I have just finished writing a final report of my research and statistical modeling analysis. I have included the introduction, methodology, results, analysis, and conclusion sections from the paper below with the regression outputs and residual plots omitted (because of formatting issues with this blog). Anyone interested in learning more, reading the full report, or seeing the full bibliography, can contact me at email@example.com or Professor Hausman at firstname.lastname@example.org. Overall, this project was extremely fulfilling and has certainly shown me how much fun it is to be a full-time academic. I cannot express how grateful I am that I was able to spend my summer and receive income to study this topic and hope that this field will continue to remain in the public eye, especially considering how much of our taxpayer money is at stake.
Any unauthorized redistribution of this paper or any of its contents without my express permission is prohibited.
In May 2014, Uruguay made history by becoming the first nation on earth to fully legalize and regulate marijuana for adult recreational use. This headline would have been thought a fantasy if it had been published only a few years prior, but a closer study of modern narcotics policy around the world reveals that Uruguay is far from being the only country to have broken its commitments to enforcing drug prohibition, an ideology that has come to dominate the policies of most nations on Earth. In point of fact two U.S. states, Colorado and Washington, passed ballot initiatives in the fall of 2012 that created legal, regulated recreational marijuana markets within their respective states, flouting the Controlled Substances Act of 1970, which classifies marijuana (among many other substances) as a schedule 1 narcotic that is illegal to grow, possess, sell, and consume. Though it remains to be seen whether the federal government will intervene to prevent these legal markets from existing, the Obama administration has largely remained inactive and has not made any direct move to challenge these laws in federal court as of the writing of this paper in July of 2014.
Perhaps even more surprising is that Portugal passed laws that decriminalized the possession and use of all illicit narcotics back in 2001 (Hughes & Stevens 2012). Though there remains some dispute about the true effects of ending harsh prohibitive policies surrounding the use of narcotics, there is a growing body of evidence that harm-reductionist strategies, like those applied in Portugal, can reduce the overall costs and burdens of societal drug abuse (Hughes & Stevens 2012). There also appears to be a growing number of nations, particularly in Europe and South America, that have taken steps to liberalize their handling of narcotics consumption, mostly through the decriminalization of marijuana (Wade 2009).
When considering why nations may be choosing to change their drug laws, it appears that a primary motivator is likely the existing costs associated with enforcing narcotics prohibition. These costs can be demonstrated in a plethora of different ways, but accurate quantification is often difficult because of unavailable data and the large variety of costs that are tenuously measured and reported by many separate organizations. One of the more reliable ways to measure a portion of these costs is through dollars budgeted by governments to maintain and enforce narcotics laws. The United States remains one of the biggest proponents and the largest financier of the international War on Drugs that was primarily implemented through three international treaties: The Single Convention on Narcotic Drugs (1961), the Convention on Psychotropic Substances (1971), and the Convention Against Illicit Trafficking of Narcotic Drugs and Psychotropic Substances (1988) (Ford, B. A. 2013). According to the Office of National Drug Control Policy, fifteen billion US dollars were spent by the United States federal government in 2010 alone to finance the War on Drugs. This number, though enormous by itself, does not even begin to account for the human costs that have arisen from mass incarceration and through violence perpetrated by black market criminal organizations. In the face of these overwhelming quantification problems, measuring the changes in expenditures to world governments becomes a necessary and useful proxy to measure a portion of the costs of international narcotics prohibition.
Though alternatives like decriminalization and legalization have long been suggested as viable policies, there simply were not enough independent cases that existed to lend insight into how these types of policies would compare to the status quo until recently. A recently published report by the London School of Economics titled The Economics of Drug Policy states in its introduction: “A more thorough cost benefit analysis of the merits of prohibition relative to the costs of enforcement… is required for a global cooperative framework. From this analysis a better appreciation of regulatory options and potential for experimentation and readjustment of resources can be decided” (Collins et all 2014). Because of these recent policy shifts and in conjunction with changing attitudes, it is likely that academics and advocates will have an increasing array of international data available to study alternative narcotics policies in the near future. For now, the primary concern of this research project is to examine the links between spending and drug policy outcomes in different nations of the world and to statistically model conditions that may be leading world governments to move away from costly prohibitive policies and toward more liberal models of control.
Approaching the modeling and quantification of narcotics policy and its costs presents a number of serious methodological challenges. In many cases drug related expenditure is distributed across multiple government sectors like health, defense, border security, etc. that often do not include item or project specific information that is required for detailed statistical analysis. In the case of international studies like this one, it is theoretically difficult to compare different countries because each uses their own methodologies when collecting data and can choose not to release relevant data to the public. Even more detrimental to this type of research is an existing data gap on expenditures: nations with some of the most hardline prohibitive narcotics policies publish little to nothing on drug-related spending (Count the Costs 2014).
Despite these challenges, there are a number of useful and reliable databases run by public and private organizations that can provide good data on many societal and economic variables that in turn can help more accurately predict outcomes in narcotics policy for a specific nation. For this project these databases were essential and provided the type of substantive data that is necessary to pursue statistical modeling. Perhaps most essentially, the dependent variable within these models had to include synthesized law data that could rate nations on a scale based on the type of narco-policies their governments have codified and implemented.
This paper makes use of an online database known as the Freedom Index that is managed by a libertarian think-tank which releases data that synthesizes information on current laws within nations and then ranks them based on their relative strictness or liberality (Drug Freedom Index 2014). The Drug Freedom Index subset database, an annually updated index that contains nation-specific rankings based on the harshness of their respective narcotics policies, provided a quantitative basis for a useful and robust policy variable for statistical modeling. Their raw variable, herein referred to as the 10-Point Policy Variable, was scaled as follows:
1 = Drugs are illegal with a death penalty punishment codified in their laws (“on the books”).
1.75 = Drugs are illegal with a death penalty codified, but never applied.
2 = Drugs are illegal and penalties for possession regularly involve imprisonment.
3 = Drugs are illegal and fines are applied to those convicted of possession.
4 = Drugs are generally illegal, medical and personal marijuana use are unofficially tolerated.
5 = Drugs are illegal, medical marijuana is legal.
5.5 = Drugs are generally illegal, marijuana unofficially allowed (decriminalization).
6 = Personal possession, use, and cultivation of marijuana is legal.
7 = Sale of non-medicinal marijuana is regulated and taxed.
8 = A hallucinogenic drug, hash and/or marijuana are legal.
9 = Hallucinogenic substances, hash, and marijuana are generally legal to possess.
10 = All psychotropic substances are legal to possess.
Though this data proved useful in itself as the dependent variable of interest, more quantitative approaches using ordered logistical models would require a more simplified ordered variable, especially because these laws tend to be highly variable causing several categories within the 10-Point scale to have only a single or no nation coded for them. As a result, the raw variable is translated into a 4-category ordered variable, herein referred to as the 4-Point Policy Variable. This variable made full use of the data from the Drug Freedom Index and reorganized the data according to the following scale:
1 = ‘Prohibition:’ Contains all nations coded from 1-3 originally
2 = ‘Reduced Penalties:’ Contains all nations coded 4-6 originally (generally no prison sentence)
3 = ‘Non-Enforcement:’ Contains all nations coded 6-7 originally (greater decriminalization)
4 = ‘Regulation:’ Contains all nations coded 7-10 originally (1 regulated narco-market exists)
The new variable reduced the number of categories causing the sample size for each remaining category to rise which produced a less specific, but broader-banded variable that is deemed more appropriate for an ordered logistical regression analysis. However, the 10-Point variable remained relevant and proved preferable to the 4-Point when using OLS linear estimation (See Results pp.11-12). Both scales were tested and the statistical models presented in this paper make use of both as the dependent variables.
The main hypothesized relationship of this study was that there existed a strong association between narco-policies internationally and the amount of expenditures devoted to enforcing them. Unfortunately, as stated previously, large data gaps arising from world governments neglecting to publish relevant expenditure data prevented a more thorough and quantitative analysis of individual nation-state expenditures relating to drug policy choice. However, because the United States sends a large amount of funds in the form of grants to foreign nations under the pretext of counter-narcotics interdiction and law enforcement, measuring the flow of these dollars to individual countries can provide a useful proxy measurement of this effect. This becomes even more plausible when one considers the enormous relative cost to less developed and poorer nations for militarizing their police forces and incarcerating their offending populations. Between 2009 and 2011, the United States provided an estimated $13.9 billion for foreign police assistance (Johnson, Foreign Police 2012). The two main federal departments that handle the processing and allocation for approximately 98% of these enforcement grants are the US Department of State and the Department of Defense (Johnson, Foreign Police 2012).
More methodological challenges presented themselves when it was determined that nation-specific data on the Defense Department’s counter-narcotics expenditures were completely unavailable and unpublished. Though there are plenty of large figures and even some nation-specific numbers for “security” spending, this type of data was determined to be too ambiguous and tenuous to be included within the dataset. Future research being conducted on these types of expenditures should make great efforts to include this missing data in their analyses.
Fortunately, the State Department’s Bureau of International Narcotics and Law Enforcement Affairs has published their nation-specific accounting of grant issuances for the years 2011(actual), 2012(estimated), and 2013(requested) (US State Department 2014). This data was used to form the variable known as ‘enforcespen’ within the dataset and provides dollar amounts issued by the Bureau in 1000’s of dollars. For all nations the most recent annual data was used, with 2011 numbers being used only when more recent data was unavailable. The sample for the variable was 61 cases, which would contribute to limiting the estimated models’ overall sample size when it was included (See Results pp.12-13).
In order to better control for the effects of the enforcement grants within the statistical models, a number of additional independent and control variables were collected for inclusion and experimentation:
- Unemployment Rate – 2012 International data courtesy of the World Bank
- Educational Spending as % of GDP – 2013 International data courtesy of the UN Human Development Report (United Nations 2013)
- Military Spending as % of GDP – 2013 UNHDR (United Nations 2013)
- Presence of a Military Conflict – Courtesy of Uppsala Conflict Data Program; nation = 1 if conflict resulted in greater than 25 deaths during 2013-Present.
- Nation is Western – Loosely interpreted from Huntington’s classification of western culture’s influence and roots in Clash of Civilizations (Huntington 1996). Nation = 1 if considered western.
- GDP per Capita – UNHDR (United Nations 2013)
- Percentage of Population Who Trust Their Government – UNHDR (United Nations 2013)
- Total Population – 2012 numbers, Courtesy of the World Bank
- Nation is a Major Narcotics Producer/Trafficker – From 2014 report from the Office of the Press Secretary (White House 2014). Nation = 1 if named by the executive branch to have met criteria to be declared a major illicit narcotics producer or trafficker (See Appendix Section 2).
These variables are used exclusively to test hypotheses of significance and association within this paper’s presented models of narcotics policy outcomes internationally.
Because there was a hypothesized linear relationship between spending and policy, OLS regressions were run using both the 10-Point and 4-Point variables as the dependent variable. When an ordered logistical model was deemed a more specific fit for these data and in conjunction with a limited sample size due to data unavailability (see Results pp.12-13), 4 additional ologit models were constructed to better investigate and support the results of the linear estimation. To test for the proportional odds assumptions that must hold for an ordered logistical regression, likelihood ratio tests were run on all 4 of the ologit models. All of the models passed this test, indicating that the proportional odds assumption is not violated.
Regression Results and Preliminary Analysis
4-Point Policy Variable OLS Regression
This simple OLS model reveals some basic relationships among the primary variables that are more thoroughly examined in following models (For a summary of significant variables and model coefficients across all models see Appendix Section 1). Enforcement grants, educational expenditures as a percentage of GDP, and military spending as a percentage of GDP are all highly significant at the .05 level, while being a western nation remains significant at the .10 level. Enforcement grants are positively associated with a more liberal policy, holding all other variables constant. This model predicts a very marginal increase in a nation’s liberalization of narcotics control with a one percentage point increase in enforcement grant dollars received, holding all variables constant. The same is true for educational expenditures, though the magnitude of association is slightly larger. Military spending is negatively associated with liberal policy while being in a western nation is positively associated. These relationships are examined more thoroughly when using the 10-Point policy variable because it functions more precisely as a continuous variable scale meaning OLS is a more appropriate estimation technique.
10-Point Policy Variable OLS Regression
Using the same variables as the previous model, the 10-point OLS regression reveals a very similar picture. Enforcement grants remain highly significant and its positive association with more liberal policy is greater than that predicted under the 4-point model. Both education and military spending remain significant at the .05 level and have beta coefficients that are consistent in direction (positive and negative respectively) with the other OLS model. Being a western nation remains significant at the .10 level and has a relatively large positive impact on more liberal policy. The regression residuals also behave much better under this model (see residual plot above).
Small Sample 4-Point Ordered Logistical Model
Under an ordered logistical model the use of the 4-point policy variable allows for a more precise model estimation based on a method of maximum likelihood. However, because of the limitations in the data, the sample size drops significantly when trying to predict the effect of enforcement grant dollars on policy. This is largely because the number of cases is limited by the unavailability of Department of Defense budget-specific data on counter-narcotics outlays to foreign law enforcement, leading their enormous share of this type of expenditure to be unaccounted for within these models (See Methodology pp.8). Any future research using these data should account for these additional grant funds when they become available.
As a result and in order to better demonstrate these relationships using ordered logistical models, two models for each policy variable are presented, one with a smaller sample that includes the enforcement spending variable and the other with a larger sample and no enforcement spending variable (see Methodology pp.9). It should be noted that based on the p-values of the chi-square test, all of these ordered logistical models are statistically significant at the .05 level.
The 4-point small sample model (see above) reinforces much of what the linear estimations predict. Specifically, enforcement grants, education, and military spending are all significant at the .05 level. Education and Enforcement are both positively associated with more liberal policy around the world while military spending is negatively associated, holding all variables constant. Being a western nation loses its statistical significance although its factor change on liberal policy remains positive. Under this model, for a unit increase in enforcement spending, the odds of a nation having more liberal policy increases by a factor of 1.0001, holding all other variables constant. Similarly, for a unit increase in educational spending as a percentage of GDP, the odds of a nation having a more liberal policy increase by a factor of 2.7344, holding all other variables constant. This model also predicts a .0682 factor decrease in the odds of a nation having more liberal policy for a unit increase in military spending, holding all other variables constant. This model concurs strongly with the direction of the relationships presented in the OLS models.
Big Sample 4-Point Ordered Logistical Model
The large sample model presents a slightly different view, though its results contain more confidence because of its larger sample size. Education spending remains positively related to liberal policy and predicts a 1.2159 factor increase in the odds of a nation having a more liberal policy for a unit increase, holding all other variables constant. Being a western nation is highly significant and predicts a 7.0315 factor increase in the odds of a nation having a more liberal policy if the condition is present, holding all other variables constant. Being a major narcotics producing or trafficking nation remains significant at the .05 level and predicts a 4.2484 factor increase in the odds of a nation having more liberal narco-policies.
Small Sample 10-Point Ordered Logistical Model
This small sample model, that includes the enforcement grant variable, is slightly less stable because logit models generally require a larger sample size than OLS models. However, its results continue to agree with the previous models. Enforcement Spending is significant at the .05 level while education spending and being a western nation are significant at the .10 level. Enforcement spending has a relatively small factor increase in the predicted odds of a nation having more liberal policy in this model, while educational spending’s factor increase in the odds for a nation having more liberal policy remains consistent with the previous models. Also consistent with the previous models: western nations have a positive association with more liberal policy.
Big Sample 10-Point Ordered Logistical Model
This larger-sample model excludes enforcement grants from estimation, but provides more information about some other potentially important variables and their relationship to narcotics policy around the world. Health spending as a percentage of GDP, military spending, a nation being involved in a major military conflict, being a western nation, and GDP per capita are all highly significant at the .01 level. For a unit increase in Health Spending, the odds of a nation having a more liberal policy increase by a factor of 1.3717, holding all other variables constant. Being a western nation increases the odds of a nation having a more liberal policy by a factor of 5.4277, holding all other variables constant. A nation being involved in a military conflict and increases in military spending reduce the odds that a nation will have more liberal drug policies, holding all other variables constant. Once again, the overlapping variable results are highly consistent with the previous models in terms of the direction of the relationships among the variables.
A review of the regression results from the statistical models constructed for this paper shows several significant variables that move closely with narco-policy outcomes. Within the four models that attempt to measure the effect of enforcement spending on policy, all of them agree that the variable is statistically significant and positively associated with more liberalized narcotics policies around the world. That is, as US enforcement grant dollars to a nation increase, these models predict that there will be an increase in the presence of liberal drug laws, holding all other variables constant. These results are contrary to pre-experimental expectations that enforcement dollars from the United States would escalate enforcement efforts in a recipient nation and encourage the continuation of status quo prohibitory policies. These results indicate that enforcement grant dollars allocated by the State Department for foreign law enforcement and counter-narcotics activities are not associated with increasing the strictness of domestic drug laws; the models indicate that there is a marginal liberal association with policy from these dollars. The author believes that this may be indicative of a failure of these dollars to incentivize policy change or, alternatively, that the use of these dollars has focused on narcotics demand reduction in a way that has incentivized a movement toward more liberal domestic policies like decriminalization by foreign governments. The true reason behind this outcome is uncertain, but would likely be described more clearly if nation-specific data on the Defense Department’s foreign law enforcement outlays could be taken into account (see Methodology pp.8). It is also possible that the goals of the State Department under the heading of “Counter-Narcotics” may be incentivizing a different sort of behavior than that intended by the Department of Defense, especially since their organizations deal with different aspects of foreign policy. Without a doubt, more research is required to better quantify this variable’s effect on international narcotics policy.
Military spending as a percentage of GDP remains statistically significant in 4 of the 6 models (see Appendix Section 1). When significant, the model predicts a negative association with liberal drug policies, indicating that as a nation increases its military spending there is a predicted decrease in the amount and degree of liberalized drug policies, holding all other variables constant. The author believes that this effect may be explained by the tendency of heavily militarized nations to favor stricter policies that generally limit personal freedoms in favor of maintaining order and/or sovereignty. The attitudes, politics, and cultures within a nation that help decide how much the government spends on the military may also work fundamentally against the implementation of more liberal drug-control policies that are more favorable to societies that value individual autonomy. More research into this phenomenon is required, but these models concur on a decidedly negative relationship between military spending and liberal drug policies.
Educational spending as a percentage of GDP remains statistically significant in 5 of the 6 models (See Appendix Section 1). All of these models agreed that educational spending is positively associated with more liberal drug policies: As educational spending increases there is a predicted increase in the degree of liberalization of narcotics control policies, holding all other variables constant. The author speculates that because educational funding reflects a nation’s basic valuation of education for its populace and that a more educated populace is more likely to be tolerant and open-minded to behaviors like drug use, nations that choose to devote greater resources to education will also be more likely to implement less harsh and tolerant alternative policies regarding illicit narcotics. Though additional research is needed, these models predict a consistent positive relationship between education spending and liberal narco-policy around the world.
Being a western nation is statistically significant in 5 out of the 6 models (See Appendix Section 1). Once again, all of the models agree on the direction of the relationship: if a nation is a western nation there is a relatively strong association with more liberal drug policies, holding all other variables constant. Because there is a strong continuing tradition of liberal democratic rule in western societies, it seems to follow that these nations would be more likely to experiment with alternative narcotics policies, especially if there exists a politically salient plurality within their population that harbors a more tolerant view toward narcotics use. These pluralities would be afforded political capital and clout, at least partially, if they were also members of a western liberal democracy. These results were consistent with pre-experimentation expectations.
A nation being a major narcotics producer and trafficker was statistically significant within the large sample ologit model that examined the 4-Point Policy dependent variable (See Results pp.14). Its predicted association with liberal policy was positive indicating that major drug producing and trafficking states are more likely to have implemented more liberal and less harsh domestic drug-control policies, holding all other variables constant within this model. The fact that the variable is only significant in a single model reduces the confidence of these predictions and more research incorporating this variable is required to determine more accurately if it is a significant factor. The author can only offer speculation that these types of nations are exceptionally affected by drug-war violence and may be taking steps to mitigate damages by pursuing alternative policies.
Health Spending as a percentage of GDP, a nation being involved in a major military conflict, and GDP per capita are all statistically significant in the large sample ologit model that examines the 10-Point Policy dependent variable only. GDP per capita and Health Spending both have predicted factor changes on liberal policy that are positive, holding all other variables constant. The conflict dummy variable is negatively associated with liberal policy. These results are once again considered tenuous at best because they remain statistically significant in only a single model. However, at a glance it would appear that a greater valuation of population health by a nation’s government and generally having a more affluent populace both are associated with more liberal policies. It is possible that the effect of these variables is already being captured by the attitudes that decide the educational spending allocation within a nation. More research is required to better isolate the effects of these particular variables on narcotics policy, if they are truly significant.
The aims of this research undertaking were to help to explain and better quantify the effect that enforcement spending has on the drug laws adopted by different world governments. Though cross-sectional statistical modeling can only offer a limited scope of how these different phenomena are related to one another, especially when concerned with narcotics policy and spending data that can be notoriously challenging to study methodologically, these models offer some perspective on the closeness between enforcement spending and law. Fundamentally, the cost of enforcing laws and maintaining national infrastructures to that end can partially show how a nation’s government values those respective laws. In the case of drug control policy, which has remained largely stagnant over the past fifty years, it appears that world governments are beginning to re-evaluate, at least partially, whether or not continuing wide-spread narcotics criminalization is worth it. If these models hold even partial validity to the true model of narcotics policy outcomes around the world, it seems clear that the United States’ attempt to continue the war on illicit narcotics by homogenizing the domestic drug policies of all other nations has failed. If this perceived trend continues, it could also be the case that the United States’ ability to influence lawmaking and enforcement regarding narcotics in other nations is not merely negligible, but potentially accelerating liberal policy shifts away from the US’s standard prohibition model.
It is suspected that a time-series analysis of policy shifts over time that incorporates this type of cross-sectional data on spending and enforcement grant allocation can potentially model the long term variation in drug policy choice by nations more substantively. Future research should also attempt to make use of the Department of Defense’s nation-specific data on enforcement grants, as they account for an enormous proportion of foreign law enforcement funding to other nations and remains unavailable at the time this paper was written because of inadequate record keeping and accounting practices within the department’s massive bureaucracy. In the interest of a healthier and more informed society, it is hoped that this type of research will continue to be conducted by those in academia and discussed in an unbiased way within the media and educational system so that future citizens will have more reliable and evidence-based policy information on the costs of criminalizing the consumption of psychotropic substances.
Bibliography Omitted — Available Upon Request