective outcomes associated with the study. This model presented R-squared value of 0.943, an indication that 94.3% of the relationship between driving under influence and other covariates explained by the model. This presented a higher level of explanation between the response and the predictor variables.

On the third model, robbery model’s R-squared=0.719, an indication that 71.9% of the regression between number of robberies, age, post 21, age* post 21, age squared, age squared*post 21 and birthday was explained by the model. This presented a higher predictive power between the response and the predictor variables. Furthermore, age to 21 had a negative effect on the model (β=-2.901), with a significant effect on number of robberies (p<0.01). Nonetheless, post 21 presented a positive effect on the number of robberies reported (β=1.732). The parameter proved statistically significant at 5% level of significance as demonstrated from the respective model findings. The interaction between age and post 21 presented a negative effect on the number of robberies noted over the respective individuals within the model (β=-2.205). The parameter also proved statistically significant at 1% level of significance as presented within the model (p<0.01). Square of age presented a positive effect on the number of robberies (β=1.232) with a significant effect on the robberies as noted from the respective findings. Age squared and post 21 interaction presented a negative effect on the number of robberies as presented from the findings (β=-0.422) with an insignificant effect on the number of robberies. Whether an individual had attained their 21st birthday presented a positive (β=5.918) significant effect on the number of robberies (p<0.1).

On the fourth model, it was noted that liquor law violation model had R-squared value of 0.990, showing that the model explained 99.0% of the regression. This gave a very high explanation between the predictor and the response variables over time. Age presented a negative effect on the liquor law violations (β=-33.608) with a significant effect on the respective liquor law violations (p<0.01). Post 21 presented a negative effect on the liquor law violations (β=-64.973). The parameter proved statistically significant at 1% level of significance (p<0.01), an indication that the parameter was desired for prediction of the liquor law violations over time. Moreover, Age and post 21 presented a positive (β=26.544) significant (p<0.01) effect on the liquor law violations over time. Furthermore, the square of age presented a negative (β=7.736) significant (p<0.01) effect on the liquor law violations as demonstrated from the figure. This implied that a rise in the square of age was associated with a decline in the liquor law violations over time as noted from the analysis. Age squared and post 21 presented a positive (β=9.548) significant effect (p<0.01) on the liquor law violations over time as presented from the model outcomes. Furthermore, it was noted that birthday had a positive (β=5.134) significant effect (p<0.01) effect on the liquor law violations over time as noted from the findings.

Simple assault model presented age to 21 as having a negative effect on the simple assaults (β=1.247) with an insignificant effect on the number of simple assaults. It was noted that post 21 had a positive significant effect on the respective model (p<0.01). Age squared and post 21 and birthday also presented a positive significant effect on the number of assaults noted from the statistical outcomes.

Table 2 Regression Table for Branch Causes

(1) | (2) | (3) | (4) | (5) | (6) | (7) | (8) | |

VARIABLES | All | Drive under influence | Robbery | Liquor Lawviolation | Simple assault | Aggravatedassault | Vagrancy & Disorder | Drunk risk |

Age to 21 | -72.782*** | 24.987*** | -2.901*** | -33.608*** | -1.247 | 1.446 | -1.505*** | 4.224 |

(9.431) | (2.807) | (0.694) | (1.189) | (1.111) | (1.181) | (0.580) | (2.705) | |

Post 21 | 97.219*** | 52.758*** | 1.732*** | -64.973*** | 5.247*** | 4.045*** | 2.012*** | 33.690*** |

(5.786) | (1.722) | (0.426) | (0.729) | (0.682) | (0.724) | (0.356) | (1.659) | |

Age & Post 21 | -20.019 | -21.927*** | -2.205** | 26.544*** | -1.091 | -2.823* | -0.830 | -19.916*** |

(13.351) | (3.974) | (0.983) | (1.683) | (1.573) | (1.672) | (0.821) | (3.829) | |

Squared Age | -24.002*** | -2.507* | 1.232*** | -7.736*** | -0.115 | -0.060 | 0.509* | -0.078 |

(4.560) | (1.357) | (0.336) | (0.575) | (0.537) | (0.571) | (0.280) | (1.308) | |

Age Squared & Post 21 | 34.856*** | 2.977 | -0.422 | 9.548*** | -0.145 | 0.599 | -0.109 | 4.538** |

(6.460) | (1.923) | (0.475) | (0.814) | (0.761) | (0.809) | (0.397) | (1.853) | |

Birthday | 627.932*** | 129.246*** | 5.198* | 5.134 | 38.005*** | 23.699*** | 14.915*** | 193.944*** |

(36.958) | (11.000) | (2.720) | (4.659) | (4.356) | (4.627) | (2.273) | (10.601) | |

Constant | 1,542.960*** | 194.210*** | 24.546*** | 81.164*** | 54.862*** | 65.266*** | 15.116*** | 104.358*** |

(4.090) | (1.217) | (0.301) | (0.516) | (0.482) | (0.512) | (0.251) | (1.173) | |

Observations | 1,460 | 1,460 | 1,460 | 1,460 | 1,460 | 1,460 | 1,460 | 1,460 |

R-squared | 0.514 | 0.943 | 0.719 | 0.990 | 0.148 | 0.249 | 0.340 | 0.688 |

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Conclusion

According to the analysis, it was noted that the Minimum Legal Drinking Age (MLDA) had a negative effect on the proportion of drinks consumed by the respective consumers from the study. This indicated that individuals outside the legal drinking age were not required to consume alcohol as noted from the study. The pattern depicts that most of alcoholic consumers are beyond the ages of 21 years. Nonetheless, it was also noted that Minimum Legal Drinking Age (MLDA) had a negative effect on the number arrests made from the study. This showed that ages above 21 years were prone to more arrests from different crimes as compared to those below the legal age of 21 years. These differences noted that Minimum Legal Drinking Age (MLDA) had a negative effect on the proportion of alcohol consumption and a positive effect on the number of arrests from various crimes around the US.

.

6

.

6

2

.

6

4

.

6

6

.

6

8

.

7

D

r

i

n

k

A

l

c

o

h

o

l

2121.221.421.621.8

Age

Graph1: bin width = 30 days

.

4

5

.

5

.

5

5

.

6

.

6

5

.

7

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph2: bin width = 30 days

.

2

.

4

.

6

.

8

1

D

r

i

n

k

A

l

c

o

h

o

l

18202224

Age

Graph3: bin width = 30 days

.

2

.

4

.

6

.

8

1

D

r

i

n

k

A

l

c

o

h

o

l

1820222426

Age

Graph4: bin width = 30 days

.

4

.

5

.

6

.

7

.

8

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph1: bin width = 10 days

.

4

5

.

5

.

5

5

.

6

.

6

5

.

7

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph2: bin width = 20 days

.

4

5

.

5

.

5

5

.

6

.

6

5

.

7

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph3: bin width = 30 days

.

5

.

5

5

.

6

.

6

5

.

7

D

r

i

n

k

A

l

c

o

h

o

l

1920212223

Age

Graph4: bin width = 80 days