Forget my approach, an even freakier way to measure cocaine use

Our last Freakonomics column was about the indirect approach that Roland Fryer, Paul Heaton, Kevin Murphy, and I used to try to measure crack cocaine use across places and over time in U.S. cities and states. Read all about it here.

Some researchers in Italy took a very different, very bizarre approach, as discussed in a British newspaper article reprinted below:

Where rivers run high on cocaine
By Nigel Hawkes
Analysis of waste water in Italy shows a startlingly high level of drug abuse

THE rivers of Italy are flowing with cocaine, say scientists who have adopted a new approach to measuring the extent of drug misuse. The biggest river, the Po, carries the equivalent of about 4kg (8lb 13oz) of the drug a day, with a street value of about ?20,000.
Cocaine users among the five million people who live in the Po River basin in northern Italy consume the drug and excrete its metabolic by-product, benzoylecgonine (BE). This goes from sewers into the river. So a team led by Dr Ettore Zuccato, of the Mario Negri Institute for Pharmacological Research in Milan, estimated the use of cocaine by testing the waters of the Po for BE, and for any cocaine that had passed through the body unaltered or reached the sewers in other ways.

What they found surprised them. They calculated that for every 1,000 young adults in the catchment area, about 30 must be taking a daily dose of 100 milligrams of cocaine, which greatly exceeds official national figures for cocaine use.

According to official Italian statistics, 1.1 per cent of people between the ages of 15 and 34 admit to having used cocaine “at least once in the preceding month”. Almost all cocaine use occurs in this age group.

Assuming that there are 1.4 million young adults in the Po River basin, the official statistics suggest that there would be 15,000 cocaine-use events per month. But the evidence from the water suggests that the real usage is about 40,000 doses a day, a vastly greater figure.

“The economic impact of trafficking such a large amount of cocaine would be staggering,” Dr Zuccato said. “The large amount of cocaine – at least 1,500kg – that our findings suggest is consumed per year in the River Po basin would amount, in fact, to about $150 million in street value, based on an average US street value of $100 per gram.”

To confirm their findings, the team also sampled urban waste water from Cagliari in Sardinia, Latina in central Italy, and from Cuneo and Varese in the north – all medium-sized cities. The values they obtained from the undiluted waste water were far higher than those in the Po, as would be expected. But when translated into likely local use of the drug, they produced very similar figures – which suggests that the Po region is not exceptional in its cocaine consumption. The results cannot be explained by assuming that some drug trafficker was panicked into dumping his stash down the lavatory. If so, much more pure cocaine would have been found, and much less of its human metabolite, BE. In fact, the ratio of cocaine to BE was consistent throughout all the samples.

If anything, Dr Zuccato said, the method would be expected to underestimate rather than to overestimate cocaine use, because some would be lost or absorbed in sediments. So the real consumption may be even higher.

This method has previously been used by the same team to measure the by-products of widely-used prescription drugs, and has produced results consistent with known prescribing patterns. So it seems to work.

The technique has been developed by the Italian team and is complex, as it needs to be to detect such tiny residues – of the order of billionths of a gram per litre of water.

The scientists say that the method needs to be tested further before being brought into general use, but suggest that it would be a more reliable and much cheaper way of tracking trends in drug use than by using population surveys.

“The approach tested here, which is in principle adaptable to other illicit drugs, could be refined and validated to become a general, rapid method to help estimate drug abuse at the local level,” they report in the journal Environmental Health.

“With its unique ability to monitor changing habits in real time, it could be helpful to social scientists and authorities for continuously updating the appraisal of drug abuse.”

The levels of the drug and the metabolite found in river water are so low that any effect on natural life is very unlikely. But this is not true of all chemicals. Research indicates that chemicals that mimic natural hormones are having an effect on fish in many rivers, including “feminising” many male fish. The sources of these chemicals include hormones excreted by the human body and industrial chemicals that reach the waterways.


If it works, it is truly a brilliant idea. One thing that makes me highly, highly skeptical: if I am reading the article right, they are estimating that cocaine use is almost 100 times higher based on their estimates than one would expect given official estimates. Besides the fact that being off by 100 times would be an amazingly bad estimate by the authorities, it seems to me that these sort of official drug use estimates are always inflated because the authorities have an incentive to exaggerate the extent of the drug problem.

On the other hand, an estimate of $150 million in cocaine expenditure for 1.4 million people doesn’t sound insanely high. That would be lower than my own back-of-the-envelope estimates of cocaine expenditure per person in Chicago.

Somebody is way off. Hard to know who for sure.

(Thanks to Jon Gemus, and someone else who’s email I lost, for pointing me to this article.)


The italian scientists findings suggest that 1000 people consume 3g a day of cocaine.
The italian stats suggest 11*x*y/30 g per 1000 people, where x is the number of times a user uses cocaine in a month, and y the average dose size in grams. The stats don't have those numbers, but if xy = 8 (a user consumes 8 grams a month, a budget of 800$ a month per user), the numbers match. I must admit to having no idea if this is exorbitant or not. It is certainly on the high side, but I fail to see how it is 100 times off. Maybe 2 to 5 times off. Unless I am missing something


How about cocaine manufacturing processes? A clandestine laboratory concealed in a basement may produce wastewater, no? Perhaps it is a problem of attribution.


Econblogger, did you read the actual explanation about what they measured? They weren't just measuring cocaine in the water, they were measuring the products of human digestion of it. Or does cocaine manufacturing produce this digested byproduct?


There's a whole freakonomics sub specialty: counting people who don't want to be counted. I remember from "Nickle and Dimed" someone wanted a more acurate assessment of the Hispanic population of the El Paso TX area (more accurate than Census figures). A researcher went to the town's bus drivers and asked them to estimate the number of maids they had as passengers every day. The black & white numbers in Census reports lead to under reporting in a way that black and white uniforms on humans don't. And of course, being a bus driver myself, intrigues me on this particular application of social science.


BE in human tissues is soley the product of the matabolism of cocaine, but BE in the environement isn't nessecarily soley the product of cocaine metabolism.

Closely related metabolites are formed of of the other medically therapeutic (and generally not abused) 'caines and these can sometimes breakdown to BE when exposed to the environment in certain circumstances. In fact, if memory serves, most of the other 'caine metabolites wouldn't be enviromentally stable unless they did breakdown into BE.

This study is interesting but not proof positive of anything, really.


I, too, question how 40,000 doses is 100 times 15,000 doses. But whatever. I read in another article on this subject that the official estimates are based on surveys. I think one would expect the number of admitted cokeheads to be significantly lower than the number of cokeheads.


40000 daily doses x 30 = 1.2 million doses monthly

The survey results say 15000 users "at least" once in the past month. If 5000 of those are abusers (3 times the daily does mentioned above) and the other 10000 average 1 dose a week, then we end up with 450000+40000 ~= .5 million doses monthly.

For differences between reported use of an undesirable drug and actual use, a 2:1 ratio seems to fall within a range that is believable...


A few thoughts:

1. Assuming the data are accurate (ie, that there is no other explanation for the presence of BE, that it is in fact BE and not something similar that has been altered by its immersion in Po River water, and that it has been accurately measured) the apparent level of measured use shouldn't surprise anyone familiar with substance abusers. It is a highly addictive drug, lives get ruined, etc. There may just be a small number of very, very -- even hyper- -- active cocaine users. Saying you have used cocaine more than once in the last month includes people who really have used it only twice, people who have used it once daily (hard to imagine) and people who have used it so many times they can't remember. That last group can account for a lot of BE.

2. The Po flows through areas very popular with travelers from other countries. Italy may not have a cocaine use problem among its residents, so much as a problem with people who use cocaine while using bathrooms in the Po River basin. Much like marijuana in Amsterdam.

3. I'm not sure why anyone would believe the numbers put forth by the authorities about estimates of actual use, or use those numbers as any sort of credible comparison or baseline for measurement. They have myriad conflicting incentives, and it is hard to imagine that their data collection techniques are inherently reliable, even without those biases.



As trigger said above, it seems likely that other closely related substances -- Novacaine, et cetera, most famously used by dentists -- may be breaking down into chemically similar metabolites. I see nothing in the article to suggest that the investigators have considered that possibility.

Todd seems to me that these sort of official drug use estimates are always inflated because the authorities have an incentive to exaggerate the extent of the drug problem.

Always? It seems to me that officials who want to be reelected have a pretty strong incentive to underreport drug usage under their watch.


Inflated numbers for sure.

I find it difficult to believe that $100 for a gram is the average US street price of cocaine.
From personal experience, I know If anyone is paying more than $150 for an eightball (3.3 grams), then they are getting ripped off. The drug isn't very addictive and the high is short, and one gram will only last a single person a few hours.


Since everyone else is postulating potential problems with the Po River study, I would like to add two additional problems: 1) the river is not perfectly "stirred" such that the concentration of the metabolite could vary enormously at various locations in the cross section of the river; and 2) the rate of linear flow differs at various sites in a cross-section of the river. Multiplying the concentration in a few sites times the mean flow could yield inaccurate results. I, also, do not understand why Levitt claims the experimental and government predictions of cocaine usage vary by "100 times".


whose email I lost.


okay, for all the people who can't read th article says:
the official statistics suggest that there would be 15,000 cocaine-use events per month. But the evidence from the water suggests that the real usage is about 40,000 doses a day

and for all the people who can't count:
40,000 doses per day times 30 days = 1,200,000. And 1,200,000/15,000=80. This is a simple calculation without adjusting for how Levitt could have weighted the "at least once in the preceding month", but 80 is close enough to a 100 so you can see where he's getting his #s from.

Also. the skeptics here seem to be disregarding what seems to be a fairly significant support argument for this research:
This method has previously been used by the same team to measure the by-products of widely-used prescription drugs, and has produced results consistent with known prescribing patterns. So it seems to work.



Consider that Italy has a massive illegal alien population. Drug dealing is probably a good way to sustain yourself if you can't get legitimate employment. If that is the case, the proportion of drug dealers might be higher than normal. Having said that, how do they quantify the 1.4 million young adults in the area, as the relaxed border control means that only the legally registered adults are included in the count.

The government in Italy changes frequently and doesn't address the aliens issue. If cocaine use is so underestimated it can't be a shattering social problem. The government has no incentive to tackle this problem.

Too many variables.


authorities have an incentive to exaggerate the extent of the drug problem.

I always figured the authorities had incentives to both exaggerate and under exaggerate drug use.

They could have an incentive to exaggerate drug use when asking congress for money to combat importation, manufacture, distribution, etc..

They could have incentive to under exaggerate drug use when justifying there continued existence to congress and the public. Why would we continue spending billions & billions of dollars if it were having no effect on drug use.

I don't think the controls implemented in the U.S. have effected drug use all that much. Certainly not enough to justify the money spent. If they can't keep drugs out of controlled confined restricted areas like prisons, how can they ever hope to do so in the U.S. where people live relatively free and private lives.



I bet $500 that Freakonomics will win the Quill Award for the business category.

J. Cross

Right, the chemist results aren't 100 X the survey results. The survey finds 15,000 USERS (1.1% of 1.4 M) where each user is using AT LEAST once a month. It doesn't say the numbers of total uses. The chemists find 40,000 USES per day or 1.2 M USES per month.

That said, the average user would have to be using 80 times a month in order for these numbers to match up. One other thing to consider, not all users are between 15 and 34 so the survey wouldn't catch all users.


One other thing to consider, is that nobody defined what a use represent? is it 0.1g or 1g?


Just a quick note... I think they measure both cocaine and BE levels in the water -- they claim that cocaine metabolism results in a consistent ratio of the two, and then they claim that they observe that ratio in the water they tested. So presumably that should account (again, they claim) for non-cocaine-usage-related spikes in either BE or cocaine (i.e., "dumping" events).

They also test both surface water (SW) from the river and standing waster water (WW).

But don't take my anonymous word for it -- you can read the original paper (open access) here.

Mark Kleiman also had a pretty decent post about this a while back, too.