Marijuana Price Gradients: Implications for Exports and Export-Generated Tax Revenue for California After Legalization
Jonathan P. Caulkins
Brittany M. Bond
Abstract
California nearly legalized commercial marijuana production in 2010; it and other states are expected to entertain similar proposals in coming years. This raises the question of whether marijuana diverted from legal production could displace current sources of marijuana in other states. Combining Kilmer et al.’s (2010a,b) estimates of legal production cost with new estimates of the cost of smuggling within the U.S. suggests that in most states diverted legal production would substantially undercut current prices. So if one state legalized, such (illegal) inter-state “exports” could depress marijuana prices throughout the U.S., even if taxes are collected before diversion and export. We proxy smuggling costs by the current gradient in prices observed for Mexican or commercial grade marijuana; based on seven different data sets it appears to be roughly $325 - $475 per pound per thousand miles as one moves north from the Mexican border.
Keywords: Drug policy, drug markets, marijuana, prices, legalization
Running Head: Marijuana price gradients and implications for legalization
Introduction
California entertained two serious proposals to legalize marijuana in 2010: a legislative bill (the “Ammiano Bill”) that died before receiving a full vote and a voter proposition (“Proposition 19”) that was narrowly defeated on November 2nd. Either could have made large-scale commercial production for recreational use legal with respect to state and local laws, something no jurisdiction in the world has done. Kilmer et al. (2010a, 2010b) offer a comprehensive analysis of implications for California budgets and marijuana consumption, and effects on Mexican drug trafficking organizations’ revenues and violence.
Here we investigate an under-appreciated implication of legalization: the possibility that marijuana diverted from legal production might be (illegally) exported to other jurisdictions. Such exports could place downward pressure on prices elsewhere and, if the diversion occurs after taxes were collected, potentially create export-based tax revenue for the state that legalized.
The viability of this scenario would depend on how legalization is implemented and on federal and other states’ law enforcement responds. We defer discussion of such qualifications to the end of the paper, but acknowledge that we are discussing only one plausible scenario.
Superficially it seems implausible that taxes might be collected on exports, given that those exports would remain illegal. However, some growers could produce for both the legitimate market and the inter-state black market, as happens with regulated opium production in India today (Paoli et al., 2008). If taxes are collected from the producer, rather than at retail, then taxes might be collected before the marijuana is diverted. Indeed, even (purely criminal) exporters might value having their purchases be above board and, hence, taxed, if that reduces their risks when acquiring the marijuana.
This issue remains topical even though both California measures failed for three reasons. First, the analysis sheds light on the cost of transporting illegal drugs within the U.S. Second, there will likely be future efforts to legalize marijuana in California. Third, legalization in some but not all states or nations in a region could happen elsewhere, and the approach might carry over, even though the parameters here are specific to the U.S.
The basic exercise is to compare for each U.S. state in the lower 48 (1) the projected post-legalization price in California bumped up by the cost of smuggling marijuana illegally from California with (2) the current price of marijuana. If the former is smaller, then it is plausible that California might end up supplying consumers in that other state. We ignore the possibility of exports to Alaska, Hawaii, and other countries; the estimates are conservative in that respect.
Kilmer et al. (2010a) estimated production costs after legalization and discuss potential taxes and tax evasion. As we explain below, smugglers might be able to acquire high-quality marijuana (sinsemilla) in California at prices ranging from $300 per pound (low-end production cost, untaxed) to $1350 per pound (high-end production cost, with sales and a $50 per ounce excise tax).
Although $300 - $1350 per pound is well below current sinsemilla prices, even $300 per pound is very high compared to most conventional goods; it is roughly the same price per unit weight as silver. At those value-to-weight ratios, conventional distribution costs (fuel, labor, etc.) become minor considerations. Rather, it is costly to transport marijuana primarily because of the need to compensate for various risks, particularly of enforcement (Reuter and Kleiman, 1986; Caulkins and Reuter, 2010).
Legalizing marijuana in California should not greatly affect the cost of smuggling from California to other states because most of the smuggling risk comes from enforcement by the federal government and other states. Hence, we infer future smuggling costs from currently observed spatial variation in marijuana prices. Essentially, we plot current price vs. distance from presumed source and observe the slope. This linear model is a simplification, but it fits the data reasonably well.
Drug prices are not measured perfectly, so this approach only works if the price gradient is large relative to the error in measurement of prices. The coefficient of variation of measured prices is roughly the same across a range of drugs (Reuter and Caulkins, 2004), so the price gradient is easiest to measure for a cheaper drug, namely Mexican or “commercial grade” marijuana. (Ditch weed can be cheaper still, but it has no single source location.) So we estimate future smuggling costs with current commercial grade marijuana data, even though post-legalization we expect essentially all domestic production to be sinsemilla (Kilmer et al., 2010a). Since marijuana sentences are based on weight, not marijuana type or potency, this may not be unreasonable.
We turn now to the data and price gradient. Section 3 estimates California market share post-legalization. Section 4 concludes.
Gradient in Marijuana Prices Observed within the U.S.
Sources of Price Data
The ideal data for examining the relationship between price and distance from source would be official reports on current wholesale prices for Mexican marijuana in every one of the lower 48 states. There is no ideal data set. So instead we use every available data set that is relevant. Inasmuch as the same general pattern holds in each, we can be reasonably confident of the basic finding.
In particular, we use data from seven sources that are described briefly in the Appendix and in more detail by Bond and Caulkins (2010): the Narcotic News website (http://www.narcoticnews.com/), the National Drug Intelligence Center (NDIC) Drug Market Analysis reports, the Drug Enforcement Administration’s (DEA’s) System to Retrieve Information from Drug Evidence (STRIDE) database, the DEA’s Illegal Drug Price/Purity Reports (IDPPR), the Arrestee Drug Abuse Monitoring (ADAM) system, High Times magazine, and a new website Price of Weed. Table 1 summarizes key attributes of the data sets.
The sources are not entirely independent; one would presume NDIC draws on information from DEA. However, none is simply a copy of another. For example, the Narcotic News website and NDIC Market Analysis prices are highly correlated (correlation 0.647), but for only one state do they quote exactly the same price (Utah, $600 - $1000 per pound).
Table 1: Seven Data Sets Used to Estimate the Marijuana Price Gradient Within the U.S.
Source / Market Level / Quality/Type / # of States / Time / SourceDEA IDPPR / Pound / Commercial Grade / 16 / 1999-2001 / Enforcement Agency
NDIC / Pound / Mexican / 39 / 2001-2009 / Enforcement Agency
STRIDE / Pound / $250 - $2000/lb. / 25 / 2005-2009 / Transaction Data
Narcotics News / Pound / Not sinsemilla or high grade / 48 / 2010 / Enforcement Agency Voluntary Report
ADAM / Pound / 200-1500 grams and average price < $1900 per pound / 27 / 2000-2003 / User (Arrestee) Survey
High Times / Ounce / Schwag / 45 / 1996-2005 / User Voluntary Report
Price of Weed / Ounce / "Low" / 47 / 2010 / User Voluntary Report
Distance Data
Distances were proxied by distance to the nearest city in Mexico, computed using the web site http://www.mapcrow.info/Distance_Mexico.html. Effectively this is akin to assuming that transport costs within Mexico between production areas and border cities are small. Marijuana is also imported from Canada (mostly sinsemilla) and Jamaica, and there is illegal domestic production. Nevertheless, Kilmer et al. (2010b) estimate that Mexican marijuana accounts for 40 – 67% of U.S. marijuana consumption, and so is the majority of the roughly 80% of that market which is not sinsemilla.
Proximity to Canada may affect prices, so we considered Canada as a potential source. In particular, we computed distances to the “nearest” source, in Mexico or Canada (specifically Vancouver or Quebec), allowing for an additional “fixed charge” representing how much harder it is to smuggle across the US-Canadian border than across the US-Mexican border. This did not improve explanatory power, so we report here only price vs. distance from Mexico. Likewise adding other independent variables (e.g., population for prices quoted at the city level) did not statistically improve the fit, although searching for better multivariate models may be an avenue for further research.
Correlations and Descriptive Statistics with State Level Prices
Table 2 shows that inter-state variation in price is positively correlated across all pairs of data sets, and prices in all seven are positively correlated with distance from Mexico. Where the data were not originally at the state level, we took a simple, unweighted average across city-specific prices. E.g., the Narcotic News California price of $416 per pound is a simple average of the prices for San Diego ($350), Los Angeles ($375), San Francisco ($438), and Fresno ($500). The four data sets supplied by enforcement agencies show the highest correlations with distance and with each other; the three data sets based on user reports appear to include more random variation.
Table 2: Correlation in Marijuana Prices Across Data Sets and with Distance from Mexico
Distance / DEA IDPPR / NDIC / STRIDE / Narcotics News / ADAM / High TimesDistance from Mexico / 1.0
DEA IDPPR / 0.596 / 1.0
NDIC / 0.547 / 0.736 / 1.0
STRIDE / 0.688 / 0.661 / 0.594 / 1.0
Narcotics News / 0.674 / 0.663 / 0.647 / 0.567 / 1.0
ADAM / 0.264 / 0.494 / 0.329 / 0.573 / 0.212 / 1.0
High Times / 0.318 / 0.604 / 0.276 / 0.521 / 0.288 / 0.297 / 1.0
Price of Weed / 0.285 / 0.364 / 0.286 / 0.493 / 0.400 / 0.629 / 0.423
Figure 1 plots state-level price per pound vs. distance from Mexico for STRIDE and Narcotics News. STRIDE is the most familiar in the academic literature, and Narcotic News is the only one that both covers all the states and is (at least purportedly) based on enforcement agency data.
The pattern of price increasing with distance is clear at a glance, as is the considerable dispersion around the trend lines. The two slopes ($343 and $463 per pound per thousand miles for NN and STRIDE, respectively) may bracket the true value, since the actual value along the Mexican border (at a distance of 0) is about $400 per pound (Kilmer et al., 2010b), which is roughly half way between the intercepts of these two linear relations.
Figure 1: Marijuana Price per Pound vs. Distance from Mexico in Data from STRIDE and Narcotics News
Price Gradient Results for Mexican/Commercial Grade Marijuana in the U.S.
We created similar plots for all the data series, keeping the analysis at the city level when possible. Table 3 summarizes the results. The gradients are between $325 and $475 per pound per thousand miles. We use that range in our analysis below, with $400 as the base case value.
Some data sets were explicitly about Mexican marijuana, but most were not so specific. That is problematic because price varies by type and quality of marijuana. As noted below and in the Appendix, we dropped observations whose prices were clearly inconsistent with other prices of Mexican marijuana. This is necessarily a somewhat subjective process, but we verified that inclusion or exclusion does not dramatically affect the estimated price gradients.
Table 3: Summary of Regressions of Price vs. Distance from Source
Source / Intercept, Constant / R2 / Price Gradient$/lb increase/ 1000 miles
Narcotic News / $482 / 0.518 / $392
STRIDE / $237 / 0.427 / $474
NDIC / $637 / 0.300 / $343
IDPPR
IDPPR All years (1986 – 2000) with 14 dummy variables for individual years / $516 / 0.621 / $437
IDPPR All years (1986 – 2000) with 4 dummy variables for clusters of 3 yrs / $638 / 0.593 / $438
IDPPR Average of 15 separate single year regressions, 1986-2000 / $681 / 0.494 / $439
ADAM / $545 / 0.113` / $351
ADAM, excluding 4 cities with avg price/lb. > $1900 / $421 / 0.306 / $322
High Times / NA / 0.166 / $337
Price of Weed / NA / 0.081 / $416
This issue was most pronounced for ADAM and STRIDE, the two data sets which do not even distinguish sinsemilla from commercial grade. Kilmer et al. (2010b) found that 80% of ADAM’s price observations were well below those associated with sinsemilla, but four of the 40 cities stand out for having pound level (200-1500 gram) prices that are high in absolute terms ($1950 - $2575 per pound) and far above the regression line. Not coincidentally, all four (San Jose, Portland OR, Spokane, and Seattle), were in or near Northern California, a region associated with extensive sinsemilla production. Table 3 shows that dropping those cities greatly improved the fit but only modestly affected the price gradient. The only other city whose price exceeded $1205 was Detroit ($1705 per pound). We retained that data point because it is not a gross outlier considering distance from Mexico; dropping it would further reduce the gradient (to $276 per pound per 1000 miles) without materially improving the fit.