Natural Resources and Local Communities:
Evidence from a Peruvian Gold Mine
On-line appendix
Fernando M. Arag´on Juan Pablo Rud
May 2012
A Data and main variables
Household data The empirical analysis uses data from repeated cross sec-
tions of the Peruvian Living Standards Survey (ENAHO), an annual house-
hold survey collected by the National Statistics Office. The survey consists
of a stratified household sample representative at the regional level. The
regions are defined for statistical purposes and consider both environmental
conditions (coast, highlands and forest) and geographical location (north,
center and south).
1
We focus on the North Highlands statistical region, the
area where the mine is located and restrict attention only to households with
an employed head.
2
Figure 1 shows the area of study and highlights in grey
the districts in the survey’s universe. Districts are the smallest political ju-
risdictions, usually composed by a main town and a surrounding rural area.
The data set covers 10 years, between 1997 and 2006, and includes in total
more than 7700 households located in 101 districts.
3
The main purpose of the survey is to measure poverty and living stan-
dards. The survey contains detailed information on income, expenditure,
1
In general, these statistical regions are larger than departments and do not necessarily
share the same boundaries.
2
This filter reduces the sample by just 46 observations and does not affect the results.
3
It represents an average of 770 observations per year.
1
socio-demographics (such as gender, age, educational attainment of indi-
viduals), composition of the household, housing characteristics like access
to public utilities and construction materials, and self-reported incidence of
health problems and exposure to crime. The data set also has extensive
information on prices and agricultural activity at household level.
To quantify exposure to the mine’s center of activities, Cajamarca city,
we construct a measure of the distance from the household’s location to the
city. This measure varies at district level. In particular, we measure distance
as the length of the shortest route between the main town of the district and
Cajamarca city using the existing road network.
4
We perform the calculation
using the ArcGIS software and maps produced by the Ministry of Transport
of Peru. The road map corresponds to the network available in 2001 and
includes only tracks usable by motorized vehicles. The measure of distance
ranges from 0 to 400 km, with an average value of 100 km. As we will
discuss below, we use this threshold to define districts close and far from
the city. Figure 1 shows the districts with households included in the survey
sample, and highlights in dark grey the districts within 100 km to the city.
Note that the sample includes districts in the vicinity of other cities, such as
Chachapoyas.
Table 1 shows some summary statistics of the main variables from the
household survey. We estimate the means and standard errors using sample
weights and clustering by primary sampling unit to account for the sampling
design.
Constructing a measure of real income Our main outcome variable
is the real income per capita. To construct this variable we divide the net
nominal income per capita by a local consumer price index. As a price index
we use the value of the poverty line, as calculated by the National Statistics
Agency (INEI).
To construct this index, the INEI sums the value of food and non-food
minimum consumption baskets (INEI, 2010 p. 13-18). The food consumption
basket reflects food required to meet minimum nutritional requirements. The
4
The results are robust to alternative measures of distance.
2
Figure 1: Districts in sample, by distance to Cajamarca city
3
Table 1: Summary statistics of household data
Variables Mean Standard
N=7,738 error
Household head
Years of education 5.4 0.1
Age 47.4 0.2
% female 15.7 0.4
Household
Income per capita 212.0 3.4
Consumption per capita 190.2 2.2
Poverty line 173.3 0.3
% poor 63.5 0.5
% extreme poor 33.9 0.5
% urban 36.5 0.5
% access to electricity 38.1 0.6
% access to piped water 59.1 0.6
Nr. Household members 4.7 0.03
Nr. Income earners 2.0 0.01
Distance to Cajamarca city (km) 97.0 0.7
Note: The mean and its standard error are calculated using
sample weights and clustering by primary sampling unit. In-
come, consumption and poverty line are measured in Nuevos
Soles. In the period of analysis, the average exchange rate
was 1 US dollar=3.2 Nuevos Soles.
4
composition of the food basket remains stable over the period of analysis, but
the prices are updated every year. The prices are obtained from the household
survey and are calculated as the average of each department’s urban and rural
area.
5
The resulting value of this minimum food basket corresponds to the
extreme poverty line threshold.
The non-food consumption basket includes goods from major consump-
tion groups such as clothing, transportation, health services, entertainment,
and housing. The value of this consumption basket is calculated using prices
collected in main cities (like Cajamarca, Trujillo, Chachapoyas and Chicla-
yo).
6
The rural prices are assumed to be the same as urban prices.
A main concern is that the poverty line may fail to capture the actual
change in local cost of living. Nonetheless, there are two reasons that justify
the use of the poverty line as a price deflator. First, the poverty headcount
in the sample is 65 percent. This implies that the median household is poor
and hence its consumption basket may not be too different from the one used
to calculated the poverty line.
Second, we compare the evolution of the poverty line in Cajamarca city
with the official consumer price index used by the National Statistics Office.
This price is used to report city’s inflation and it is only available for major
cities (such as Cajamarca). Figure 2 depicts both variables, normalized to be
equal to 100 in year 1997 for Cajamarca city. Note that the poverty line has
a similar trend than the official consumer price index. This evidence suggests
that the poverty line captures a relevant dimension of the local cost of living.
We further explore the robustness of the results to alternative price de-
flators in two ways. First, we compare Cajamarca city to other cities. In
that case we use the official consumer price index as a price deflator (see
Section B.3.3). Second, we combine information from the poverty lines with
proxies of self-reported housing rents to construct price deflators at district
level. The results are similar to the ones using the poverty line (see Table 5
in the Appendix A2).
5
In our sample, it means there are 7 different values of the poverty line each year.
6
The data used to calculate this index, however, is not available in the household survey.
5