Alcohol and Marijuana: Substitutes or Complements (8 to 12 Minute Read)

Update: This post was originally written in 2013. After 5 more years of evidence, its conclusion of substitution between alcohol and marijuana and the public health benefits of marijuana legalization have even more evidence and relevance.

Over at Reason.com, there's an article up about whether pot and alcohol are substitutes or complements. As author Jacob Sullum notes, this could be very important for policy makers to determine whether or not marijuana should be legalized. Since many view weed as less harmful than alcohol, whether legalizing will lead to net public health benefits hinges on this economic question.

So, what are "substitutes" and "complements," how do we classify weed and alcohol, and why does this affect what policy should be pursued?

Substitutes and Complements
Substitutes and complements are terms that refer to how people purchase/ consume two different products or services. Do you buy them together (complements) or are they used in place of each other (substitutes)? Technically, economists are interested in whether price changes vs amount purchased are positively or inversely correlated.

For example, if the price of oranges goes up, you will probably buy more apples because they are somewhat similar. In manufacturing, aluminum is often a substitute for steel. On the other hand, complements are products that go well together. Examples of these include peanut butter and jelly, television sets and DVD players (increasingly becoming an obsolete example), or popcorn and a movie. Interestingly enough, if the government banned Netflix and other streaming platforms (quit hyperventilating it's only hypothetical), economists would expect the price of popcorn to decline.

There are at least a few pieces of evidence that should lead us, perhaps contrary to popular opinion, to conclude marijuana and alcohol are substitutes. While not always true (think cross-fading), from a public policy perspective, the evidence is abundant.

Economics of Alcohol and Marijuana Purchasing
Two economists, D. Anderson and Daniel Rees, examine whether marijuana and alcohol are complements or substitutes in The Legalization of Recreational Marijuana.  Previous research attempted to estimate the relationship between Mary Jane and alcohol by using price changes or comparing nearby city or state policies (using cross-sectional studies), but as they point out this can often lead to spurious results since many unseen, hidden, correlated, or random factors can significantly affect the conclusions. For example, studies that compared Utah vs. Colorado do not consider the cultural differences between the people of these states. (AKA those Mormon's don't drink coffee, smoke weed or tobacco, consume alcohol, or otherwise sin in any fashion.)

The Gold Standard of Microeconomics
This is why economists generally prefer to look at natural experiments. Ideally, we could run a truly  randomized experiment like what is done in controlled medical studies.  Wouldn't it be nice if we could settle the perennial question of minimum wage's effect on employment by randomly assigning each US city with a different wage rate? A randomized experiment is so powerful because there is a control group and a experimental group who's variables of interest vary as defined by researchers. The researchers can then retroactively go back and look at how their tweaks resulted in differences between the control group and experimental groups, and if the groups studied are quite large (or large enough), we can be very confident in the statistical validity of those differences.

Here's an interesting blog post about the change in economics from a "theoretical" disciple to a more scientific one. Credit Noah Smith

While there has been a push to attempt truly randomized control trials (RCTs) in economics, and particularly in developmental economics, this is usually quite tricky. Instead, it is often easier to look at "natural" experiments. In this case, a researcher isn't controlling variables, but is looking at observational (historical) data and trying to define or find a natural change that looks as similar as possible to a randomly controlled experiment.

In general, policy changes are often great natural experiments since many other factors remain constant, and we can simply look at before and after the policy took place. Additionally, when the policy at hand is decided at a more local level, we can compare changes across a larger group. If a federal law is passed, we don't necessarily have a control group. But if a few states change laws while a few others don't change, we can tease out some of the hidden/ random variables by comparing the resulting differences and feeling assured that the policy is in fact the cause. In the case at hand, economists have looked at changes in the minimum legal drinking age and changes in medical marijuana laws.

Changes in Minimum Legal Drinking Age
As Anderson and Rees note, there is a decrease in marijuana consumption starting at age 21, which would appear to be evidence that marijuana and alcohol are substitutes. However, it is possible this is due to other random factors; that is, until we look at places where there used to be the more sane drinking age of 18. The data suggests in places where the drinking age was 18 people would start smoking less weed at 18; then, when those states raised the drinking age to 21, people wouldn't start smoking less weed until 21. That is quite strong evidence that marijuana and alcohol are substitutes, as well as elucidating the effectiveness of natural experiments.

Changes in Medical Marijuana Laws
Since lots of states changed their medical marijuana laws at different times, this is a great natural experiment based on state by state differences and time differences. This allows us to see drops in heavy drinking among younger age groups of 18-29, as well as drops in beer sales by as much as 5%.

Additional Evidence of Substitution
Stock Prices are Inversely Correlated
Additionally, investors understand that legalizing marijuana isn't going to lead to more consumption of alcohol and opioids but less. In other words, marijuana stocks are competing for investor's money away from these industries. As such, we would expect the stock prices of these industries to be inversely related. This is what we see in the data.

Tilray is a Canadian marijuana company that became the first reefer madness company to be listed on the Nasdaq in June 2018. Its been one of the highest performing IPOs in recent history, and also one of the most overvalued. Compare that to good 'ole Budweiser stock and you can see the decline. While the graph change seems subtle, Anhueser Busch is an extremely mature stock, with a market cap of ~$150 Billion, meaning that subtle 16% decline over 3 months corresponds to a drop of about $20 Billion in the value of BUD in anticipation of Canadian marijuana legalization October 17th. 



Surveys Indicate Declines in Alcohol Usage
The graph below is of middle and high school students drinking habits before and after marijuana was legalized in Colorado. The drop is significant, but perhaps contradicted by the fact that marijuana usage is also down among this younger age group. I leave it as a puzzle for the reader to decide why this anomaly of decreases in both alcohol and marijuana post legalization exists. (Hint, did you ever feel rebellious in high school?)

Source: https://www.colorado.gov/pacific/sites/default/files/PF_Youth_HKCS_Substance-Infographic.pdf

More direct evidence from a more relevant age group in Colorado is that "the number of 18-to-25-year-olds using alcohol on a monthly basis fell by four percentage points between 2014-2015 and 2015-2016." It's pretty obvious that this age group is definitely substituting alcohol with more of something. Try walking through Denver for more than 10 minutes without smelling weed. 

Public Policy
All of this makes sense, even if initially counter-intuitive. While it may seem like cross-fading is extremely popular, in the aggregate this is a minority of consumption. People who choose to drink alcohol or smoke marijuana are usually trying to either relieve stress, have fun, loose control, and/or relax. Since both alcohol and marijuana achieve these same goals, people tend to use them as replacements. While obviously some people use them together to enhance some of these effects, this is a minority rather than a majority, as is indicated in the above facts. Cross-fading normally happens while partying or binging; most days you (who said anything about me) pick one or the other.

Traffic Fatalities
Why does this matter? As Steve Levitt finds, "legally drunk drivers pose a risk 13 times greater than sober drivers" for getting into an accident. On the other hand while studies are much sparser, people who drive soon after consuming marijuana are only 2 times more likely to get into an accident compared to a sober driver. While still impaired, these drivers often drive slower and take less risks. This means that if people use marijuana in place of alcohol then we would most likely expect lower rates of car accidents. This is what Anderson and Rees find; states where medical marijuana is legalized see on average a 13% drop in traffic fatalities. This is certainly unexpected, and most people would prefer to not even consider these facts in such a cold-hearted economical manner, but it's nonetheless true. The youngest age groups (18 to 29) are most likely to be involved in accidents, and they are also most likely age group to shift their consumption towards marijuana in states where it becomes legal.

Addiction 
Also, looking at how addictive marijuana is compared to alcohol and other drugs, marijuana generally comes out at or near the bottom of the list:

Source: https://www.drugwarfacts.org/cms/Addictive_Properties#sthash.vUS1SpC8.dpbs

This means that if people substitute alcohol with marijuana we can expect there to be drops in addiction levels/ alcoholism. If they were complements, then we would expect an increase in these issues.

Opioid Usage
Thinking of the opioid crisis is depressing no matter what. But again, Colorado's legalization of marijuana appears to have lowered what would have been a dramatic increase compared to the rest of the US. This is why big pharma is leading the push against marijuana legalization.



Violence
Finally, people who are under the influence of alcohol are usually more violent than people who are under the influence of marijuana, unless you believe Reefer Madness. This means that there should be lower levels of violence as marijuana becomes more available. While the evidence is less clear in cases of domestic violence, cartel related crime has fallen massively in states bordering Mexico after legalization efforts. Researchers "found that among the border states the effect of the change in law was largest in California, where there was a reduction of 15% in violent crime, and weakest in Arizona, where there was a fall of 7%. The crimes most strongly affected were robbery, which fell by 19%, and murder, which dropped by 10%. Homicides specifically related to the drug trade fell by an astonishing 41%."

As any economist would tell you, there are forces working in the opposite direction. You might think that as marijuana became more available people would become more lazy, and there may even be higher rates of psychosis. Obviously, many people are concerned about marijuana being a gateway drug. One thing is certain; more people would have the munchies. But come on, are these issues really worth the aforementioned public health benefits?

The conclusion to draw from this in evaluating public policy is that it is crucial to ask "compared to what?" Someone might argue that legalizing marijuana will lead to more of the bad things associated with smoking marijuana. However, it also lead to less of the much worse things associated with drinking alcohol, such as violence, traffic fatalities, addictions and overdoses. It seems abundantly discernible which is the better outcome.