Is there a “wine premium” in
Chilean rural land values?
William Foster, Gustavo Anríquez,
Oscar Melo, and Jorge Ortega
Some questions regarding Chilean rural
land values
• Main question: What factors explain rural land
values in Chile?
• A motivating concern: Can we anticipate the
impact on land values of possible future climate
scenarios?
• An additional question: Beyond characteristics of
a geographic area related to future incomegenerating potential (e.g., soil types, climate,
distance to markets and likely future demographic
trends), does information regarding specific
activities add to our ability to predict land values?
Specifically, to what degree might wine grape
cultivation contribute to a rural areas land values?
• The value of rural land in a geographic area (such
as a county) should reflect the present discounted
value of future income streams from farming
activities and (as yet unrealized) potential
residential use.
– We look at Chilean “comunas” – municipalities.
• Once one has accounted for all of a geographic
area’s relevant characteristics, it would be
reasonable to presume that per-hectare land
values should be predictable in a cross-section
without reference to the specific composition of
agricultural activities taking place at a given time.
If land use decisions are optimal, basic
characteristics should determine profit-maximizing
crop mix at a specific place and point in time.
• There is, however, the possibility that a particular
type of land use, such as wine grape cultivation,
might be associated both with
– significant immobile investments, such as plantations,
the value of which should be bid into vineyard values,
– and with other activities, such as wineries and tourism,
which have external effects on the value of all land in
the surrounding area.
• So land use composition could offer additional
information regarding area-wide land values
beyond that of the area’s intrinsic characteristics
of climate, market access, soil types, etc.
Example: Simple von Thunen model : land value is proportional
to yearly rent, which depends on activity selected.
Activity net income depends on value of
crop sales at center and cost of transport.
Activity A is a high-value perishable, C is
durable, bulk commodity.
Yearly rent
per hectare
Activity A
Activity B
Activity C
0
Distance from markets / market potential
Empirical approach
• Let’s look at the correlation between rural land
values at the municipal level and basic factors:
climate, market potential, and others.
• Then add farm activities in “comunas” – does
this add information or is it redundant?
• Account for geographic correlation and
possible spillover effects.
• Results are “exploratory” and needs refining.
Land value data
• Two sources of Chilean rural land value data:
– Classified ads for sales offers (Revista del Campo)
– Transactions recorded in CBRs – registries.
• We use data from 37 land registry offices out of the 118
CBRs in total covering full area of study.
• Registries were selected randomly in a stratified
sampling framework to ensure adequate geographic
coverage. The geographic unit of analysis is the
municipality.
• 15 agro-ecological zones and 5 macro zones (north to
south) combined to 28 strata.
• In each stratum, the sample of 1-3 CBR were chosen
with probability proportional to number of farms.
Land value data
• Data for 1980, 1990, 1997 and 2007 were
gathered from recorded land transactions in 37
land registries covering 9 regions and 178
municipalities from Region III to Region X.
• After processing the basic data, information is
available for land area and per-hectare values
for 32,453 individual land transactions.
• Initial analysis of the real (cost-of-livingcorrected) value per hectare of land shows that
parcel values vary significantly according size
and region.
Count of recorded rural transactions by Chilean Region and Year
Región
3
4
5
6
7
8
9
14
10
Total
1980
63
307
329
1,040
1,639
824
753
267
319
5,541
1990
143
242
485
971
1,345
1,092
1,062
310
424
6,074
1997
120
238
560
1,367
2,033
1,436
1,685
252
1,097
8,788
2007
195
201
872
1,237
2,561
2,576
2,543
423
1,442
12,050
Total
521
988
2,246
4,615
7,578
5,928
6,043
1,252
3,282
32,453
Median number of hectares in recorded transactions
Región
3
4
5
6
7
8
9
14
10
Total
1980
31.00
16.70
9.09
10.40
15.85
34.40
40.00
47.50
44.20
17.80
1990
8.50
2.50
0.76
2.00
2.62
5.00
8.74
9.91
8.40
3.81
1997
5.42
2.26
0.75
2.04
1.50
3.00
4.21
5.48
1.26
2.04
2007
1.96
0.51
0.73
1.00
1.45
2.10
4.20
3.00
1.09
1.78
Total
5.20
3.60
0.97
2.98
3.15
3.70
6.00
9.94
2.23
3.40
Our variable of interest:
Median value per hectare in UF in recorded transactions.
Región
3
4
5
6
7
8
9
14
10
Total
1980
13.25
15.92
27.77
39.32
14.03
7.57
7.86
14.30
10.30
14.86
1990
6.15
52.12
238.03
89.65
64.50
17.85
12.86
21.97
18.26
32.16
1997
20.88
62.07
703.20
194.63
125.12
42.60
47.28
41.61
142.00
90.40
2007
144.44
652.70
591.88
366.34
154.01
66.36
51.84
91.90
108.77
107.80
Total
38.73
49.05
375.38
131.14
73.68
38.41
30.27
34.06
71.12
59.17
We use the median values per hectare in UF in recorded
transactions at the municipality level – 89 comunas for 4 years
Basic conceptual model of land values
The final assumption is that the observed value of parcel j , Vj , is the maximum value over all
possible activities (a = 1,2,…,A):
ln V j  max ln V1 j , ln V2 j ,..., ln VAj 
As Schlenker, Hanemann and Fisher (2009) note that, if in the decomposition of the error
term into a parcel effect and an activity effect (uaj = vj + εa), the error associated with the
activity effect is not extreme-value distributed, then there is no closed-form solution for the
expected value of the value equation; and the generic hedonic expression
ln V j  X j    j
is an approximation to the land-value envelope of the possible rents over all activities.
Land value per hectare in the Municipalities of Chile
1997
Looking only at activities,
certainly there is a strong
correlation with median
municipal land values and
whether or not comuna has
vineyards and fruit production.
Or can we explain this by more
basic characteristics?
2007
Land value per hectare in the Municipalities of Chile
yes
500
If the municipality has Fruits more than 5% of agricultural area
200
Fruit
300
400
Wine grapes
100
No
UF/HA
100
200
300
400
500
600
If the municipality has Vineyard more than 1% of agricultural area
1997
2007
No
yes
Spatial Durbin model – SDM – available in Stata – includes both
endogenous and exogenous interaction effects (LeSage and Pace,
2009; Elhorst, 2010; Vega and Elhorst, 2013).
W is determined by distance (inverse). We use random effects,
because with fixed we lose the climate information.
SEM : spatial error model
  M
X’s – What explains per-hectare land values?
- Size of the transaction (parcel size), climate, soils,
distance to markets and population centers, local
population density.
Land prices and Land size
Median data at Comuna level
-5
-5
0
0
5
ln(UF/Ha)
5
10
10
15
Mean data at Comuna level
0
2
4
6
ln(Hectares)
lnUF_HA
R-sq=0.05; Beta=-0.26; n=503
Fitted values
8
10
0
2
lnUF_HA
R-sq=0.45; Beta=-0.77; n=503
4
ln(Hectares)
6
Fitted values
8
Market potential is an index summarizing the
distance-weighted incomes of markets near and far:
Region
3
4
5
6
7
8
9
10
14
mean lnMP
13.84
14.78
18.09
17.94
17.21
17.27
16.15
15.34
16.58
R
MPr   Ys e
s 1
 (1 ) d rs
Gs
Gs  G  ws , psh 
Incomes and prices vary over time, parameters from
Félix Modrego, Philip McCann, William E. Foster, M. Rose Olfert. 2014.
“Regional Market Potential and the Number and Size of Firms: Observations and
Evidence from Chile.” Spatial Economic Analysis, 9(3): 327-348.
------- 2015. "Regional entrepreneurship and innovation in Chile: a knowledge
matching approach." Small Business Economics. 44(3): 685-703.
Climate variables
Meteorological stations are not optimally
positioned for used by economists 
Climate variables
5
10
15
20
• Average and standard deviation of temperature and
precipitation, annual, based on monthly records for 533
weather stations from 1964-2012. Municipal stats based on
weighting stations based on distance.
1
2
3
4
5
6
7
Month
Curicó_100km
Parral_100km
Pitrufquén_100km
Talca_100km
San_Felipe_100km
8
9
10
11
Rancagua_100km
Osorno_100km
Los_Angeles_100km
Carahue_100km
12
More X variables
• Municipalities are classified by urban/rural-ness, essentially
according to population. 6 types from very rural to
metropolitan. RIMISP – Berdegue, et al., 2011.
• Whether or not there is mining in the comuna.
• 7 agro-ecological zones: soils, valley vs. piedmont, dryland vs.
irrigation availability, and volcanic/sandy, good for grains.
• Percentages of various types of ag activities from Ag census:
–
–
–
–
–
–
Annual field crops
Problem: While there is
variation cross-section, the % of
Grasslands, pasture
fruits and vineyard do not
Forestry
change very much over time.
Fruits
But, yes, for crops and forestry.
Vineyards
Others: fallow, marginal, structures, unused.
Summary of results
• With a dummy variable (comuna > 1% of land in
vineyards) – close to SAG classification.
• Two periods with Ag Census data for crop
proportions by municipality: 1997 and 2007.
• Using four periods (1980, 1990, 1997, 2007) with
simple % of plantings in 1997 or 2007, or their
average.
• Comparison of results : OLS, RE, SAR, SEM, and
SDM.
SDM with random-effects Number
of obs
356
Group variable: Comuna2 Number of
groups
89
Time variable: periodo
Panel
length
4
lnUF_HA
Coef.
Std. Err.
z
P>z
lnDimensión
-0,40
0,13
6,50
-0,26
3,29
-0,14
-0,38
-0,15
0,00
-2,56
0,04
0,03
2,36
0,83
2,00
0,07
11,65
0,04
0,00
4,80
-8,95
3,87
2,75
-0,32
1,64
-2,02
-0,03
-3,63
3,58
-0,53
0,00
0,00
0,01
0,75
0,10
0,04
0,97
0,00
0,00
0,59
Ln(market pot. Index)
Vinas % ave 9707
Fruit % ave 9707
T_mean
T2_mean
T_cv
PP_mean
PP2_mean
PP_cv
within
= 0,76
between = 0,85
overall
= 0,80
lnUF_HA
Coef.
Región2
5
6
7
8
9
14
Macro_zonas2r
Precordillera
Secano costero
Secano interior
Trumao lomaje
Valle de secano
Valle riego
TipoTF
Metropolitano (mayor 250m)
Rural pluricomunal
Rural unicomunal
Rural-Urbano (C18-40m)
Rural-Urbano (C40-80m)
Std. Err. z
P>z
-1,10
-0,97
-0,39
-0,67
-0,58
0,06
0,92
0,69
0,57
0,40
0,28
0,96
-1,200,23
-1,410,16
-0,680,50
-1,660,10
-2,060,04
0,060,95
0,48
0,09
-0,09
0,65
0,60
0,21
0,21
0,26
0,26
0,25
0,20
0,23
2,290,02
0,350,73
-0,350,73
2,640,01
2,930,00
0,920,36
-0,03
-0,35
-0,22
-0,28
0,19
0,22
0,20
0,20
0,25
0,22
-0,130,90
-1,730,08
-1,120,26
-1,140,26
0,880,38
Bottom line: there is something valuable about having
vineyards.
• From dummy model: Municipalities with vineyards have
25-50% higher median land values than otherwise,
controlling for climate, soils, market potential, etc.
• From % vineyard model: A 1 point increase in percentage
in vineyards increases median value by at least 3%, but
in most models 5-7%, relative to “other” uses – mainly
marginal land, fallow and structures.
• Plus indirect effects, total impact reaches 10 to 20%.
• Ceteris paribus, the proportion of land in fruit production
does not add any information. Ditto for other activities.
This suggests that the added value of vineyard is more
than the value of plantations.
Other results
• Climate important: higher median
temperatures, higher land values, decreasing
rate. Lower precipitation, higher land values,
decreasing.
• As expected, market potential important.
Ceteris paribus, increasing MP from that of the
region with the least to that of the highest, the
gain in value is on the order of 40%
• Elasticity of transaction size is about -0.4.
Possible sources of wine premium in a
municipality’s median per-hectare value.
• Missing variables: endogeneity everywhere? Terroir?
• Special and significantly larger land-tied investments
associated with wine grape production, more than
fruit plantations.
• Spillovers onto value of all land of having local wine
and wineries
– the cachet, prestige associated with having a property or
(second?) home in a wine region.
– Wineries, tourism
Future detective work
• Is it just wine as an aggregate? We have
plantings by variety and asking prices from
classified ads, which contain more information
regarding property aptitudes. Shorter time span.
• Tourism - is it the presence of wineries, not just
the vineyards? Locate wineries, but all or just
of a “prestige” level?
• More, better information on basic
characteristics.
thanks