# RECORD 28fd8140cfc928298ce0dcc59bc4a1264ace9c6c
# TITLE: Exploring the quantitative impact of medical marijuana dispensaries on residential sale prices in Oklahoma
# YEAR: 2026.0  VENUE: Empirical Economics  DOI: 10.1007/s00181-026-02894-6
# AUTHORS: Joshua Clark; Michael S. Delgado
# PAGES: 25  CANDIDATE TABLE PAGES (1-based): [10, 11, 16, 17, 19, 20, 22]
# ABSTRACT: 
 Medical marijuana was legalized in Oklahoma as recently as 2018, and since then Oklahoma has rapidly grown to have the largest number of medical marijuana dispensaries of any states in the USA. What have been the impacts of this rapid proliferation of legalized medical marijuana? We use a hedonic pricing model to assess how legalized marijuana dispensary activity (both distance to the nearest dispensary and the number of nearby dispensaries) has impacted residential property values. The dataset spans six different counties from across the State of Oklahoma, and we explore a variety of different regression specifications to rigorously explore the effects of dispensary activity on local residents. While we find evidence of some heterogeneity in the dispensary effects, one broad finding is that residents prefer not to live in immediate proximity to a dispensary but prefer access to dispensaries moderately distanced from their home. Understanding these broad trends and the localized heterogeneity in effects are important for residents and policymakers alike in Oklahoma and in other states that may be considering similar legislation.


===== PAGE 1 (score 3, numeric-paren 0, table-words 0, var-words 3) =====

Empirical Economics (2026) 70:61
https://doi.org/10.1007/s00181-026-02894-6
Exploring the quantitative impact of medical marijuana
dispensaries on residential sale prices in Oklahoma
Joshua Clark1 · Michael S. Delgado1
Received: 20 December 2024 / Accepted: 28 January 2026 / Published online: 21 March 2026
© The Author(s) 2026
Abstract
Medical marijuana was legalized in Oklahoma as recently as 2018, and since then
Oklahoma has rapidly grown to have the largest number of medical marijuana dispen-
saries of any states in the USA. What have been the impacts of this rapid proliferation
of legalized medical marijuana? We use a hedonic pricing model to assess how legal-
ized marijuana dispensary activity (both distance to the nearest dispensary and the
number of nearby dispensaries) has impacted residential property values. The dataset
spans six different counties from across the State of Oklahoma, and we explore a
variety of different regression specifications to rigorously explore the effects of dis-
pensary activity on local residents. While we find evidence of some heterogeneity in
the dispensary effects, one broad finding is that residents prefer not to live in immedi-
ate proximity to a dispensary but prefer access to dispensaries moderately distanced
from their home. Understanding these broad trends and the localized heterogeneity in
effects are important for residents and policymakers alike in Oklahoma and in other
states that may be considering similar legislation.
Keywords Medical marijuana · Hedonic · Public policy
1 Introduction
Oklahoma State Question 788 (Initiative Petition No. 412), or SQ 788, was filed with
the Oklahoma Secretary of State on April 11, 2016, and proposed the legalization of the
purchase and consumption of medical marijuana for adults with a state issued medical
marijuana license. SQ 788 established that a regulatory office—the Oklahoma Medical
Marijuana Authority (OMMA)—be opened under the Oklahoma State Department of
Health to regulate and receive applications for dispensaries, consumption, growing,
and packaging licenses. SQ 788 passed on June 26, 2018, with a 56.86 percent major-
B Michael S. Delgado
delgado2@purdue.edu
Joshua Clark
joshclarkstudent1@gmail.com
1
Department of Agricultural Economics, Purdue University, West Lafayette, IN 47907-2056, USA
123


===== PAGE 10 (score 60, numeric-paren 41, table-words 1, var-words 8) =====

61
Page 10 of 25
J. Clark, M. S. Delgado
Table 1 Descriptive statistics for the pooled sample of four counties and separately by county
Pooled
Comanche
Wagoner
Washington
Oklahoma
Tulsa
Payne
Sample
County
County
County
County
County
County
Statistic
(n = 28, 429)
(n = 5, 941)
(n = 396)
(n = 286)
(n = 21, 806)
(n = 27, 374)
(n = 676)
Property price
Sale price
253,060.30
209,286.10
232,053.00
182,763.20
266,290.00
302,299.50
325,171.10
(243,574.30)
(284,027.50)
(150,395.20)
(106,298.60)
(232,476.20)
(450,721.40)
(466,785.50)
Property characteristics
Beds
3.196
3.079
3.354
3.395
3.223
3.083
(0.666)
(0.599)
(0.629)
(0.755)
(0.679)
(0.968)
Baths
2.215
1.775
2.182
1.944
2.339
1.820
2.052
(0.898)
(0.578)
(0.702)
(0.742)
(0.936)
(0.764)
(0.874)
Age
40.296
43.301
37.346
60.699
39.264
50.042
(23.885)
(24.852)
(13.651)
(19.337)
(23.625)
(24.564)
SQFT
1,904.987
1,643.816
2,107.545
2,020.227
1,970.952
(895.220)
(599.074)
(854.591)
(968.378)
(947.487)
Stories
1.154
1.149
1.181
(0.260)
(0.287)
(0.321)
Garage area
388.317
521.715
403.213
(245.387)
(161.457)
(246.642)
Garage Dummy
0.787
0.997
0.825
(0.410)
(0.050)
(0.380)
Non-Ag acres
0.557
(0.700)
Depreciation loss
−51,702.83
(33,454.04)
Dist to CBD (km)
39.787
124.586
33.039
65.149
16.473
14.524
77.832
(44.451)
(8.678)
(15.225)
(3.227)
(6.255)
(6.886)
(6.199)
123


===== PAGE 11 (score 80, numeric-paren 62, table-words 1, var-words 3) =====

Exploring the quantitative impact of medical marijuana…
Page 11 of 25
61
Table 1 continued
Pooled
Comanche
Wagoner
Washington
Oklahoma
Tulsa
Payne
Sample
County
County
County
County
County
County
Statistic
(n = 28, 429)
(n = 5, 941)
(n = 396)
(n = 286)
(n = 21, 806)
(n = 27, 374)
(n = 676)
Neighborhood
203,309.40
139,660.60
177,009.70
139,379.10
221,966.40
241,517.80
173,123.90
(155,612.00)
(58,822.90)
(101,887.20)
(60,546.30)
(169,922.10)
(250,292.60)
(135,425.30)
Density
24.801
25.176
3.407
3.703
25.364
13.497
36.544
(18.367)
(15.728)
(4.823)
(2.259)
(18.906)
(8.618)
(42.698)
Dispensary exposure
minDist
1,409.414
1,726.025
2,624.329
1,260.135
1,303.049
1,224.962
3,209.315
(1385.480)
(1866.693)
(1893.913)
(983.981)
(1187.434)
(1064.254)
(2953.432)
Count500
0.340
0.309
0.131
0.245
0.353
0.319
0.109
(0.894)
(0.923)
(0.567)
(0.612)
(0.893)
(0.776)
(0.525)
Count1000
1.527
1.197
0.444
1.171
1.642
1.451
0.361
(2.387)
(2.282)
(1.857)
(2.101)
(2.414)
(2.103)
(0.941)
Count2000
5.657
4.152
1.472
3.413
6.172
5.564
1.370
(6.614)
(6.106)
(3.714)
(3.208)
(6.725)
(8.618)
(3.155)
Year indicators
Year2020
0.248
0.131
0.169
0.154
0.283
0.223
0.392
(0.432)
(0.338)
(0.375)
(0.361)
(0.450)
(0.417)
(0.489)
Year2021
0.468
0.323
0.412
0.350
0.509
0.444
0.229
(0.499)
(0.468)
(0.493)
(0.478)
(0.500)
(0.497)
(0.421)
Year2022
0.236
0.325
0.290
0.409
0.208
0.333
0.272
(0.424)
(0.468)
(0.455)
(0.493)
(0.406)
(0.471)
(0.445)
Year2023
0.049
0.220
0.129
0.087
0.000
0.107
(0.215)
(0.414)
(0.335)
(0.283)
(0.000)
(0.309)
Numbers reported are sample means, with sample standard deviations in parentheses below
The variable “neighborhood” refers to our historical neighborhood value index calculated prior to the passage of SQ788
123


===== PAGE 12 (score 24, numeric-paren 2, table-words 9, var-words 4) =====

61
Page 12 of 25
J. Clark, M. S. Delgado
Fig. 1 Scatterplot of historical neighborhood values and current property prices
Table 2 Univariate correlation between neighborhood values and ln(Price/100)
Dependent variable: ln(Price/100)
ln(Neighborhood)
0.708∗∗∗
(0.006)
Constant
−0.936∗∗∗
(0.072)
Observations
28,022
R2
0.333
Adjusted R2
0.333
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
correlation—thus, a descriptive statistic that motivates our use of neighborhood as a
control variable—and should not be taken as evidence that dispensaries are located in
areas characterized in any particular way (e.g., by race, income, or other status).
Turning to the dispensary-related variables, we see that, on average, the minimum
distance to the nearest dispensary is about 1.4 kilometers (Table 1). These properties
have, on average, 0.34 dispensaries within a 500m radius of the property; 1.5 dis-
pensaries within 1km of the property; and 5.7 dispensaries within 2 kilometers of the
property. It is worth noting that Washington, Oklahoma and Tulsa counties have prop-
erties that are in closest proximity, on average, to dispensaries and with the highest
average density of proximate dispensaries.
123


===== PAGE 16 (score 105, numeric-paren 82, table-words 3, var-words 5) =====

61
Page 16 of 25
J. Clark, M. S. Delgado
Table 4 Benchmark model and sub-models to further assess minDist
Dependent variable: ln(Price/100)
No counts &
No counts &
Benchmark
No counts
minDist2
Interaction
minDist
2.197e-05***
1.467e-05***
1.135e-06
−1.613e-05**
(2.881e-06)
(2.598e-06)
(7.023e-06)
(7.761e-06)
minDist2
3.041e-09**
2.437e-09*
(1.466e-09)
(1.470e-09)
minDist ×
8.908e-11***
Neighborhood
(1.707e-11)
Count500
0.003
(0.003)
Count1000
−0.004**
(0.002)
Count2000
0.004***
(0.001)
Age
−0.012***
−0.012***
−0.012***
−0.012***
(0.0004)
(0.0004)
(0.0004)
(0.0004)
Age2
8.527e-05***
8.834e-05***
8.786e-05***
8.726e-05***
(4.010e-06)
(3.975e-06)
(3.981e-06)
(3.981e-06)
SQFT
0.001***
0.001***
0.001***
0.001***
(0.00001)
(0.00001)
(0.00001)
(0.00001)
SQFT2
−3.323e-08***
−3.328e-08***
−3.318e-08***
−3.355e-08***
(1.372e-09)
(1.373e-09)
(1.374e-09)
(1.375e-09)
Beds
−0.043***
−0.043***
−0.043***
−0.043***
(0.005)
(0.005)
(0.005)
(0.005)
Baths
0.069***
0.069***
0.069***
0.068***
(0.005)
(0.005)
(0.005)
(0.005)
Density
0.0003**
0.0003*
0.0003**
0.0003**
(0.0001)
(0.0001)
(0.0001)
(0.0001)
Dist to CBD (km)
−0.0007*
−0.001***
−0.001***
−0.001***
(0.0004)
(0.0004)
(0.0004)
(0.0004)
Neighborhood
3.823e-07***
3.736e-07***
3.3739e-07***
2.815e-07***
(1.721e-08)
(1.713e-08)
(1.713e-08)
(2.463e-08)
Year2020
−0.169***
−0.169***
−0.169***
−0.170***
(0.013)
(0.013)
(0.013)
(0.013)
Year2021
−0.075***
−0.076***
−0.075***
−0.076***
(0.012)
(0.012)
(0.012)
(0.012)
Year2022
0.087***
0.087***
0.087***
0.086***
(0.012)
(0.012)
(0.012)
(0.012)
Oklahoma Co
0.086**
0.031
0.046
0.022
(0.043)
(0.042)
(0.043)
(0.043)
123


===== PAGE 17 (score 53, numeric-paren 20, table-words 10, var-words 10) =====

Exploring the quantitative impact of medical marijuana…
Page 17 of 25
61
Table 4 continued
Dependent variable: ln(Price/100)
No counts &
No counts &
Benchmark
No counts
minDist2
Interaction
Washington Co
−0.080**
−0.127***
−0.116***
−0.126***
(0.033)
(0.032)
(0.033)
(0.033)
Wagoner Co
0.012
−0.041
−0.029
−0.040
(0.043)
(0.043)
(0.042)
(0.042)
Constant
6.814***
6.903***
6.894***
6.941***
(0.053)
(0.051)
(0.052)
(0.052)
Observations
28,022
28,022
28,022
28,022
R2
0.640
0.639
0.639
0.640
Adjusted R2
0.640
0.639
0.639
0.640
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
the average home price, this is a decrease of about 1085.63 dollars. At the same time,
the coefficient on Count2000 indicates that adding one more dispensary in the 1000–
2000m range, holding minDist, the other Count variables, and controls constant,
results in a 0.427 percent increase in property values (or about 1080.57 dollars at the
average). Taken at face value, these estimates indicate that homeowners prefer to have
a marginal dispensary located farther away from the home.
It is worth discussing the effects and significance of the control variables on prop-
erty prices before digging deeper into the dispensary exposure variables and their
relationship to property prices. Still looking at the benchmark model in Column 1 of
Table 4, we see statistically significant quadratic relationships between the age and
square footage of the property and property values. In particular, property age has an
inverse-U-shaped relationship with property prices, with a minimum property value
occurring at a property age of 70 years. The average age of the home in the pooled
sample is 40 years (Table 1). Square footage has a U-shaped relationship with property
values, with the maximum occurring at 15,046 square feet. From Table 1, the average
square footage in the pooled benchmark sample is 1904.99, and the detailed descrip-
tive statistics tables in the appendix indicate that this turning point is greater than the
maximum square footage in our data; in effect, the square footage effect on property
prices is monotonically increasing but at a decreasing rate. We see that the number of
bedrooms is significantly negative; given that we control for square footage, the neg-
ative coefficient on bedrooms means that home buyers prefer fewer but larger rooms
for a home of a given size. The number of bathrooms and neighborhood home density
are significantly positive, while distance to the nearest CBD is significantly negative.
The historical neighborhood value is significantly positive, which is in line with the
correlations shown in Fig.1 and Table 2, and indicates that historical average neigh-
borhood values are positively and significantly correlated with current sales prices.
This estimate, therefore, reinforces our intuition to include this variable as an impor-
tant control for neighborhood quality that in part explains current property prices but
123


===== PAGE 18 (score 23, numeric-paren 0, table-words 10, var-words 3) =====

61
Page 18 of 25
J. Clark, M. S. Delgado
might also correlate with dispensary activity (distance and exposure). Finally, we see
some significance of the year of sale indicators relative to the baseline year of 2023,
and significance of the county indicators relative to the Comanche County baseline.
We note that the R2 is 0.640, reflecting a reasonable fit of the regression.
4.2 Distance to the nearest dispensary
Moving beyond the benchmark regression, in Columns 2–4 of Table 4 we explore
more deeply the relationship between minDist and property values. In Column 2, we
remove the Count variables and explore the linear effect of minDist on the natural
log of property values: We find the results remain qualitatively unchanged from the
benchmark regression (Column 1), with a significantly positive effect of minDist on
property values. This estimate is fairly smaller in magnitude compared to the estimate
in Column 1, this time implying an effect of 3.71 dollars per meter, or about 371
dollars for a 100m distance.
In Column 3, we consider a quadratic specification and find that only the quadratic
term is statistically significant, and in Column 4 we consider both the quadratic
minDist specificationincludinganinteractionwiththehistorical neighborhoodindex.
In this latter model, we find that all coefficients are significant. Exploring these esti-
mates further, using both the average property value and average neighborhood value
index (203,309.40 from Table 1), we find that on average the minDist marginal effect
on property prices is about 224 dollars per 100m. In looking across the models in
Table 4, we see that sharpening our estimates (i.e., Column 4) results in a smaller
estimated average effect of minDist on property prices. Yet, from the perspective of
these models, dispensaries are viewed by homeowners as a negative amenity.
4.3 Intensity of dispensary exposure
Next, we look more deeply into the relationship between the Count variables and
property values; regression results are found in Table 5. The first model is a version
of the benchmark regression from Table 4 except that we omit minDist. Models 2–4
consider the three Count variables separately, one at a time. The dependent variable
remains ln(Price/100).
The results in the first column are very similar to the results shown in the bench-
mark regression (Column 1 of Table 4). That is, removing minDist does not change
the results from the benchmark model. This result mirrors Column 2 of Table 4 in
that removing the Count variables did not change the effect of minDist on property
price. That is, we continue to find an insignificant coefficient on Count500, a sig-
nificantly negative coefficient on Count1000, and a significantly positive coefficient
on Count2000 all with magnitudes very similar to those discussed in the previous
subsection.
It is interesting that when considering the Count variables separately, as shown in
Models 2–4, only Count2000 remains significant (and still has a positive coefficient).
It is important to remember that in these three regressions, the interpretations of the
Count variables coefficients are different from the joint model. Since the other bands
123


===== PAGE 19 (score 81, numeric-paren 63, table-words 3, var-words 5) =====

Exploring the quantitative impact of medical marijuana…
Page 19 of 25
61
Table 5 Models to further assess significance of count variables
Dependent variable: ln(Price/100)
All
Count500
Count1000
Count2000
Count500
0.0001
0.001
(0.003)
(0.003)
Count1000
−0.005***
0.0004
(0.002)
(0.001)
Count2000
0.003***
0.002***
(0.001)
(0.0004)
Age
−0.013***
−0.012***
−0.012***
−0.013***
(0.0004)
(0.0004)
(0.0004)
(0.0004)
Age2
0.00009***
0.00009***
0.00009***
0.00009***
(0.000004)
(0.000004)
(0.000004)
(0.000004)
SQFT
0.0006***
0.0006***
0.0006***
0.0006***
(0.00001)
(0.00001)
(0.00001)
(0.00001)
SQFT2
−0.0000000***
−0.0000000***
−0.0000000***
−0.0000000***
(0.00000)
(0.00000)
(0.00000)
(0.00000)
Beds
−0.043***
−0.043***
−0.043***
−0.043***
(0.005)
(0.005)
(0.005)
(0.005)
Baths
0.070***
0.070***
0.070***
0.070***
(0.005)
(0.005)
(0.005)
(0.005)
Density
0.00006
0.00008
0.00008
0.00006
(0.0001)
(0.0001)
(0.0001)
(0.0001)
Dist to CBD (km)
−0.0002
−0.001*
−0.001*
−0.0002
123


===== PAGE 20 (score 85, numeric-paren 66, table-words 6, var-words 0) =====

61
Page 20 of 25
J. Clark, M. S. Delgado
Table 5 continued
Dependent variable: ln(Price/100)
All
Count500
Count1000
Count2000
(0.0004)
(0.0004)
(0.0004)
(0.0004)
Neighborhood
0.0000004***
0.0000004***
0.0000004***
0.0000004***
(0.0000000)
(0.0000000)
(0.0000000)
(0.0000000)
Year2020
−0.171***
−0.171***
−0.171***
−0.171***
(0.013)
(0.013)
(0.013)
(0.013)
Year2021
−0.077***
−0.077***
−0.077***
−0.078***
(0.012)
(0.012)
(0.012)
(0.012)
Year2022
0.086***
0.086***
0.086***
0.086***
(0.012)
(0.012)
(0.012)
(0.012)
Oklahoma Co
0.141***
0.094**
0.096**
0.143***
(0.043)
(0.041)
(0.042)
(0.043)
Washington Co
−0.059*
−0.094***
−0.093***
−0.059*
(0.033)
(0.032)
(0.032)
(0.033)
Wagoner Co
0.072*
0.025
0.027
0.074*
(0.042)
(0.041)
(0.041)
(0.042)
Constant
6.803***
6.860***
6.858***
6.797***
(0.053)
(0.051)
(0.052)
(0.053)
Observations
28,022
28,022
28,022
28,022
R2
0.639
0.639
0.639
0.639
Adjusted R2
0.639
0.639
0.639
0.639
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
123


===== PAGE 21 (score 14, numeric-paren 0, table-words 4, var-words 6) =====

Exploring the quantitative impact of medical marijuana…
Page 21 of 25
61
are not held constant, the coefficient on Count1000, for instance, indicates the effect
on log price of an additional dispensary located anywhere within 1000m of the prop-
erty, whereas previously it pertained only to an additional dispensary being located
within the 500–1000m distance band from the property. When taken separately, these
intensity variables indicate that a change in intensity of dispensary exposure does not
significantly affect property values within 500m (see Column 1) or up to 1000m (see
Columns 2–3) of the property. A marginal increase in intensity up to 2000m from the
property results in an increase in the property value of about 0.19 percent, or about
480.81 dollars based on the average property sale price. Broadly speaking, these results
also indicate that residents prefer a marginal dispensary to be located farther from the
property, though they may be indifferent to a marginal dispensary being located nearer
to the property.
4.4 County-specific regressions
Thus far we have focused exclusively on both distance and intensity measures of
dispensary exposure for our four-county benchmark sample. To ascertain the extent
to which there may be heterogeneity in the dispensary effects across counties, and to
exploitdataontwoothercounties,wenowdividethesampleandturntocounty-specific
regressions. Note that from the four benchmark counties, the majority of observations
come from Oklahoma County, with a large group also coming from Comanche County.
Relatively, few observations come from Wagoner and Washington counties. Thus, it
is interesting to assess the extent to which the pooled sample results merely reflect
Oklahoma and perhaps Comanche counties, or are representative of all counties more
generally. For these county-specific models, we are no longer bound by a common set
of control variables across counties; for each county-specific regression, we include
as many control variables as are available for that county. Also, for each county we
use the Fitzpatrick and Parmeter (2021) cross-validation method to select the optimal
minDist threshold, and then use all property sales within 2 times that threshold for our
county-optimal sample. We report the county-specific regression results in Table 6. To
save space, we withhold estimates of all control variables, and provide the estimates
of the dispensary effects.9
Looking first the the coefficients on minDist across Table 6, we see that minDist
is significantly negative in Comanche and Payne counties, insignificant in Wagoner
and Washington counties, and significantly positive in Oklahoma and Tulsa counties.
Count500 is only significant in Tulsa County, wherein we find the estimate is positive.
Count1000 is significantly negative in Oklahoma County and significantly positive
in Tulsa County, while Count2000 is significantly positive in Oklahoma County and
significantly negative in Payne County. Taking stock of the first four columns, relative
9 In terms of additional variables per county, we note that Comanche County has a dummy variable
indicating whether the property has a garage, and a township location indicator; Wagoner County has
the number of stories per home, the garage dummy, amount of non-ag acreage of the property, and measure
of property depreciation loss; Washington County has the number of stories and garage dummy; Oklahoma
County has the number of stories and rooms in the home; Payne County lacks data the age and square
footage of the home; and Tulsa County lacks data on square footage but has available the number of stories
in the home.
123


===== PAGE 22 (score 65, numeric-paren 45, table-words 6, var-words 0) =====

61
Page 22 of 25
J. Clark, M. S. Delgado
Table 6 County-specific benchmark regressions
Dependent variable: ln(Price/100)
Comanche
Wagoner
Washington
Oklahoma
Payne
Tulsa
minDist
−0.00013***
0.00001
−0.00001
0.00002***
−0.00005**
0.00011***
(0.00003)
(0.00001)
(0.00002)
(0.000003)
(0.00002)
(0.00001)
Count500
−0.005
0.054
−0.008
0.001
0.023
0.019***
(0.011)
(0.055)
(0.035)
(0.003)
(0.110)
(0.006)
Count1000
0.004
−0.008
0.001
−0.008***
0.012
0.011***
(0.006)
(0.027)
(0.013)
(0.002)
(0.077)
(0.003)
Count2000
−0.003
−0.003
−0.007
0.006***
−0.044***
−0.002
(0.003)
(0.011)
(0.008)
(0.001)
(0.015)
(0.001)
Constant
7.768***
6.501***
5.715***
6.815***
4.989***
7.043***
(0.465)
(0.345)
(0.458)
(0.021)
(0.464)
(0.038)
Control variables?
Yes
Yes
Yes
Yes
Yes
Yes
minDist threshold
907.87
4884.91
4884.91
2655.66
3719.51
1192.78
Observations
4,068
396
286
21,463
622
24,233
R2
0.246
0.536
0.772
0.732
0.307
0.197
Adjusted R2
0.243
0.511
0.756
0.732
0.293
0.196
∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01
123


===== PAGE 23 (score 12, numeric-paren 0, table-words 3, var-words 6) =====

Exploring the quantitative impact of medical marijuana…
Page 23 of 25
61
to the benchmark pooled sample regressions, it seems apparent that the results found
with the four pooled counties are largely driven by the Oklahoma County effects. Table
6 shows that residents in Comanche County experience a negative minDist effect,
indicating that property prices fall when dispensaries are located farther away from
the home, though the intensity of dispensary activity does not affect property values.
Looking toward Payne and Tulsa counties, we see that Payne County bears similar
effects to Comanche County, while Tulsa County accords with Oklahoma County. It
is worth pointing out the following. First, there is quite a bit of heterogeneity across
counties in terms of both the optimal minDist threshold and the R2 measures for
each regression, and readers should bear in mind that both Payne and Tulsa counties
are missing several key variables (which is why they were left out of the baseline
four-county model). In terms of the R2, Oklahoma County seems to have the best fit,
relative to the regressions for the other counties.
5 Discussion and conclusion
We study the effects of medical marijuana dispensaries on residential property prices in
Oklahoma as a means of measuring the extent to which dispensaries are seen as desir-
able or undesirable neighborhood amenities and thereby affect property prices. We
have motivated the uniqueness of Oklahoma for our study setting: not only has Okla-
homaveryrecentlylegalizedmedicalmarijuana(effectiveAugust2018)butOklahoma
has seen a rapid rise in the proliferation of medical marijuana dispensaries and has
become a leader in the nation in terms of the number of operating dispensaries. More-
over, Oklahoma is an interesting case because of the strongly conservative electorate,
despite marijuana legalization being typically seen as a politically liberal interest.
Thus, we believe that Oklahoma is an ideal setting for studying the effects of medical
marijuana legalization on nearby property values with lessons for citizens of Okla-
homa as well as the public or policymakers in other states. Using data from six counties
in Oklahoma, we conduct a state-wide hedonic pricing analysis to investigate the net
effects, as capitalized in property values, of local dispensary activities on residential
property prices. We measure dispensary activity with both distance to nearest dispen-
sary as well as intensity of dispensary activity in the proximate area (within 500m,
1000m, and 2000m from the property). We consider pooled sample regressions as well
as county-specific regressions, linear and nonlinear specifications. In looking across
these six counties, we find some heterogeneity in our results. However, our results
point to a few broad trends across most of Oklahoma. First, a general pattern is that
residents view local proximity to a medical marijuana dispensary as a net negative,
with property prices being significantly lower when either the minimum distance to
the nearest dispensary is closer to the home or when a marginal dispensary is opened
within 0.5 kilometers from the home. We find the effects to be quite small, with even
the largest effects being less than one percent of the property’s value. At the same time,
residents seem to prefer relatively easy access to medical marijuana dispensaries, with
property prices increasing for marginal dispensaries being located at a medium dis-
tance (i.e., between 1 and 2 kilometers) from the home. Second, it is worth noting that
Washington and Wagoner counties generally have insignificant effects of dispensary
123