aHa
n, C
nit
Available online 29 April 2015
rva
fits
ran
pec
based on growth data from actual small scale landholders who were incentivized to convert lands to teak planta-
t yield models to solve for the optimal Faustmann rotation and compute the net pres-
nagers
patterns. Economic theory suggests that, on
making land use decisions. Understanding these private tradeoffs, as design or add new incentive programs.
oncern to the Panama
ng and water delivery
s believed to influence
available to the canal
Forest Policy and Economics 56 (2015) 48–56
Contents lists available at ScienceDirect
Forest Policy an
e linclude habitat for biodiversity conservation, carbon storage, and
hydrological regulation (Millennium Ecosystem Assessment, 2005;
Balvanera et al., 2013). One approach to realizing the public benefits
(Wohl et al., 2012; Ogden et al., 2013; Ogden and Stallard, 2013). Forest
cover is believed to act like a sponge, reducing flooding in the wet
season and delivering water more slowly to the canal during the dryingly recognize that working landscapes, producing commercial timber,
can also provide multiple public benefits relative to more intensive
agricultural uses, such as cattle ranching (Kirby and Potvin, 2007;
Paquette and Messier, 2010; Hall et al., 2011a). These benefits may
Panama's economy. Low flows and floods are a c
Canal Authority (ACP). Both can disrupt shippi
from the canal. Land use surrounding the canal i
strongly the rate and timing that water becomesthey exist, is an important starting point for policy reform and for
projecting land use change.
Land managers can capture the private value of timber production,
and timber production can be awealth enhancing activity for landman-
agers in many tropical regions (Montagnini and Mendelsohn, 1997;
Piotto et al., 2010; Hall et al., 2011b). Public agencies and NGOs increas-
Forests generate particularly important ecosystem services in the
Panama Canal Watershed (PCW) and have been targeted for incentive
programs (Wallander et al., 2007; Simonit and Perrings, 2013). The
Panama Canal Watershed is important because the Panama Canal pro-
vides clean water to many of the residents of Panama, in addition to
the canal's role as a globally important shipping route, central toof forested landscapes is to incentivize landow
ranching to forestry. Such incentive schemes
⁎ Corresponding author at: Yale School of Forestry an
Prospect St., New Haven, CT 06511, United States.
http://dx.doi.org/10.1016/j.forpol.2015.04.001
1389-9341/© 2015 Elsevier B.V. All rights reserved.average, land managers
ernative land uses while
cated (Mason and Plantinga, 2013), and it is imperative to understand
the private tradeoffs that land managers face prior to attempting toweigh the benefits of one land use against altYield projection
Faustmann model
Cattle ranching
Landowner compensation
Ecosystem services
1. Introduction
The decisions of individual land mapatterns. We compare the NPV from forestry to cattle and find that site characteristics, discount rates, and market
prices are all important factors in influencing the land manager's decision to switch from cattle ranching to planta-
tion forestry. We find that traditional cattle ranching is economically competitive, in terms of NPV, with and may
often outperform teak plantations within the PCW. This result is robust to the teak yield model selected.
© 2015 Elsevier B.V. All rights reserved.
drive landscape land use
the expectation that once the land has been converted to forest, it will
remain in that state even if payments stop (Alix-Garcia and Wolff,
2014). Designing cost effective incentive schemes for forests is compli-Tectona grandis ent value (NPV) of a teak plantation to a private land manager as a starting point for understanding land-use
Keywords: tions.We use a suite of well fiTeak–cattle production tradeoffs for Panam
scale producers
Stephanie F. Stefanski a,b, Xiangying Shi a,c, Jefferson S.
a Yale University, Yale School of Forestry & Environmental Studies, 195 Prospect St., New Have
b Duke University, Nicholas School of the Environment, 450 Research Dr., Durham, NC 27708, U
c Shan Shui Conservation Center, Beijing, China
d Smithsonian Tropical Research Institute, Apartado 0843-03092, Panamá, Panama
a b s t r a c ta r t i c l e i n f o
Article history:
Received 1 December 2014
Received in revised form 31 March 2015
Accepted 8 April 2015
Reforestation and forest conse
mains about relative net bene
bent land uses such as cattle
(Tectona grandis) and native s
j ourna l homepage: www.ners to switch from cattle
are often designed with
d Environmental Studies, 195Canal Watershed small
ll d, Andres Hernandez d, Eli P. Fenichel a,⁎
T, United States
ed States
tion are important issues in the Panama Canal Watershed (PCW). Uncertainty re-
of profit-maximizing timber rotations compared to the net present value of incum-
ching. The scientific and popular literatures have displayed enthusiasm for teak
ies plantations. We estimate a realistic yield model for teak, an exotic tree species,
d Economics
sev ie r .com/ locate / fo rpo lseason; particularly towards the end of the dry season, when low flows
are likely to be most severe (Ibáñez et al., 2002; Ogden et al., 2013).
The government of Panama (GOP) and the Panama Canal Authority
are undertakingmeasures to combat deforestation and encourage refor-
estation in the canal watershed. In 1992, the GOP enacted law 24
(Ley 24) that incentivized landowners to reforest (FAO, 2002). The GOP
and theACP secured loans and grants from the international community
and invested significant funds of their own in land management activi-
traditional cattle ranching, the incumbent activity. Finally, we compare
the private NPV from forestry to cattle. This deterministic comparison
represents a “best case” scenario for lands likely to be targeted with
incentive schemes and provides a lower bound on the size of payments
required to encourage land managers to switch from cattle ranching to
teak plantation forestry. While land holders are subject to institutional
conditions and may also have nonmarket preferences, NPV analysis
is a critical starting point for understanding the tradeoffs land holders
3
analysis, and cedro espino is a low plantation value product. In 2013,
49S.F. Stefanski et al. / Forest Policy and Economics 56 (2015) 48–56ties, aimed at increasing forest cover and improving rural livelihoods in
the PCW. Programs have covered the onetime fixed costs of land con-
version from pasture to forests, i.e., establishment costs (personal com-
munication A. Cerezo, ACP). Through the 2000s forest cover increased in
Panama (Wright and Samaniego, 2008). Data from 2008 indicate that
over 6000 ha of the PCW have been planted with forest plantations.
Nevertheless, only 2% of thewatershed is currently in forest plantations.
Hansen et al. (2013) show that the earlier reforestation trend in Panama
has reversed, raising questions about the state of forests in the Panama
CanalWatershed. This is in stark contrast to Dale et al. (2003), who pre-
dicted that much of the cleared land east of the Panama Canal would be
converted to plantation forestry by 2020. So where are the trees, and
why haven't land managers converted fields to plantation forestry as
predicted? Are the existing establishment incentives insufficient? Per-
haps forestry is not as profitable, relative to alternative land uses, as ear-
lier predictions suggested.
The profitability of plantations in the PCW, relative to alternative
land uses, such as traditional cattle ranching, is not well understood.
Few studies using data from Panama have modeled tree growth and
yield, determined profit maximizing rotations, and/or compared the
net present value of forestry to alternative land uses, e.g., cattle
ranching. Moreover, to our knowledge, no such studies have focused
on the PCW. Ugalde and Gomez (2006) completed a countrywide
assessment of teak (Tectona grandis) growth and yield for the National
Environmental Authority (ANAM) and USAID. They found that most
teak plantations in Panamawere poorlymanaged, but that under inten-
sive management teak plantation forestry can be profitable. Griess and
Knoke (2011) analyzed the return on teak and native species in well
managed plantations in western Panama and concluded that some
native species can outperform teak under appropriate conditions. How-
ever, Griess and Knoke still found teak to be profitable. Finally, Coomes
et al. (2008) found that the benefits of bundling teak production with
carbon credits could be a superior financial strategy than traditional cat-
tle management in the Ipeti-Emberá community east of Panama City.
There have been a wide variety of incentive schemes and tax breaks
and even immigration visas in Panama to encourage reforestation,
though these are not restricted to teak.1 Currently, there are three
large commercial plantations in the PCW. These are located on lowacid-
ity limestone that is unusual for the PCW. However, these commercial
operations face substantially different incentives and possess substan-
tial different soils than most small scale farmers, which are the focus
of our study. Finally, international timber investment and
management organization operate in the Darien, which has low acidity
soil (pH of 6.0–6.9), and teak appears to growwell there. The Darien is a
region to the east and well removed from the PCW.
We contribute to the literature in three ways. First, we provide the
first yield estimates for teak based on data from a sample of actual
farm sites that converted to forestry within the PCW as part of an
incentive program. We develop realistic yield projections for forest
operations in the PCW using tree growth data from 11 stands managed
by local landmanagers as part of a portfolio to generate income. Second,
we contribute to the literature by analyzing the conversion of tradition-
ally managed cattle pasture to teak timber production in the PCW,
conditional on the existing incentive structures. To do this, we solve
for the optimal Faustmann forest rotation based on our new yield
models. This enables us to calculate the present value of net revenue
to the land holder, or net present value (NPV), sometimes called the
land expectation value (LEV) (Chang, 1998), to a land manager from
engaging in plantation forestry.2 Then, we calculate the NPV from
1 http://vivatropical.com/panama/panama-law-24-reforestation-investment/.
2 LEV is common in forestry, but since we are comparing forestry and non-forestry landuses we use NPV.stands ranged between 19 to 25 years of age. This panel data is
unbalanced because of the loss of initial stands through premature
(non-profitable) harvest or land conversion. Therefore, there are
fewer older stands in the sample. This suggests censorship of poor
performing stands. As a result, our yield estimates will be biased
towards greater yields. Regardless of this bias, our estimates are
3 In future work, we plan to address credit constraints, subsidize schemes, uncertainty,
and non-market preferences, all of which may shift decision thresholds relative to the
standard NPV analysis (Fenichel et al., 2014).
4 FundaciónNaturawas established in 1991with themandate of undertaking extension
projects within the PCW in order to enhance biodiversity conservation.
5 Logistic regression results suggest that observed site characteristics are uncorrelated
with the probability that the site remained in plantation forestry over the study period.
6 High valued cedro espino can be “mined” from native forests. High quality wood is be-
lieved to come from plantations exceeding 50 years, but the value may not be greatface and how large other factors must be to shift decisions.
Understanding yield and profitability of plantations under real world
conditions is important because realistic yield models and measures of
the opportunity cost of planting forests are critical to determining the
mix of land uses and whether or not incentive programs are likely to
be an efficientway of restoring forests that generate ecosystem services.
2. Methods and data
2.1. Description of data
We assemble a novel dataset to analyze tree growth. The dataset
comes from the Native Species Reforestation Project (PRORENA) on
behalf of Fundación Natura, a local non-governmental organization
(NGO) (Johnson et al., 2007).4 In 2006, PRORENAmeasured tree planta-
tions established on small landholder farms with a combination of
grants and loans between 1996 and 2003. PRORENA sampled all 41
farms recruited into the loan program. The farms were distributed
throughout the watershed and the distribution does not follow a clear
pattern. These were primarily low income, rural farmers, and details
about the participants are reported in Johnson et al. (2007).We extend-
ed the PRORENAdataset by re-measuringplantations in 2013. However,
6 of the 11 stands had been repurposed. Fivewere transitioned to native
species and onewas converted to agriculture, suggesting that teak plan-
tation forestry was not profitable on those lands.5 Some land managers
whose stands did not persist to re-measurement expressed a very dis-
couraging view of teak during informal conversations. We developed
an extended PRORENA PCW dataset consisting of the 11 original stand
plus 5 re-measurements, which we simply refer to as the PRORENA
dataset hereafter.
In 2006, plantation plots were sampled using a random selection of
10-meter radius circular plots within each block. For large blocks,
multiple plots were used, though a block represents a single stand and
plots were later aggregated. The centroids of multiple plots were 80 m
apart. The age, diameter at breast height (1.3 m, dbh), and total height
of all trees greater than 1 cm dbh within each plot were measured,
generating 1744 observations. The PRORENA project collected data on
all observed tree species when plantations were between 4 and
10 years old. The two most common species were cedro espino
(Pachira quinata) and teak. Data analysis focuses on the 11 teak stands.
Other species were not planted in sufficient quantities to support
6enough to offset the forbearance.
informative about the upper bound of potential production of teak on
working landscapes in the PCW.
Official loan rates in Panama may differ from loan rates accessible to
farmers and low-income families. Surveys of Panamanians give an an-
nual interest rate of 12% (Duffy et al., 2001). Official bank loan rates
given by the Banca Panameña in 2010 are 7.53–8.11% (Durante,
2011), but Banco Nacional de Panama advertises subsidized rates be-
tween 2–6% for livestock operations and as low as 4.5% for forestry.10
Based on these estimates, we test our model using a range of discount
rates.
Table 1
Summary statistics for 11 teak plantationsmeasured by PRORENA (Johnson et al., 2007) in
the Panama Canal Watershed. Five plantations persisted to be re-measured in 2013.
Variable Observations Mean Std. dev. Min Max
Age at first measurement 11 8.45 1.21 6 10
9 Grazing decisions are mostly driven by beef cattle. Other animal products require
much greater degree of capital investment and specialization.
10 https://www.banconal.com.pa/, however it is unclear if small scale farmers can access
50 S.F. Stefanski et al. / Forest Policy and Economics 56 (2015) 48–56Using the extended PRORENA PCW data, we constructed a volume
variable. Volumetric coefficients are not available for PCW plantation
teak so we use the coefficients from Perez Cordero and Kanninen
(2003). Volume per hectare is calculated using a three-step process
based. First we calculated the volume per reported stem, V, according to
ffiffiffiffi
V
p
¼ f 1 þ f 2  dbh ð1Þ
where dbh is diameter at breast height (cm) and f1 and f2 are regression
coefficients from the dbh–volume relations and are −0.0884 and
0.0297 respectively (Pérez Cordero and Kanninen, 2003). 7Second, the
merchantable volume per stem, Vm, was calculated following Perez
Cordero and Kanninen (2003) according to
Vm ¼ V  1−g1  qið Þ
g2
 dbhð Þg3
 
: ð2Þ
The Pérez Cordero and Kanninen (2003) regression coefficients are
g1 = 0.7839, g2 = 2.4149 and g3 =−2.4175.8
Themerchantable volume per hectare, qi, depends on the number of
stems per hectare. PRORENA plots often contained more than one
species and often contained large residual trees that were present
prior to planting. The sum of all trees of one species in one plot may
not be comparable across plots because of the different densities of
species. Therefore, the third step was to normalize stands to a per
hectare basis by tree count. For the first set of measurements, we
assume that the maximum tree count among all plots is the pure
stand density. This assumption biases our volume calculations upwards,
leading to over optimistic growth. For the 2013 measurements stem
adjustments were made individually for each stand. Table 1 gives
summary statistics for the PRORENA teak stands.
In addition to tree data, the PRORENA project collected soil property
data at the farm level, when unavailable at the farm level we used
national soil maps (Instituto de Investigacion Agropecueria de Panama,
2006). These data include soil texture (percentage of sand, clay and
silt) and pH. Slope, aspect, and elevation were recovered using a digital
elevation model (DEM) generated using the software package ENVI
(Exelis).
We assembled financial data from various sources (Table 2). Griess
and Knoke (2011) report that total cost to harvest a plantation in
Panama at age 25 is $1,300/ha, and that the initial establishment cost
(the cost of starting a plantation) is about $3,700. Griess and Knoke
(2011) and González (2004) present a stand establishment cost of
about $5,000. The Tropical Timber Market Report (ITTO, 2013) lists cur-
rent import prices for Panama teak logs as $350–550 per m3. This range
of teak prices can be explained in part by a size at harvest premium. In
the literature, the value for teak depends on the dbh. For example,
Bermejo et al. (2004) state that commercial teak must have a diameter
of 10 cm under the bark at both ends. A manual for teak growers in
Costa Rica (González, 2004) lists 2003 teak timber stumpage prices for
three different categories of diameter size in three different zones. The
literature clearly suggests that a constant price for teak would not
reflect market conditions.
We adopt an approach that allows theprices of harvested teak to rise
with the size of logs. Actual teak grading depends on diameter. In what
follows, we use yield models that provide a harvestable volume per
hectare at a given age. However, those models preserve a one-to-one
mapping between dbh and age, and between age and volume per hect-
are. Therefore, the pricing equation can be expressed in terms of volume
per hectare with a rescaling of parameters. This is convenient because it
helps us reduce the number of variables needed to be tracked in the
model. Using our data, and verifying general relationships with
7 Sensitivity analysis to these parameters did not qualitatively affect our results.
8 Sensitivity analysis to these parameters did not qualitatively affect our results.proprietary timber contracts, we find that price per cubic meter in dol-
lars, p, conditional on the cubic meters coming from a representative
hectare of PCW teak plantation, is approximately p =
131.10 + 0.89qi(t), where qi(t) is harvest volume per hectare as a func-
tion of the age, t, that the timber comes from. In the analysis below price
m−3 works out to range between about $175 and $200 per cubic meter,
reflecting the low value of Panama plantation teak — especially teak
grown in the Panama Canal Watershed (these prices correspond with
prices informally reported by teak growers). The price equation is
then multiplied by the harvested volume to calculate gross revenue.
Moreover, by using this price equation we implicitly assume that teak
is grown to a marketable size, and this assumption is satisfied in our
numerical results.
In 2000, nearly 79% of agricultural land in Panamawas used for cattle
grazing — mostly for beef (Wright and Samaniego, 2008).9 Wright and
Samaniego (2008) argue that cattle are less profitable than crops, but
that cattle are less risky and can be held indefinitely until cash is needed.
There is a long history of using cattle as part of a capital portfolio in de-
veloping countries and in Latin America in particular (Jarvis, 1974;
Zimmerman and Carter, 2003). The flexibility, liquidity, and “low risk”
nature of cattle, relative to plantations and crops, may explain why cat-
tle ranching continues despite reforestation incentives. This suggests
that a necessary, but not sufficient condition, to convert from cattle to
other land uses, is that the alternative land uses must provide a greater
NPV.
Duffy et al. (2001) use survey data to parameterize land-use decision
simulation models. They report that an average head of cattle weighs
266 kg and a maximum stocking rate of two heads of cattle per hectare.
However, observations of operations in the region suggest that one
head of cattle per hectare or fewer is a more accurate stocking density.
Current reports estimate a price per head of cattle in Panama at
$1.06–$1.64 per kg (Cigarruista, 2013). The average cost to maintain a
hectare of pasture is $150 per year (personal communication). With
these parameters, we compute the revenue per head to be $281.96–
436.24. Taking the midpoint price, we calculate the profit per hectare
of land in cattle to be (359.10–150) × 1 = $209.10. This is in line
with, but slightly less than, the per hectare net revenue from cattle
used by Simonit and Perrings (2013).
We use the private opportunity cost of capital for the discount rate.
Age at second measurement 5 22.20 2.28 19 25
Slope percent 11 0.86 0.78 0.5 0.2
Elevation 11 68.14 28.24 34.8 130.0
Sand 11 53.25 8.00 44 68
Silt 11 20.25 4.20 16 28
Clay 11 26.50 9.55 14 40
pH 11 4.48 0.50 4.00 5.90this preferred financing.
therefore we do not present these results.
Table 2
Financial parameters for revenue estimation of cattle and plantations in Panama.
Submodel Variable Price (US$) Source
Teak Price of timber $250–550/m3 “Tropical Timber Market Report” (2013),
González (2004), Griess and Knoke (2011)Production cost $5000
ed o
al h
andi
rna
51S.F. Stefanski et al. / Forest Policy and Economics 56 (2015) 48–562.2. Modeling yield
The basis for economic analysis for reforestation in the PCW is the
yield function. We model and forecast volume, qi, for teak per hectare
for a representative stand in the PCW based on locally derived data
from farmers in the watershed who actually responded to reforestation
incentives.11 We focus on the yield function
qi ¼ e
b0þb1 f tð Þþxβþϵ
ð3Þ
which relates forest age t, where f is a function of t, to volume, qi, and
allows for a vector of covariates x, potentially including percentage of
sand, silt, and clay in the soil, elevation, slope, and pH.12 b1 is a parame-
ter to be estimated in association with age dependent growth, and the
site specific vector β is a vector of parameters to be estimated in associ-
ationwith observable site covariates, b0 is an additional parameter to be
estimated, and ϵ is a mean zero error term. Site indices are not well
developed for the PCW, therefore we use site characteristics as explicit
covariates in our yield models. The form of model (3) nests Chang's
(1984) expanded logarithmic-reciprocal yield function. Conrad (2010)
suggests f = t−1, and Chang (1984) suggests f = t−2. It is generally
good practice to include base terms when including higher order
terms in a regression model (Wooldridge, 2002; Gelman and Hill,
2007), so we also try a specification with both t−1 and t−2. Conrad
(2010) makes no suggestions about the elements of x. Chang (1984)
suggests that x contain site index or characteristics along with site
characteristics interacted and planting density interacted with age. We
do not observe planting density at the sites. In order to capture this
unobserved variation we use random intercept (random effects) and
random slope models. These enable us to preserve degrees of freedom
while capturing unobserved site specific variation, including the
variation thatmay come from planting density. The random slope spec-
ification acts like interacting unobserved site specific characteristics
with the inverse of age.
Model (3) is log transformed and estimated by linear regression.We
first estimated models by ordinary least squares (OLS). The PRORENA
Estimate bas
$1,300/ha fin
Return on investment Discount rate 5–13% (T. gr
11–12% (inte
Bank loan rates 7.53–8.11%
Annual interest rate 12%
Livestock operations Average cow weight 266 kg
Max stocking rate/ha 1
Pasture maintenance/ha/year $150
Market price of beef/kg $1.06–1.64teak dataset displays heteroskedasticity over time and across stands,
as evidenced through residual stands and the Breusch–Pagan/Cook–
Weisberg test for heteroskedasticity (Cameron and Trivedi, 2010,
pp. 100–104). Heteroskedasticity could be caused by underlying differ-
ences across stands such asmanagement variation or site-specific char-
acteristics. Heteroskedasticity can lead to inefficient OLS estimates and
is addressed through obtaining robust standard errors. In order to ac-
count for heteroskedasticity across stands and to account for possible
clustered errors correlated with age groups, we obtain cluster-robust
11 This creates a selection bias, because these farmers likely had greater expectations of
teak outperforming cattle than farmers who chose not to adopt the incentives. This bias
favors teak.
12 Conrad (2010) also suggests the functional form Vm = a1t+ a2t2− a3t3+ xβ+ ϵ. We
find that estimating this form does not produce parameters with theoretically consistent
signs.Second, repeated samples of the sameplantation are not independent.
These repeat observations form a panel dataset.We account for the panel
nature of thedata andunobservable site characteristics byusing a random
effects generalized least squares (GLS) estimation approach.13
Third, because unobserved heterogeneity, including planting densi-
ty, can affect growth rates in varying ways overtime (Chang, 1984),
we estimate a mixed effects model that includes a random slope for
the inverse of age or age squared. This enables us to account for the
fact that unobserved site characteristics such as planting stem density
can affect slope terms as well as the intercept (Gelman and Hill, 2007).
Each specification is run with and without a full suite of covariates
(slope percent, elevation, sand, clay, silt, and pH). The dataset is too
small to estimate the teak models with the full set of covariates. We
select pH as the focal covariate because we find that it generally has a
precisely estimated relationship to growth.
Following estimation we use the model to forecast yield. Model
(3) is estimated as ln qið Þ ¼ b0 þ 1t b1 þ xβþ ϵ , where we have
expressed the error term as a vector since some of our specifica-
tions partition the variance. Conditional on estimated parameters
and mean levels for covariates E ln qið Þð Þ ¼ b
̂
0 þ b
̂
1 f tð Þ þ xβ̂ , hats
represent the best estimates of parameters and the bar is
the mean level of site characteristics. To use the model to make
projections we back transform model (3). By Jensen's inequality
expðE ln qið Þð Þ≠Eð exp ln qið Þð Þ (Cameron and Trivedi, 2005). To cor-
rect for the bias introduced by Jensen's inequality we back transform
model (3) as Q tð Þ ¼ E qi tð Þð Þ ¼ exp b
̂
0 þ b
̂
1 f tð Þ þ xβ̂ þ σ
2
2
 
, where σ2
is the regression's residual variance estimate (Quinn and Deriso, 1999;
Davidson and MacKinnon, 2004).
2.3. The profit maximizing rotationstandard errors (Cameron and Trivedi, 2010, pp. 84–85). However, the
cluster-robust errors do not affect the inference from the OLS, and
n: $3,709 for, 20 years
arvest operations, (t = 25)
s plantations in South America)
l rate of return with t = 25)
Griess and Knoke (2011)
Durante (2011)
Duffy et al. (2001)
Duffy et al. (2001)
Duffy et al. (2001), Personal comm.
Personal comm.
Cigarruista (2013)We use the Faustmann rotation (Conrad, 2010) to find the profit
maximizing rotation conditional on our estimated yield models. The
Faustmann rotation takes into account the expected yield of specific
trees on a plantation,Q(t), themarket price of timber p(Q(t)), which in-
creases with increased wood size in a one-to-one relationship with the
volume harvested, the discount rate, or required rate of return, δ, and
harvest and regeneration costs c to determine the optimal rotation
length, T, for a hectare of forest.14 A landmanager also considers the ini-
tial establishment cost, IC, when computing the switch from cattle to
13 A Hausman test demonstrates that random effects outperforms the fixed effects mod-
el (Cameron and Trivedi, 2010, pp. 235–239).
14 The Faustmannmodel includes any pre-commercial expenses. It is convenient to sim-
ply move this in time to harvest. Our yield models also account for any pre-commercial
management and to the best of our knowledge commercial thinning is uncommon for teak
in the PCW.
forestry (Chang, 1998). However, it has been common for outside
agents to cover the initial conversion cost in Panama. Furthermore,
the initial establishment costs are fixed and do not affect the first
order conditions for an optimal rotation, conditional on engaging in
forestry. Therefore, it is easy to consider the role of initial cost on con-
version after the analysis is complete. The optimal rotation length max-
imizes the net present value, π, of the parcel by choosing rotation
length, T (Conrad, 2010).
π ¼
p Q Tð Þð ÞQ Tð Þ−c½ 
eδT−1
−IC ð4Þ
When the maximizing value of T is substituted into Eq. (4), then
Eq. (4) provides the maximum achievable net present value of the
land under teak plantation management. The Faustmann land value
π(T*) must exceed the net present value associated with alternative
land uses. Otherwise, the land manager will devote the land to an
alternative use that is more profitable than forestry. This is important
in relation to the Panama case study, where farmers are being encour-
aged to switch from livestock to other land uses, which may include
teak plantations.
The net present value of a profit maximizing forest rotation is
compared to the net present value of a traditional cattle operation.
The net present value of a traditional cattle operation is computed as
NPVcattle ¼
n price−costð Þ
δ
, where n is the number of cattle per hectare.
There are no establishment costs for cattle ranching, because traditional
pasture is the incumbent land state.
3. Results
3.1. Yield
The age–volume relationship is critical for forecasting yield.Wehave
a large number of potential specifications, and model selection among
non-nested models with random effects and random slopes is compli-
cated. First, we use Wald tests (F-tests in the case of OLS) against a
null model to eliminate poorly preforming models. Model fits had to
yield Wald statistics with p-values less than 0.05 to be considered
candidate models. All OLS regressions failed this test and generally
provided imprecise estimates of coefficients. We present OLS models
in Table 3 for completeness. All random effects and random slope
models satisfied the Wald statistic criterion, suggesting the importance
of unobservable heterogeneity across stands (Table 3). Second, we
focused on models where the null hypothesis that the coefficient
associated with the inverse of age or inverse age squared was equal to
zero is rejected with a p-value less than 0.05. pH was the only covariate
that increased explanatory power. Furthermore, we found that
Table 3
Model 2 select regression results for teak in the Panama Canal Watershed: OLS is ordinary least squares, RE includes a random effect of site, and RS includes a site specific random slope
associated with the age variable. Standard errors are in parenthesis.
Model number estimation M1 OLS M2 OLS M3 OLS M4 OLS M5 RE M6 RE M7 RE M8 RE
Inverse age, t−1 −10.77 −11.62 −13.40⁎⁎ −13.55⁎⁎
(8.05) (7.45) (6.08) (5.99)
Inverse age squared, t−2 −68.16 −69.05⁎ −80.22⁎⁎⁎ −79.66⁎⁎⁎
(41.32) (38.31) (28.51) (28.52)
pH 1.27⁎ 1.22⁎ 1.22⁎ 1.16⁎
(0.69) (0.69) (0.68) (0.66)
Constant 4.25⁎⁎⁎ 3.95⁎⁎⁎ −1.32 −1.47 4.63⁎⁎⁎ 4.22⁎⁎⁎ −0.82 −0.98
(0.84) (0.54) (3.13) (3.03) (0.69) (0.46) (3.08) (2.98)
Volume at age 30 (m3/ha) 104.9 98.6 99.2 89.9 91.6 83.2 88.6 79.1
N 16 16 16 16 16 16 16 16
σ
2 1.522 1.437 1.300 1.235 0.668 0.591 0.688 0.591
Wald statistic 1.79 2.72 2.74 3.23 4.86 7.92 7.89 10.67
0.0
$25
$9,7
$3,3
1 RS
3.72
1)
⁎
4)
.52
7)
2
9
4
,483
10# (20) $5,393# (17) $8,381# (12) $8,647# (13) $5,228
62
52 S.F. Stefanski et al. / Forest Policy and Economics 56 (2015) 48–56p-Value 0.202 0.121 0.102
Optimal NPV 2% (years) $27,414# (18) $30,828# (15) $23,703# (16)
Optimal NPV 4% (years) $11,439# (16) $13,154# (14) $10,826# (15)
Optimal NPV 8% (years) $4,080# (14) $4,814# (13) $3,928# (14)
Model number estimation M9 RS M10 RS M1
Inverse age, t−1 −11.67⁎⁎⁎ −1
(4.62) (4.7
Inverse age squared, t−2 −61.19⁎⁎⁎
(24.68)
pH 1.17
(0.6
1/pH
1/(t × pH)
Constant 4.51⁎⁎⁎ 4.07⁎⁎⁎ −0
(0.49) (0.34) (2.8
Volume at age 30 (m3/ha) 72.0 56.8 82.8
N 16 16 16
σ
2 0.310 0.074 0.43
Wald statistic 6.38 6.15 11.0
p-Value 0.012 0.013 0.00
Optimal NPV 2% (years) $12,058# (22) $8,895 (17) $15
Faustmann Optimal NPV 4% (years) $4,810 (20) $3,701 (17) $6,1
Optimal NPV 8% (years) $1,570 (17) $1,284 (15) $1,9
⁎ p b 0.1.
⁎⁎ p b 0.05.
⁎⁎⁎ p b 0.01.
# Exceeds cattle NPV.(17) $1,867 (15) $3,186# (12) $3,276# (12) $2,61473 0.028 0.005 0.019 0.005
,481# (19) $19,164# (22) $20,377# (17) $17,872# (22) $18,322# (17)
45# (17) $67,656# (19) $8,525# (16) $7,107# (20) $7,650# (16)
81# (15) $2,522 (17) $3,000# (15) $2,320 (17) $2,680# (15)
M12 RS M13 RS Chang RS Cattle
⁎⁎⁎ 45.36
(28.76)
−75.17⁎⁎⁎ −319.67⁎ −314.10⁎
(24.48) (158.25) (185.93)
1.08⁎ 0.96
(0.62) (0.60)
−43.24⁎⁎
(19.98)
191.43
(146.70)
−0.62 −1.93 12.29⁎⁎⁎
(2.75) (2.79) (3.80)
67.5 37.4 39.9
16 16 16
0.282 0.254 0.305
11.64 14.80 13.50
0.003 0.002 0.004
# (23) $12,985# (18) $19,128# (13) $19,769# (13) $10,455
including an interaction term between the age effect and pH (or the in-
verse of pH) did not improve model forecasts. Table 3 provides a range
of models estimates that capture the breadth of model uncertainty.
Observations of teak size cover a narrow age range. The teak data
points between ages 5 and 10 are widely distributed whereas there
are only a few data points for older age stands (Fig. 1). The different
model specifications vary in how they weigh repeat observations from
the same farm (i.e., older stands) and in how they control for variations
in pH. This leads to variation in model forecasts (Fig. 1), which reflects
limited data and continued uncertainty with respect to teak growth
on rural small scale farms in the PCW.
Chang's approach of using f(t)= t−2 producedmodels that general-
ly had a better fit and lead to growth rates reaching an asymptote at
younger ages than the traditional approach suggested by Conrad of
using t−1. Including both terms leads to a peak and a decline in volume
(Fig. 1, Chang and M13). Excluding pH (or pH−1) led to greater yield
forecasts for the OLS and random effects models and lower yield fore-
casts for the random slope models. Overall, including pH led to more
consistent estimates across models. Our approximation of Chang's
model yields declining growth after a peak, and appears to be over
fitting our relatively small dataset. Despite the variation in forecasts,
all models forecast substantially less volume by age 30 than those
reported from professionally managed plantations outside of the PCW,
3.2. Rotation length, NPV, and comparison to livestock operations
Under the Faustmann model optimal rotation lengths for teak are es-
timated to be between 15 (for a 8% discount rate and M12) and 23 years
(2%discount rate forM11). Randomeffect and randomslopemodelswith
discount rates b 4% suggest rotations between 16 and 23 years. This com-
pares to Ugalde and Gomez (2006) reported rotations for teak plantation
forestry of 20 to 25 years. As the discount rate declines, the net present
value of teak plantation increases, reaching a NPV of $17,872 and
$12,985 per hectare under a 2% discount rate forM7 andM12, respective-
ly. As the discount rate increases to 4% (8%), these values forM7 andM12
fall to $7,107 ($2,320) and $5,393 ($1,867), respectively. We have as-
sumed that establishment costs are covered by someone other than the
land managers, as was the case for the stands in our dataset. If the land
manager had to cover $3,700 in establishment costs, these plantations
would operate at a loss with discount rates of 8%, which is in the likely
range of small scale farms opportunity cost of capital.
The NPV of livestock operations also depends on the discount rate.
We use a baseline profit of $209.10 per cow per hectare per year. The
net present value of a single cow per hectare reaches a high of
$10,455 at a 2% discount rate and declines to $5,228 ($2,614) for a 4%
(8%) discount rate (Table 3). In the best case scenario, a low tomoderate
discount rate is required to encourage the land manager to replace a
2
53S.F. Stefanski et al. / Forest Policy and Economics 56 (2015) 48–56e.g., approximately 500 m3 at age 30 (Griess and Knoke, 2011),
250 m3 at age 33 (Ugalde and Gomez, 2006) and approximately
300m3 at age 30 (Simonit and Perrings, 2013). Themost optimistic pro-
jection, given by OLSM1without covariates, only reaches 105m3/ha by
age 30. However, this model has poor statistical properties and should
not be used for forecasting. Models that have acceptable statistical
(i.e., Wald p-values b 0.05) and theoretical (i.e., non-shrinking stand
volume) properties average 78 m3 by age 30.
In our economic analysis we focus on M7 and M12, which have
precise parameter estimates and good overall modelfit. Bothmodels in-
clude pH as covariate. Furthermore, these twomodels have the greatest
divergence in growthpaths formodels includingpH (Fig. 1). By focusing
on these two models, we can make robust general conclusions about
teak–cattle tradeoffs accounting for model uncertainty. These two
models suggest between 88.6 and 67.5 m3/ha at age 30. Nevertheless,
we prefer models M11 and M12 because the random slope models
provide more precise estimates and better capture the unobserved
heterogeneity. Yet, focusing on M7 relative to M11 provides for greater
contrast, which is important given the uncertainty.
0
20
40
60
80
100
120
140
0 5 10 15 20
Vo
lu
m
e
,m
3 /h
a
AgeFig. 1. Teak growth models from Table 3 and datcattle pasture with a teak plantation, conditional on M7 (Fig. 2), but
this requires an establishment costs subsidy. If the landholder believes
that M12 is more accurate, even with a subsidy for establishment, the
landholder prefers cattle for all but low rates (i.e., rates below 4%). At
an 8% discount rate both yield models suggest that teak is not competi-
tive with cattle.
The comparison between activities suggests that the discount rate,
which is the opportunity cost of capital, is an important feature in
whether or not a land manager would optimally choose to switch
from livestock operations to teak. We have used the same discount
rate for forestry and livestock in our analysis, but (Fig. 2) enables the
reader to make comparisons across discount rates. A risk adverse land
manager may demand a risk premium on forestry leading him to opt
for cattle, which is more liquid and a less risky asset than a plantation.
This is particularly true given the model uncertainty associated with
teak yields. For example, all of our models suggest that teak growth
could not come close to covering a 2 percentage point risk premium
relative to cattle. Finally, the National Bank of Panama appears to offer
subsidized loans for agricultural activities, such as cattle ranching. It is
5 30 35 40
M5
M6
M7
M8
M9
M10
M11
M12
M13
Chang's model
dataa based on the extended PRORENA dataset.
a54 S.F. Stefanski et al. / Forest Policy and Economics 56 (2015) 48–56bless clear if these loans are available for forestry plantations, suggesting
another reason a farmer may use lower discount rates for cattle than
forestry. The issue of the appropriate rate of discount and frictions in
the credit markets is therefore important for explaining the forest
transition or lack thereof in Panama.
4. Discussion and conclusion
We present yield models for teak based on data from plantations
established on small and medium sized farms. These plantations were
established with the goal of improving livelihoods and environmental
benefits in the PCW, and these farms represent areas most likely to be
targeted by incentive programs for land use conversion. Therefore, the
models we present provide accurate, if somewhat imprecise, forecasts
of expected teak yields on PCW.We use a range of models to character-
ize the robustness of our finds to model uncertainty. We use these
models to inform the expected land use value of teak plantations as
alternative land uses to the current cattle pastures.15
There has been enthusiasm in the literature and popular media for
establishing teak and native species plantations in Panama and else-
where. Teak is highly valued. High quality teak, not teak grown in
Fig. 2. Solid curves are theprivateNPVof Teak, assuming that initial establishment costs are paid by
of teak if the land holder has to fund the initial establishment. The NPV of cattle at a given discou
15 Where biases do enter our models, these biases work in favor of teak and against cat-
tle, which makes our result that teak plantations have a hard time economically competi-
tion with cattle in the PCW all the more striking.Panama, commands up to $600 m−3, stoking the excitement for teak.
Furthermore, teak plantations have been successful in Panama's Darien,
which has soil acidity between 6.0–6.9 (Instituto de Investigacion
Agropecueria de Panama, 2006), and is well outside of the PCW. But
teak grows poorly on the acidic soils of the farms of the PCW. We
found average commercial volume at age 30 to be much less than half
of that reported by Ugalde and Gomez (2006) for an average of sites
across Panama. Our volumes are one third to one quarter the volumes
used in recent assessments of PCW ecosystem services (Simonit and
Perrings, 2013) that used these unrealistically high projections for the
soils of the PCW (Lugo et al., 1997). Designing incentives needs to be
based on an accurate portrayal of the land use and land performance
without and with the incentive.
Teak grows best on fertile soils with high values for base cations
(particularly Ca++ and Mg++) and poorly on clay soils (Weaver,
1993). The former condition is often associated with neutral to slightly
acidic soils, while the latter is often associated with low pH acidic soils
in the tropics. While insufficient training and follow through on man-
agement interventions may have played a role, we hypothesize that
teak on these farms underperformed largely due to poor site conditions
(average pH 4.5, Table 1). Follow-up with farmers suggests that they
continued management activities after the end of the financial support
provided by the project. Johnson et al. (2007) found that our teak
study sites had far inferior performance relative to teak on comparison
sites in the Soberania National Park that has a moderate pH (van
anextensionprogram.All curves canbe reduced by approximately $3,700 to recover theNPV
nt rate are by a horizontal dashed line. The top figure is M7 and bottom figure is M12.
Breugel et al., 2011). Given that Ugalde and Gomez (2006) suggest that
teak can provide vastly superior financial performance under intensive
native trees: implications for reforestation and land restoration in Mesoamerica.
55S.F. Stefanski et al. / Forest Policy and Economics 56 (2015) 48–56management regimes than unmanaged stands, it is also important
that extension efforts continue far beyond plantation establishment
(also see Hall et al., 2011b). If the policy goal is to shift a greater fraction
of the PCW into treed landscapes, then forecasts need to be based on
species appropriate for the chosen sites. “Appropriate” must include
biophysical characteristics and behavioral and management programs
of the land managers who actually manage the would-be converted
lands. This latter requirementmay rule out highly intensive silvicultural
or land management techniques.
Given the apparently limited sites appropriate for teak in the PCW,
our results suggest that from a private landmanager's perspective cattle
ranching is highly competitivewith teak, andwhen risk, experience and
norms, and other factors beyond the deterministic NPV, which all work
against teak in the PCW are considered, the NPV of teak likely needs to
exceed the NPV or cattle by a comfortable margin for land conversation
to happen. This means that plantation teak will likely only be able to
play aminor role in reforestation of the PCW, particularly under current
market conditions. One alternative is to select native tree species with
better yields (Piotto et al., 2009). However, native species markets for
plantation timber are generally less developed in Panama. So the
increased growth will have to offset the “price discount” relative to
teak that these products provide. Experience with native species is
also limited. For example, cedro espino was widely planted in the
watershed. This species appears to grow rapidly, but does not produce
heartwood until an advanced age, which leads to very low valued
wood on 20 to 30 year rotations. Mixed production of cattle and trees
is also a promising possibility (Murgueitio et al., 2011; Riedel et al.,
2013).
Teak plantations may provide many of the ecosystems services, and
land managers may be able to participate in ecosystem service markets
or receive payments for plantations (Kraenzel et al., 2003).16 Such
payments would have to be sufficient to get managers to move from
livestock to teak. Structuring payments for ecosystem services is
complicated (Jack et al., 2008; Kaczan et al., 2013). However, back of
the envelop calculations, using our more optimistic M7, suggests that
if annual participation payments are used, then to incentivize the aver-
age land manager in our data, which is likely biased towards slightly
better than average sites, with a 12% discount rate to switch from cattle
to forestry, would require the payer to cover all of the conversion and
establishment costs and then pay the land manager $85/ year in perpe-
tuity (e.g., an ecosystem services rental fee). This sum seems unreason-
ably high when one considers that it is 41% of the profit margin of
raising cattle. Fenichel et al. (2014) suggest that there may be opportu-
nities to incentivize land use shift through reforms in credit and savings
markets. This approach makes use of the incentivized land use,
e.g., plantations, outperforming the current land use, i.e., cattle pasture
at low discount rates.
Our results are disappointing from a timber production and from a
conservation and ecosystem service perspective. However, it would be
even more disappointing to move forward with teak plantations or
teak plantation incentive programswithout recognizing the strong eco-
nomic headwinds that such investments are likely to face. Conversely,
recognizing these headwinds may provide incentives to develop native
species plantation programs that may actually provide greater
ecosystem services than teak plantations.
Acknowledgments
This study was funded by a Faculty Research Grant through the F.K.
Weyerhaeuser Memorial Fund administered by the Center for Business
and the Environment at Yale, from NSF award EAR-1360369
16 Simonit and Perrings (2013) suggest that native species plantations would provide
more ecosystem services than teak, though their yield models are questionable.For. Ecol. Manag. 261, 1553–1557.
Hall, J.S., Love, B.E., Garen, E.J., Slusser, J.L., Saltonstall, K., Mathias, S., van Breugel, M.,
Ibarra, D., Wishnie, M.H., Ashton, M.S., 2011b. Tree plantations on farms: evaluating
growth and potential for success. For. Ecol. Manag. 261, 1675–1683.
Hansen, M.C., Potapov, P.V., Moore, R., Hancher, M., Turubanova, S.A., Tyukavina, A., Thau,
D., Stehman, S.V., Goetz, S.J., Loveland, T.R., Kommareddy, A., Egorov, A., Chini, L.,
Justice, C.O., Townshend, J.R.G., 2013. High-resolution global maps of 21st-century
forest cover change. Science 342, 850–853.
Ibáñez, R., Condit, R., Angehr, G., Aguilar, S., GarcÍa, T., MartÍnez, R., Sanjur, A., Stallard, R.,
Wright, S.J., Rand, A.S., Heckadon, S., 2002. An ecosystem report on the Panama Canal:
monitoring the status of the forest communities and the watershed. Environ. Monit.
Assess. 80, 65–95.
Instituto de Investigacion Agropecueria de Panama, 2006. Zonificacion de Suelos de
Panama por Nivelele de Nutrientes. Instituto de Investigacion Agropecueria de
Panama, p. 24.
ITTO, 2013. Tropical timber market report. ITTO 17, 1–15.
Jack, B.K., Kousky, C., Sims, K.R.E., 2008. Designing payments for ecosystem services:
lessons from previous experience with incentive-based mechanisms. Proc. Natl.Collaborative Research: Planning and Land Management in Tropical
Ecosystem; Complexities of land-use and hydrology coupling in the
Panama Canal Watershed, and with support from ForestGEO, the forest
component of the Smithsonian Institution Global Observatory (SIGEO).
The project also forms part of the Agua Salud Project, a collaboration be-
tween the Smithsonian Tropical Research Institute, the Panama Canal
Authority, and other institutions. We thank Estrella Yanguas for logisti-
cal support and Edwin Hernandez, Erick Agrazal, Guillermo Fernandez,
Miguel Nunez, Carlos Diaz, Sam Somerville, and Anabel Rivas for help
in data collection and entry. Evan Ray provided research assistance.
Mark Ashton provided helpful insights during the research effort. Sun
Joseph Chang provided helpful comments on an earlier version of this
paper.
References
Alix-Garcia, J., Wolff, H., 2014. Payments for ecosystem services from forests. Ann. Rev.
Resour. Econ. 6 (online early).
Balvanera, P., Uriarte, M., Almeida-Lenero, L., Altesor, A., DeClerk, F., Gardner, T., Hall, J.,
Lara, A., Laterrra, P., Pena-Claros, M., Matos, D.M.S., Romero-Duque, L.P., Vogl, A.,
Arreola, L.F., Carro-Borrero, A.P., Gallego, F., Jain, M., Little, C., Xavier, R.O., Paruelo,
J.M., Peinado, J.E., Poorter, L., Azcarrunz, N., Correa, F., Cunha-Santino, M.B.,
Cunha-Santino, M.B., Hernandez-Sanchez, A.P., Vallejos, M., 2013. Ecosystem services
research in Latin America: the state of the art. Ecosyst. Services 2, 56–70.
Bermejo, I., Cañellas, I., Miguel, A.S., 2004. Growth and yield models for teak plantations in
Costa Rica. For. Ecol. Manag. 189, 97–110.
Cameron, A.C., Trivedi, P.K., 2005. Microeconometrics Methods and Applications.
Cambridge University Press, New York.
Cameron, A.C., Trivedi, P.K., 2010. Microeconometrics Using Stata Revised Edition. Stata
Press, College Station, Texas.
Chang, S.J., 1984. A simple production function model for variable density growth and
yield modeling. Can. J. For. Res. 14, 783–788.
Chang, S.J., 1998. A generalized Faustmann model for the determination of optimal
harvest age. Can. J. For. Res. 28, 652–659.
Cigarruista, H., 2013. Precios del ganado en pie se disparan [WWWDocument]. Cap. Finance
http://www.capital.com.pa/precios-del-ganado-en-pie-se-disparan 3/30/2014.
Conrad, J.M., 2010. Resource Economics. Cambridge University Press, New York.
Coomes, O.T., Grimard, F., Potvin, C., Sima, P., 2008. The fate of the tropical forest: carbon
or cattle? Ecol. Econ. 65, 207–212.
Dale, V.H., Brown, S., Calderon, M., Montoya, A., Martinez, R.E., 2003. Estimating baseline
carbon emissions for the eastern Panama Canal Watershed. Mitig. Adapt. Strateg.
Glob. Chang. 8, 323–348.
Davidson, R., MacKinnon, J.G., 2004. Econometric Theory and Methods. Oxford University
Press, New York.
Duffy, S.B., Corson, M.S., Grant, W.E., 2001. Simulating land-use decisions in the La
Amistad Biosphere Reserve buffer zone in Costa Rica and Panama. Ecol. Model. 140,
9–29.
Durante, E.E., 2011. Panamá: Evolución económica durante 2010. In: México, D.F. (Ed.),
pp. 1–40.
FAO, 2002. Teak (Tectona grandis) in Central America. Forest Plantations Working Paper.
19. Food and Agriculture Organization of the United Nations, Rome.
Fenichel, E.P., Adamowicz, W.L.V., Ashton, M.S., Hall, J.S., 2014. Incentive systems for
forest-based ecosystem services with missing financial service markets. Working
Paper. School of Forestry & Environmental Studies, New Haven, CT, Yale.
Gelman, A., Hill, J., 2007. Data Analysis Using Regression and Multilevel/Hierarchical
Models. Cambridge University Press, New York.
González,W.F., 2004.Manual para productores de Teca (Tectona grandis L.f) en Costa Rica.
pp. 1–115.
Griess, V.C., Knoke, T., 2011. Can native tree species plantations in Panama compete with
teak plantations? An economic estimation. New For. 41, 13–39.
Hall, J.S., Ashton, M.S., Garen, E.J., Jose, S., 2011a. The ecology and ecosystem services ofAcad. Sci. U. S. A. 105, 9465–9470.
Jarvis, L.S., 1974. Cattle as capital goods and ranchers as portfolio managers: an applica-
tion to the Argentine cattle sector. J. Polit. Econ. 82, 489–520.
Johnson, A.E., Cedeno, N., Craven, D., 2007. Una Evaluación Del Programa: “Reforestación
En Fincas Ganaderas En La Cuenca Hidrográfica Del Canal De Panamá”. PRORENA
(Native Species Reforestation Project in Panama) to Fundación Natura, p. 133.
Kaczan, D., Swallow, B.M., Adamowicz, W.L.V., 2013. Designing payments for ecosystem
services (PES) program to reduce deforestation in Tanzania: an assessment of
payment approaches. Ecol. Econ. 95, 20–30.
Kirby, K.R., Potvin, C., 2007. Variation in carbon storage among tree species: implications
for the management of a small-scale carbon sink project. For. Ecol. Manag. 246.
Kraenzel, M., Castillo, A., Moore, T., Potvin, C., 2003. Carbon storage of harvest-age teak
(Tectona grandis) plantations, Panama. For. Ecol. Manag. 173, 213–225.
Lugo, A., Gillespie, A., Brown, S., Chapman, J., 1997. Caracteristicas de plantactiones
arboreas en los tropicos. In: Lugo, A.E. (Ed.), Rendimiento y aspectos silviculturales
de plantaciones maderas en America Latina. US Forest Service International Institute
of Tropical Forestry, Puerto Rico, pp. 1–62.
Mason, C.F., Plantinga, A.J., 2013. The additionality problem with offsets: optimal
contracts for carbon sequestration in forests. J. Environ. Econ. Manag. 66, 1–14.
Millennium Ecosystem Assessment, 2005. Ecosystems and Human Well-Being: Synthesis.
Island Press, Washington D.C.
Montagnini, F., Mendelsohn, R.O., 1997. Managing forest fallows: improving the econom-
ics of swidden agriculture. Ambio 26, 118–123.
Murgueitio, E., Calle, Z., Uribe, F., Calle, A., Solorio, B., 2011. Native trees and shrubs for the
productive rehabilitation of cattle ranching lands. For. Ecol. Manag. 261, 1654–1663.
Ogden, F., Stallard, R.F., 2013. Land use effects on ecosystem service provisioning in
tropical watersheds, still an important unsolved problem. Proc. Natl. Acad. Sci. U. S.
A. 110 E5037.
Ogden, F., Crouch, T.D., Stallard, R.F., Hall, J.S., 2013. Effect of land cover and use on dry
season river runoff, runoff efficiency and peak storm runoff in the seasonal tropics
of central Panama. Water Resour. Manag. 49, 1–20.
Paquette, A., Messier, C., 2010. The role of plantations in managing the world's forests in
the Anthropocene. Front. Ecol. Environ. 8, 27–34.
Perez Cordero, L.D., Kanninen, M., 2003. Provisional equations for estimating total and
merchantable volume of Tectona grandis trees in Costa Rica. Forest Trees Livelihoods
13, 345–359.
Piotto, D., Craven, D., Montagnini, F., Alice, F., 2009. Silvicultural and economic aspects of
pure and mixed native tree species plantations on degraded pasturelands in humid
Costa Rica. New For. 39, 369–385.
Piotto, D., Craven, D., Montagnini, F., Federico, A., 2010. Silvicultural and economic aspects
of pure andmixed native tree species plantations on degraded pasturelands in humid
Costa Rica. New For. 39, 369–385.
Quinn, T.J., Deriso, R.B., 1999. Quantitative Fish Dynamics. Oxford University Press, New
York.
Riedel, J., Dorn, S., Plath, M., Mody, K., 2013. Growth, herbivore distribution, and herbivore
damage of timber trees in a tropical silvopastoral reforestation system. Ann. For. Sci.
70, 75–86.
Simonit, S., Perrings, C., 2013. Bundling ecosystem services in the Panama Canal
watershed. Proc. Natl. Acad. Sci. U. S. A. 110, 9326–9331.
Ugalde, A.L., Gomez, F.M., 2006. Perspectivas económicas y ambientales de las
plantaciones de teca bajo manejo sostenible, en Panamá. In: USAIDC.,
Environmental Authority Panama (ANAM), National (Eds.), USAID, CATIE, and
National Environmental Authority, Panama (ANAM), p. 76.
van Breugel, M., Hall, J.S., Craven, D., Gregoire, T.G., Park, A., Dent, D.H., Wishnie, M.H.,
Mariscal, E., Deago, J., Ibarra, D., Cedeno, N., Ashton, M.S., 2011. Early growth and
survival of 49 tropical tree species across four sites differing in soil fertility and rainfall.
For. Ecol. Manag. 261, 1580–1589.
Wallander, S., Lauterbach, S., Anderson, K., Chou, F., Muir Grossman, J., Scholegel, C., 2007.
Existing markets for ecosystem services in the Panama canal watershed. J. Sustain.
For. 25, 311–336.
Weaver, P.L., 1993. Tectona grandis L.f. teak. Silviculture of Puerto Rico. U.S. Forest Service.
Wohl, E., Barros, A., Brunsell, N., Chappell, N.A., Coe, M., Giambelluca, T., Goldsmith, S.,
Harmon, R., Hendrickx, J.M.H., Juvik, J., McDonnell, J., Ogden, F., 2012. The hydrology
of the humid tropics. Nat. Clim. Chang. 2, 655–662.
Wooldridge, J.M., 2002. Econometric Analysis of Cross Section and Panel Data. The MIT
Press, Cambridge, Massachusetts.
Wright, S.J., Samaniego, M.J., 2008. Historical, demographic, and economic correlates of
land-use change in the Republic of Panama. Ecol. Soc. 13.
Zimmerman, F.J., Carter, M.R., 2003. Asset smoothing, consumption smoothing and the
reproduction of inequality under risk and subsistence constraints. J. Dev. Econ. 71,
233–260.
56 S.F. Stefanski et al. / Forest Policy and Economics 56 (2015) 48–56