Forest Policy and Economics 57 (2015) 3137

Contents lists available at ScienceDirect

Forest Policy and Economics

journal homepage: www.elsevier.com/locate/forpol

A comprehensive analysis of teak plantation investment in Colombia

Hctor I. Restrepo , Sergio A. Orrego
Universidad Nacional de Colombia, Sede Medelln, Forest and Climate Change Research Group, Colombia

a r t i c l e i n f o a b s t r a c t

Article history: A nancial assessment of forest investments is comprehensive if the analysis includes reliable yield estimates,
Received 29 August 2014 land expectation value (LEV) and risk calculation. All of these aspects were considered and applied to teak plan-
Received in revised form 5 May 2015 tations in Colombia, an emergent economy where high forest productivity, low opportunity cost of land, and de-
Accepted 6 May 2015
creased nancial/economic risk have substantially contributed to promote forest investments. The von
Available online 16 May 2015
Bertalanffy non-linear mixed effect model was used to estimate forest yields using data collected from 31 perma-
Keywords:
nent sample plots, measured over a 17 year period. A stochastic version of LEV along with other nancial criteria
Forest yield was calculated by using a computer algorithm and Monte Carlo simulation. Finally, probabilities obtained from
von Bertalanffy stochastic nancial calculations were used in logistic models to estimate probabilities of success for a forest plan-
Faustmann tation project, a measure of risk assessment, after changing land prices. Results suggest that the potential forest
Financial risk analysis productivity (i.e., the biological asymptote) ranges from 93 to 372 m3 ha1. The mean annual increment is
27.8 m3 ha1 year1, which is attained 6 years after the forest plantation is established. Protability analyses
for teak plantations in Colombia suggest a LEV of US$7000 ha1. The risk analyses indicate negligible nancial
risk for forestlands whose prices are lower than US$2000 ha1.
2015 Elsevier B.V. All rights reserved.

1. Introduction potential and environmental constraints are related to site specic

Forest management is likely to be inuenced by both variables relat-
ed to the production process (e.g., species type, yield prediction, silvicul-
tural treatments, harvesting rates) and exogenous factors (e.g., public
policy, macroeconomics, and performance of global forest product mar-
kets). Although exogenous factors are crucial to evaluate the potential
success of forest investments, these factors are not determined by land-
owner decisions. Unlike exogenous factors, some technical and nancial
variables related to forest production seem to be more inuenced by
forest product producers. Two critical technical issues are the estima-
tion of forest yield and its corresponding uncertainty, and a robust anal-
ysis of techniques and their outcomes.
Forest yield estimation is appropriately determined by using
biologically-based statistical methods. An empirical and well-known
method is the von Bertalanffy model (Fekedulegn et al., 1999; Moser
and Hall, 1969; Pienaar and Turnbull, 1973; Pienaar, 1979; Torres,
2004; Torres and del Valle, 2007; Zapata, 2007; Zeide, 1993). The von
Bertalanffy model is based on a biological theory which conceives that
the growth of any organism is the result of two main and opposite forces
known as anabolism and catabolism (Pienaar and Turnbull, 1973). Like-
wise, forest growth could be thought of as the outcome of two opposite
factors, the unlimited trend of growth or biotic potential and the growth
constraints imposed by the environment (Zeide, 1993). Both biotic
Corresponding author at: Calle 59A # 63-20 Of. 20, 306, Medelln, Colombia.
E-mail addresses: hirestrepoo@unal.edu.co (H.I. Restrepo), saorrego@unal.edu.co
(S.A. Orrego).
factors such as climate, soil and topography or physiography
(Bravo-Oviedo et al., 2007; Clutter et al., 1992; Gnter et al., 2009;
Lugo et al., 1988; Louw and Scholes, 2002, 2006; Vanclay, 1992; Walters
et al., 1989; Wang et al., 2005). All these factors can be included in the
von Bertalanffy model to express forest growth and yield as a function
of covariates in the mixed effect models (Bravo-Oviedo et al., 2007;
Hall and Bailey, 2001; Louw and Scholes, 2006; Wang et al., 2005).
On the other hand, in order to calculate economically optimal rota-
tion lengths and investment protability, nancial analysis is essential.
Even though some researchers have used different nancial criteria to
assess the economic performance of forest projects, the Faustmann
formula is, indeed, the correct solution (Amacher et al., 2009;
Diaz-Balteiro, 1997; Newman, 2002; Samuelson, 1995). The Faustmann
formula, or land expectation value (LEV), allows for calculation of net
present value (NPV) of innite rotations and represents the maximum
amount of money that an investor ought to pay for a unit of land and
earn the minimum acceptable rate of return (Klemperer, 2003). The
age at which LEV attains its maximum value corresponds to the optimal
economic rotation, and indicates the appropriate time to harvest a
stand. Other important nancial criteria, such as NPV and IRR, are also
used in forest nancial analyses, providing complementary and useful
information to evaluate the forest investment performance.
Forest nancial analyses should not be considered exhaustive with-
out a risk evaluation. Risk is especially important when returns to forest
investments are reaped many years after establishing a forest planta-
tion. This unique and crucial characteristic of forest investments has, es-
sentially, two important features. First, investors would feel that forest

http://dx.doi.org/10.1016/j.forpol.2015.05.001
1389-9341/ 2015 Elsevier B.V. All rights reserved.
32 H.I. Restrepo, S.A. Orrego / Forest Policy and Economics 57 (2015) 3137
investments often seem riskier than other types of investments (Ashton
et al., 2001). Second, economic analysis may be difcult due to uncer-
tainties related to nancial variables such as price and interest rate
(Hildebrandt and Knoke, 2011). A method commonly used in risk eval-
uations is the Monte Carlo simulation (Acua and Drake, 2003; Reeves
and Haight, 2000; Rodrguez and Diaz-Balteiro, 2006). Using this com-
putational procedure, samples of input variables are randomly generat-
ed by assuming probability distributions and then computing a matrix
of outcomes that represent the possible values of LEV and other nancial
criteria.
This research aims to propose a comprehensive approach to esti-
mate yield, rotation age, and prots based on biology and economic the-
ory. To the best of our knowledge, this research is likely the rst in Latin
America that explicitly deals with a complete forest investment analy-
sis, including a nancial risk assessment. Indeed, protability analyses
tend to be scarce or limited and few research have been done in
Colombia (Cubbage et al., 2010; Henao, 1982; Lpez et al., 2007, 2010;
Restrepo, 2010; Restrepo et al., 2012), and risk evaluation is virtually
nonexistent. Estimations of forest yield and optimal rotations are
complemented with uncertainty analyses using Monte Carlo simula-
tion. This approach was applied to analyze both technical and economic
issues for forest plantations of teak (Tectona grandis) in Colombia, using
a reliable and unique dataset consisting of biometric measurements
over a 17 year period.
2. Material and methods
2.1. Model estimation
von Bertalanffy's model can be expressed as:
f T A1 expkT ; with n 1m1 ;
n (Romero, 1997; Samuelson, 1995). This time corresponds to the age
where f(T) represents the yield (m3 ha 1), A is the asymptote
(m3 ha 1), T is the time or the stand age (years), k is the intrinsic
growth rate, n is a shape parameter and m is an allometric constant.
The value of the parameter m indicates how a specic forest species re-
produces and survives (Gonzlez, 1994). Thus, the m value indicates the
photosynthetic ability, being specic for each organism and environ-
ment (Pienaar and Turnbull, 1973). The parameter m generally has a
range of [0,1) and (1,3] (Gonzlez, 1994), which also makes von
Bertalanffy's model more exible. A value of m greater than 1 corre-
sponds to a logistic shape for the yield curve; when m is less than 1, it
corresponds to the well-known and concave Mitscherlich curve; when
m is equal to 1, the model takes the name of the Gompertz function
(Pienaar and Turnbull, 1973; Zapata, 2007). Additionally, the m param-
eter is related to the curve's inection point and represents the age at
which the maximum current annual increment is attained, thereby
inuencing the rotation age. The parameter A is consistent with the
equinal principle of system general theory. It suggests that stands
with the same site quality in a wide range of tree densities would ap-
proach similar yields, given time (Pienaar, 1979). Finally, parameters A
and k represent potential environmental effects on growth and yield
since they are related to metabolic outcomes of resource availability
(Brown and Lugo, 1982; Gonzlez, 1994; Lugo et al., 1988; Pienaar,
1979; Restrepo and Alviar, 2010).
The mixed effect models are often one of the statistical methods
used to estimate forest yield (Bravo-Oviedo et al., 2007; Calegario
et al., 2005; Diguez-Aranda et al., 2006; Hall and Bailey, 2001; Hall
and Clutter, 2004; Torres, 2004; Torres and del Valle, 2007; Torres
et al., 2012; Wang et al., 2007; Zapata, 2007), and can be written as:


yi j f i j ; i j i j ; i 1; ; M; j 1; ; ni
where M is the number of groups, ni is the number of observations in
each group, ij is a vector of xed parameters and random values, vij is
a vector of covariates, and i is the error term, assumed to be distributed
as i ~ N(0,2I). The ij vector might be specied as:
i j X i j W i j bi ; 3
where Xij and Wij are matrices of covariates and groups of random ef-
fects, respectively, is a xed effect vector and bi is a random effect vec-
tor. It is usually assumed that all bi are independent, are uncorrelated
with the error term, are distributed as N(0,2b) and have E[b] = 0
(McCulloch and Searle, 2001). Following the mixed effect model theory,
the von Bertalanffy model can be written as:





 ni j
f i j ; i j A i j 1 exp k i j ; i j T : 4
Maximum likelihood estimation methods are used to estimate
Eq. (4), and the statistical performance is evaluated using the Schwarz
Bayesian Criteria (SBC) (Pinheiro and Bates, 2000), expressed as:
AIC 2 logLik 2npar ;
5
BIC 2 logLik 2npar logN;
where AIC is the Akaike Information Criterion, BIC is the Bayesian Infor-
mation Criterion, npar is the number of parameters included in the
model, N is the number of observations and log represents the natural
logarithm of the likelihood function.
2.2. Rotation age and prot calculations
An important aspect in forest economics and management is the de-
termination of the time at which the forest or stand should be harvested
1
that maximizes the LEV, denoted as (Amacher et al., 2009):
LEV 1 exprT 1 pf T exprT c; 6
where LEV is the land expectation value (US$ ha1), p is the stumpage
price (US$ m3), f(T) is a yield function (m3 ha1), r is the discount
rate, and c is the establishment costs (US$ ha1). The present value of
management cost is also included in Eq. (6).
2.3. Financial risk determination
Most of the empirical literature tends to calculate LEV in a determin-
istic way (Amacher et al., 2009; Samuelson, 1995). However, the dy-
namic nature of some relevant variables (e.g., prices, yields, interest
rate), would lead to assess the potential effect of changes in those vari-
ables. A method commonly reported in the literature to accomplish this
type of analysis is the Monte Carlo simulation with which it is possible
to calculate a stochastic LEV. Moreover, it is possible to evaluate the po-
tential success of a forest investment, calculated as the probability that
stochastic LEVmax exceeds a determined land price. The calculated prob-
ability is considered as a measurement of success for a forest invest-
ment, and also can be considered a risk measurement (Clutter et al.,
1992; Klemperer, 2003).
An algorithm written in Visual Basic for Applications was used to
calculate both deterministic and stochastic LEV, rotation age, NPV,
IRR, and their corresponding condence intervals and success prob-
abilities. The algorithm does not use a xed number of iterations.
On the contrary, each simulation stops when a criterion of variance
stabilization is met, which implies a distinct number of iterations in
each simulation. The variance stabilization criterion used was the
2 following:
 . 
s2 abs si siz 2
2 2
7
siz
 H.I. Restrepo, S.A. Orrego / Forest Policy and Economics 57 (2015) 3137 33

where s2 represents the change in the variance which is calculated
with the current (s 2i ) and z previous (s 2i z) variance, and is the
allowed change of variance measured in percentage. Values of z
and of 10 and 0.0001, respectively, were arbitrarily dened.
Finally, the probabilities of success and values of stochastic LEV are
used to estimate the logistic model:

p1 ; 8
1 expabLEV
where a and b are parameters to estimate. The estimated model allows
to evaluate how changes in land prices can affect the shape of the tted
curve of probabilities of success, evaluation that can be of interest to po-
tential investors.
2.4. A case study: teak plantations in Colombia
Forest yield data were measured over the 19902007 period in 31
teak (T. grandis) permanent plots in the Caribbean region of Colombia.
Two to eight careful measurements were available for each plot and
consistency of eld records was veried. The dataset also included
each plot's management activities. Data used to estimate the model
corresponded to unthinned plots and identical management practices.
The geographic coordinates of each plot were used to calculate envi-
ronmental variables such as rainfall, temperature, elevation and slope.
Climate and slope were obtained from WorldClim, version 1.4
(Hijmans et al., 2005) and STRM from NASA (Farr et al., 2007), respec-
tively. The complete database is summarized in Table 1. A number of
23 observations were randomly chosen and used to validate the esti-
mated statistical model using t-student and Wilcoxon tests.
Market prices and local establishment and management costs of
teak plantations in Colombia were used for rotation age and prot calcu-
lations. Stumpage prices were obtained as residual price, after
subtracting all costs and expenses related to harvesting, extraction,
transportation and timber delivery from CIF-America's prices for teak
in the range US$400600 m 3 (ITTO, 2013). In this analysis we also
took into account that timber from plantations tends to have a lower
price than timber from natural forest (Keogh, 2008). Since the stumpage
price may vary with the stand age (Klemperer, 2003), which seems to
be the case for teak plantations in America, the following price equation
was used:

US$6T m3 ; T 20 years
pT ; 9
US$ 120 m3 ; TN20 years
where p(T) is the average stumpage price (US$ m3) for a forest plan-
tation of age T (years). Although the price equation seems to be dened
in an arbitrary way, it is intended to capture the appreciation of the teak
timber over time. The price specication was inserted into Eq. (6), and
the income term, therefore, was rewritten as p(T)f(T)exp( rt). Addi-
tionally, Eq. (6) uses a continuous discount rate (r) which is related to
discrete interest rate as ln(1 + rd). A common rd used by researches to
calculate LEV and NPV is 8% (Cubbage et al., 2007, 2010; Rodrguez
and Diaz-Balteiro, 2006), which corresponds to an r of 7.7% and was
employed here. Monte Carlo simulations were made by assuming a uni-
form distribution (Hildebrandt and Knoke, 2011), a variable number of
iterations, and the variance stabilization criterion denoted by Eq. (7).
Since future values for r and p(T) are more uncertain in Colombia com-
pared to developed countries, values were randomly selected in the
range of mean times 0.8, and mean times 1.2 (Table 2). Finally, all
models were estimated using R, version 2.15.3 (R Development Core
Team, 2013).
3. Results
3.1. Forest yield estimation model
The von Bertalanffy forest yield model estimated is shown in Table 3.
It is a function of four xed effect parameters (1, 2, 3 and 4), an en-
vironmental covariate that corresponds to the annual mean tempera-
ture (Amt), the age of the stand (T), and two random effect values (b1
and b2).
The model can be written as:
f T 242:94 b1 1 exp10:750:38Amt b2 T
4:3
: 10
Given that by assumption random effects have a mean equal to 0, the
estimated asymptote was about 243 m3 ha 1. By using the mean of
the covariate Amt, a maximum mean annual increment (MAI) of
27.8 m3 ha1 year1 was attained at 6 years and denes the optimal
rotation age from a biological perspective. This model was validated by
using 23 observations which were not included in the estimation of the
model. Evidence was not found to reject the estimated model using t-
student and Wilcoxon tests (p N 0.05).
3.2. LEV and nancial risk
LEV, economic rotation age and nancial criteria were also calculat-
ed. Given that ~ 39,800 iterations were performed for all 31 plots, the

Table 1
Summary of dataset used to estimate the forest yield of teak in Colombia.

Variable Short-name Units Mean s.d.a Min. Max.

Forest
Time T Years 6.3 3.1 2.1 16.0
Volume V m3 ha1 125.5 77.6 2.1 362.3

Environmentalclimatic
Annual mean rainfall Amr mm 2176.7 253.6 1186.0 2326.0
Moister month's rainfall Mmr mm 286.2 18.3 217.0 300.0
Driest month's rainfall Dmr mm 33.0 5.9 11.0 38.0
Moister quarter's rainfall Mqr mm 840.2 86.7 501.0 885.0
Driest quarter's rainfall Dqr mm 130.9 25.2 33.0 144.0
Annual mean temperature Amt C 27.2 0.1 26.9 27.4
Hottest quarter's temperature Hqt C 27.7 0.2 27.5 28.3
Coolest quarter's temperature Cqt C 26.8 0.1 26.5 26.9
Temperaturerainfall ratio TR 100 C mm1 1.3 0.3 1.2 2.3

Environmentaltopographic
Elevation E masl 89.0 20.2 40.0 164.1
Slope S Degrees 3.9 1.6 1.3 7.9
a
s.d.: Standard deviation.
34 H.I. Restrepo, S.A. Orrego / Forest Policy and Economics 57 (2015) 3137

Table 2
Range for r and p(T) from which values were randomly selected using a uniform distribu-
tion in Monte Carlo simulations of nancial and risk analyses. T corresponds to stand age.

Variable Units Range (mean)

Discount rate, r % 6.169.24 (7.7)

Stumpage price, p(T) US$ m3 4.8T7.2T (6T)

condence interval for the mean of LEV is virtually the mean itself
due the number of iterations on the standard error. The mean of sto-
chastic LEVmax was US$7000 ha 1, which was reached at 11 years.
However, by considering the maximum and minimum for each
time (hypothetical rotation ages), the stochastic LEV varied from
negative to positive values, with the highest value being
US$25,000 ha 1 (Fig. 1). Fig. 1. Mean, lowest and highest stochastic LEVmax.
The results for each plot shown in Table 4 reveal some interesting
ndings. First, economic rotation ages are different among plots.
While some plots reached the LEVmax in year nine, indicating good site
quality, other plots reached it just before year twenty, indicating poor The estimated model logistic models, using Eq. (8) and values of
site quality. Second, the LEVmax and nancial criteria are also different probabilities of success and stochastic LEVmax contained in Table 4,
among plots. were:
Third, there is a signicant difference in the magnitude of the LEVmax
 1
1
calculated in a deterministic or traditional way and LEVmax computed in P LEVNUS$2000 ha ; 11
1 e3:8130:00178LEV
stochastic form. Since the deterministic LEVmax is not in the condence
interval of the stochastic LEVmax in most of the analyzed plots, there is
 1
1
a signicant statistical difference between them (p b 0.05). In fact, P LEVNUS$5000 ha ; 12
1 e5:010:00096LEV
most of the deterministic LEVmax tended to be lower than those of
their corresponding stochastic LEVmax. If landowners made decisions
 1
1
P LEVNUS$10; 000 ha : 13
based on deterministic instead of stochastic values, they would under- 1 e7:2440:0007LEV
estimate the protability of teak investments as indicated by the sto-
chastic calculations. Results suggest that high values of stochastic It works for a specic location with its corresponding reference land
LEVmax are associated with high probabilities (Table 4). price of US$2000, US$5000, and US$10,000 and could be a tool to make

Table 3
Estimation results for the non-linear von Bertalanffy mixed effect model to predict Tectona grandis forest yield.

Fixed effects

Parameter Value Standard error Degree of freedom t-Value p-Value

1 242.94 10.78 128 22.54 0.00

2 10.75 3.44 128 3.12 0.00
3 0.38 0.13 128 3.02 0.00
4 4.30 0.18 128 23.95 0.00

Random effects

Plot b1 b2 Plot b1 b2
P01 68.33 0.12 P17 76.85 0.07
P02 0.78 0.15 P18 10.47 0.05
P03 41.47 0.08 P19 64.91 0.01
P04 20.95 0.05 P20 6.13 0.02
P05 129.10 0.18 P21 23.49 0.00
P06 4.96 0.00 P22 47.81 0.06
P07 43.18 0.05 P23 27.29 0.03
P08 36.34 0.02 P24 10.89 0.06
P09 45.68 0.06 P25 26.43 0.07
P10 3.10 0.09 P26 19.23 0.10
P11 24.77 0.02 P27 18.40 0.07
P12 72.99 0.06 P28 149.61 0.23
P13 49.10 0.02 P29 7.66 0.05
P14 2.78 0.03 P30 50.76 0.00
P15 11.48 0.06 P31 63.24 0.11
P16 52.44 0.20

Schwarz's Bayesian Criterion

AIC 1264.5 logLik 623.2
BIC 1292.3 N 162
 H.I. Restrepo, S.A. Orrego / Forest Policy and Economics 57 (2015) 3137 35

Table 4
Deterministic LEVmax, economic rotation age, mean, standard deviation and condence interval of the stochastic LEVmax, probabilities to obtain different land prices, net value present and
internal return rate. All monetary values are in 2012 US$ ha1.

Plot Deterministic LEVmax Rotation Stochastic LEVmax (US$ ha1) Probability of successa NPV IRR (%)
(US$ ha1) (years) (US$ ha1)
Mean s.d. 95% CI N2000 N5000 N10,000
b
P01 1582 19 1768 957 1718 1817 0.36 0.00 0.00 1324 10.8
P02 3632 18 3772 1468 3529 4015 0.93 0.25 0.00 2810 13.5
P03 3853b 15 4025 1414 3925 4125 0.94 0.22 0.00 2705 14.8
P04 6740b 13 6917 2004 6811 7023 1.00 0.81 0.09 4401 18.5
P05 14,878b 9 15,155 3402 14,940 15,370 1.00 1.00 0.96 7389 31.5
P06 6832b 12 7095 1958 7021 7170 1.00 0.86 0.10 4175 19.9
P07 4244b 14 4376 1442 4297 4455 0.99 0.31 0.00 2820 15.7
P08 9036b 11 9311 2412 9191 9432 1.00 0.99 0.36 5155 23.2
P09 9941b 10 10,150 2501 10,001 10,298 1.00 1.00 0.49 5400 24.9
P10 5809b 14 6024 1836 5920 6128 1.00 0.67 0.03 3838 17.6
P11 6232b 11 6417 1822 6331 6503 1.00 0.76 0.04 3583 19.9
P12 11,573b 10 11,975 2925 11,685 12,265 1.00 1.00 0.73 6133 27.3
P13 9023b 11 9297 2352 9165 9428 1.00 1.00 0.35 5257 22.7
P14 7568b 11 7835 2116 7664 8005 1.00 0.93 0.16 4334 21.8
P15 5914b 13 6117 1817 6029 6205 1.00 0.69 0.03 3735 18.3
P16 5811b 9 6036 1701 5966 6106 1.00 0.69 0.02 2958 21.2
P17 4123b 10 4355 1399 4272 4437 0.98 0.29 0.00 2363 17.4
P18 6034b 12 6198 1846 6111 6284 1.00 0.71 0.04 3731 18.7
P19 10,668 10 10,686 2648 10,015 11,356 1.00 1.00 0.58 5781 25.4
P20 7225b 11 7423 2019 7327 7518 1.00 0.90 0.12 4194 20.9
P21 8258b 11 8495 2228 8415 8574 1.00 0.97 0.25 4761 22.2
P22 5659b 10 5800 1664 5722 5878 1.00 0.65 0.01 3062 19.9
P23 6489b 10 6651 1843 6549 6753 1.00 0.80 0.06 3574 20.8
P24 7501 10 7625 2028 7431 7819 1.00 0.91 0.15 3952 22.6
P25 9876b 9 10,096 2432 9934 10,259 1.00 1.00 0.47 4850 27.1
P26 6048 11 6183 1734 6035 6331 1.00 0.69 0.02 3535 19.4
P27 7372b 9 7607 1974 7485 7729 1.00 0.91 0.13 3819 23.1
P28 15b 16 532 415 502 562 0.00 0.00 0.00 66 8.1
P29 7820b 11 8085 2139 8017 8153 1.00 0.94 0.20 4453 22.1
P30 9926b 10 10,669 2856 10,212 11,126 1.00 1.00 0.60 5742 24.7
P31 10,482b 10 10,774 2709 10,629 10,918 1.00 1.00 0.57 5849 25.0
a
Probability that stochastic LEV exceeds plausible land prices (US$ ha1) in Colombia.
b
Indicates signicant statistical difference between deterministic and stochastic LEVmax.

decisions if an investor should buy this land or, instead, look for another
piece of land or investment. Curves of the estimated Eqs. (11), (12), and
(13) are shown in Fig. 2. Observe the difference among curves and how
the number of probabilities of success diminishes when land price
changes.
4. Discussion
A comprehensive bioeconomic forest investment analysis, by includ-
ing appropriate yield modeling and economic and risk evaluation, was
proposed and applied to teak forest plantations. The estimated model
successfully predicted the yield of teak in Colombia. The asymptote
Fig. 2. Stochastic LEVmax and its estimated probability of exceeding three specic thresh-
olds. A curve for each threshold price is drawn and contains the same number of points
which corresponds to teak plots.
value of 243 m3 ha 1 was in the range of previous studies:
200 m3 ha1 (Henao, 1982), 200524 m3 ha1 (Bermejo et al., 2004;
Prez and Kanninen, 2005), 144567 m3 ha 1 (Nunifu and
Murchinson, 1999), 193337 m3 ha 1 (Quintero et al., 2012),
172 m3 ha1 (Pandey and Brown, 2000) and 153 m3 ha1 (Thapa and
Gautam, 2005). The MAI obtained of 27.8 m3 ha1 year1 was greater
than the values reported by previous studies: 15 m3 ha1 year 1
(Henao, 1982), 925 m3 ha 1 year 1 (Bermejo et al., 2004; Prez
and Kanninen, 2005), 414 m 3 ha 1 year 1 (Nunifu and
Murchinson, 1999), 2.712.3 m3 ha 1 year 1 (Pandey and Brown,
2000) and 14.5 m3 ha 1 year 1 (Thapa and Gautam, 2005).
The obtained teak rotation ages were lower than those reported by
previous studies in Asia, Africa and America. While in Asia and Africa
teak rotations ranged from 40 to 60 years (Bailey and Harjanto, 2005;
Cubbage et al., 2010; Pandey and Brown, 2000; Thapa and Gautam,
2005), rotation ages in tropical America were lower, 2025 years
(Bermejo et al., 2004; Cubbage et al., 2010; Henao, 1982; Prez and
Kanninen, 2005; Zanin, 2005). The lower rotation ages for most of the
sites could be explained by lower establishment and management
costs and higher forest productivity. On the other hand, nancial opti-
mal rotation age is longer than biological rotation age which is contrary
to the common trend in Northern hemisphere forest investments. It is
attributable to the stumpage price function, p(T), which was used to
capture the effect of the differences of unitary prices among ages of
teak plantations.
The obtained LEVmax was in the range reported by previous studies.
For example, in Indonesia the LEVmax ranged between US$96 (loss)
and US$1851 ha1 (Cubbage et al., 2010), while in Venezuela
and Panama it ranged between US$9800 ha1 (Cubbage et al., 2010)
and US$3500 ha1 (Griess and Knoke, 2011), respectively. Another im-
portant nding regarding LEV was the difference between deterministic
36 H.I. Restrepo, S.A. Orrego / Forest Policy and Economics 57 (2015) 3137
and stochastic calculations. Unlike the results of this research, the deter-
ministic LEVmax was higher than the corresponding stochastic LEVmax in
Brazil (Rodrguez and Diaz-Balteiro, 2006).
The probability results showed some interesting facts. A forest pro-
ject with land prices lower than US$2000 ha1 is likely to be nancially
successful. To the best of our knowledge, our research is probably the
rst in Latin America that incorporates explicitly nancial risk in teak
plantation economic analysis. Although previous risk analyses for teak
plantations were not found, results were compared with risk calcula-
tions for other forest species. For example, Pinus radiata plantations in
Chile had a probability equal to 50% to obtain a positive NPV (Acua
and Drake, 2003), the probability value that is required in order to en-
sure the success of a project (Reeves and Haight, 2000). Finally, the lo-
gistic model allows for the prediction of the success probability for
each reference land price, by using a land price and the LEVmax calculat-
ed with stochastic methods. No previous research was found regarding
the use of logistic models to estimate nancial risk in any forest
plantations.
5. Conclusions
Most of the emerging economies are in the tropics where relative
high rates of biomass growth are observed. However, economic, nan-
cial, political and legal uncertainties may be critical to potential forest
investments. A comprehensive analysis was proposed, including bio-
metrics, economic and risk as determinants of the potential success of
any forest investment in emerging economies. Regarding biometrics, a
unique dataset and a biological model were used to predict forest
yield. Detailed cash ows and prices were used to determine the opti-
mal rotation of teak forest plantations. Finally, evaluation of the risk as-
sociated with teak forest plantations was used to estimate probabilities
of success. Results of the risk assessment of teak plantations in Colombia
indicate that higher values of LEVmax are associated with higher proba-
bilities of success, and that there is no risk at all for forestlands whose
prices are lower than US$2000 ha1. Moreover, the LEVmax average for
teak plantation projects in Colombia was US$7000 ha1, with values
in the range of US$53215,155 ha1.
Acknowledgments
We would like to thank Lawson Creekmore who carefully read the
draft and provided valuable suggestions and comments. We also
thank El Caribe, a forest company which provided data for the estima-
tion of forest yield models.
References
Acua, E., Drake, F., 2003. Anlisis de riesgo en la gestin forestal e inversiones
silviculturales: una revisin bibliogrca. Bosque (Valdivia) 24, 113124. http://dx.
doi.org/10.4067/S0717-92002003000100009.
Amacher, G.S., Ollikainen, M., Koskela, E., 2009. Economics of Forest Resources. Massachu-
setts Institute of Technology, Cambridge.
Ashton, M.S., Mendelsohn, R., Singhakumara, B.M., Gunatilleke, C.V., Gunatilleke, I.A., Evans,
A., 2001. A nancial analysis of rain forest silviculture in southwestern Sri Lanka. For.
Ecol. Manag 154, 431441. http://dx.doi.org/10.1016/S0378-1127(01)00513-8.
Bailey, J.D., Harjanto, N.A., 2005. Teak (Tectona grandis L.) tree growth, stem quality and
health in coppiced plantations in Java, Indonesia. New For. 30, 5565. http://dx.doi.
org/10.1007/s11056-004-1116-5.
Bermejo, I., Caellas, I., San Miguel, A., 2004. Growth and yield models for teak plantations
in Costa Rica. For. Ecol. Manag. 189, 97110. http://dx.doi.org/10.1016/j.foreco.2003.
07.031.
Bravo-Oviedo, A., del Ro, M., Montero, G., 2007. Geographic variation and parameter as-
sessment in generalized algebraic difference site index modelling. For. Ecol. Manag.
247, 107119. http://dx.doi.org/10.1016/j.foreco.2007.04.034.
Brown, S., Lugo, A.E., 1982. The storage and production of organic matter in tropical forest
and their role in the global carbon cycle. Biotropica 14, 161187.
Calegario, N., Daniels, R.F., Maestri, R., Neiva, R., 2005. Modeling dominant height growth
based on nonlinear mixed-effects model: a clonal Eucalyptus plantation case study.
For. Ecol. Manag. 204, 1120. http://dx.doi.org/10.1016/j.foreco.2004.07.051.
Clutter, J.L., Fortson, J.C., Pienaar, L.V., Brister, G.H., Bailey, R.L., 1992. Timber Management:
A Quantitative Approach. John Wiley & Sons, Inc., New York (Reprint edition).
Cubbage, F.W., Mac Donagh, P., Sawinski, J., Rubilar, R., Donoso, P., Ferreira, A., Hoeich, V.,
Morales, V., Ferreira, G., Balmelli, G., Siry, J., Baez, M., lvarez, J., 2007. Timber invest-
ment returns for selected plantations and native forest in South America and the
Southern United States. New For. 33, 237255. http://dx.doi.org/10.1007/s11056-
006-9025-4.
Cubbage, F.W., Koesbandana, S., Mac Donagh, P., Rubilar, R., Balmelli, G., Morales, V.,
de la Torre, R., Murara, M., Hoeich, V.A., Kotze, H., Gonzlez, R., Carrero, O., Frey,
G., Adams, T., Turner, J., Lord, R., Huang, J., MacIntyre, C., McGinley, K., Abt, R.,
Phillips, R., 2010. Global timber investments, wood costs, regulation, and risk.
Biomass Bioenergy 34, 16671678. http://dx.doi.org/10.1016/j.biombioe.2010.
05.008.
Development Core Team, R., 2013. R: A Language and Environment for Statistical Com-
puting. R Foundation for Statistical Computing, Vienna, Austria.
Diaz-Balteiro, L., 1997. Turno forestal econmicamente ptimo: una revisin. Econ. Agrar.
180, 181224.
Diguez-Aranda, U., Granadas-Arias, J., lvarez-Gonzlez, J.G., von Gadow, K., 2006. Site
quality curves for birch stands in north-western Spain. Silva Fenn. 40, 631644
(http://www.metla./silvafennica/full/sf40/sf404631.pdf).
Farr, T.G., Rosen, P.A., Caro, E., Crippen, R., Duren, R., Hensley, S., Kobrick, M., Paller, M.,
Rodrguez, E., Roth, L., Seal, D., Shaffer, S., Shimada, J., Umland, J., Werner, M., Oskin,
M., Burbank, D., Alsdorf, D., 2007. The Shuttle Radar Topography Mission. Rev.
Geophys. 45, 133. http://dx.doi.org/10.1029/2005RG000183.
Fekedulegn, D., Mac Siurtain, M.P., Colbert, J.J., 1999. Parameter Estimation of Nonlinear
Growth Models in Forestry. Silva Fenn 33, 327336 (http://www.silvafennica./pdf/
article653.pdf).
Gonzlez, H., 1994. Anlisis del crecimiento en condiciones naturales de Prioria copaifera
Grisebach por medio de modelos matemticos. Universidad Nacional de Colombia,
Sede Medelln, Medelln.
Griess, V.C., Knoke, T., 2011. Can native tree species plantations in Panama compete with
Teak plantations? An economic estimation. New For. 41, 1339. http://dx.doi.org/10.
1007/s11056-010-9207-y.
Gnter, S., Gonzlez, P., lvarez, G., Aguirre, N., Palomeque, X., Haubrich, F., Weber, M.,
2009. Determinants for successful reforestation of abandoned pastures in the
Andes: soil conditions and vegetation cover. For. Ecol. Manag. 258, 8191. http://
dx.doi.org/10.1016/j.foreco.2009.03.042.
Hall, D.B., Bailey, R.L., 2001. Modeling and prediction of forest growth variables based on
multilevel nonlinear mixed models. For. Sci. 47, 311321.
Hall, D.B., Clutter, M., 2004. Multivariate multilevel nonlinear mixed effects models for
timber yield predictions. Biometrics 60, 1624. http://dx.doi.org/10.1111/j.0006-
341X.2004.00163.x.
Henao, I., 1982. Estudio de rendimientos y rentabilidad en una plantacin de teca (Tectona
grandis L. F.) del departamento de Crdoba, Colombia. Cron. Forest. Medio Amb. 2
(1,2), 178.
Hijmans, R., Cameron, S., Parra, J., Jones, P., Jarvis, A., 2005. Very high resolution interpo-
lated climate surfaces for global land areas. Int. J. Climatol. 25, 19651978. http://
dx.doi.org/10.1002/joc.1276.
Hildebrandt, P., Knoke, T., 2011. Investment decisions under uncertaintya methodolog-
ical review on forest science studies. For. Policy Econ. 13, 115. http://dx.doi.org/10.
1016/j.forpol.2010.09.001.
ITTO, 2013. Tropical Timber Market Report. 17(3). International Tropical Timber
Organization.
Keogh, R., 2008. International pricing mechanism for plantation teak: a proposal to bring
transparency to log markets. ITTO Trop. For. Update 18, 2426.
Klemperer, W.D., 2003. Forest Resource Economics and Finance. McGraw-Hill.
Lpez, A.M., Barrios, A., Ortega, A., Nieto, V.M., Gasca, G., Salamanca, M., 2007. Empleo de
un modelo de crecimiento y rendimiento para la determinacin de la edad ptima de
rotacin de Pinus caribaea var. hondurensis creciendo en la regin oriental de colom-
bia. Rev. Colomb. For. 10, 119126 (http://revistas.udistrital.edu.co/ojs/index.php/
colfor/article/view/2980).
Lpez, J., de la Torre, R., Cubbage, F., 2010. Effect of land prices, transportation costs, and
site productivity on timber investment returns for pine plantations in Colombia. New
For. 39, 313328. http://dx.doi.org/10.1007/s11056-009-9173-4.
Louw, J.H., Scholes, M.C., 2002. Forest site classication and evaluation: a South
African perspective. For. Ecol. Manag. 171, 153168. http://dx.doi.org/10.1016/
S0378-1127(02)00469-3.
Louw, J.H., Scholes, M.C., 2006. Site index functions using site descriptors for Pinus patula
plantations in South Africa. For. Ecol. Manag. 225, 94103. http://dx.doi.org/10.1016/
j.foreco.2005.12.048.
Lugo, A.E., Brown, S., Chapman, J., 1988. An analytical review of production rates and
stemwood biomass of tropical forest plantations. For. Ecol. Manag. 23, 179200.
http://dx.doi.org/10.1016/0378-1127(88)90081-3.
McCulloch, C.E., Searle, S.R., 2001. Generalized, Linear, and Mixed Models. John Wiley &
Sons, Inc., New York.
Moser, J.W., Hall, O.F., 1969. Deriving growth and yield functions for uneven-aged forest
stands. For. Sci. 15, 183188.
Newman, D.H., 2002. Forestry's golden rule and the development of the optimal for-
est rotation literature. J. For. Econ. 8, 527. http://dx.doi.org/10.1078/1104-
6899-00002.
Nunifu, T.K., Murchinson, H.G., 1999. Provisional yield models of teak (Tectona grandis Lim
F.) plantations in northern Ghana. For. Ecol. Manag. 120, 171178.
Pandey, D., Brown, C., 2000. Teak: a global overview. Unasylva 201, 313.
Prez, D., Kanninen, M., 2005. Stand growth scenarios for Tectona grandis plantations in
Costa Rica. For. Ecol. Manag. 210, 425441. http://dx.doi.org/10.1016/j.foreco.2005.
02.037.
Pienaar, L.V., 1979. An approximation of basal area growth after thinning based on growth
in unthinned plantations. For. Sci. 25, 223232.
 H.I. Restrepo, S.A. Orrego / Forest Policy and Economics 57 (2015) 3137
Pienaar, L.V., Turnbull, K.J., 1973. The ChapmanRichards generalization of von
Bertalanffy's growth model for basal area growth and yield in even-aged stands.
For. Sci. 19, 222.
Pinheiro, J.C., Bates, D.M., 2000. Mixed-effects Models in S and S-PLUS. Springer, New
York.
Quintero, M.A., Jerez, M., Flores, J., 2012. Modelo de crecimiento y rendimiento para
plantaciones de teca (Tectona grandis L.) usando el enfoque de espacio de estados.
Rev. Cienc. Ing. 33, 3342.
Reeves, L.H., Haight, R.G., 2000. Timber harvest scheduling with price uncertainty using
Markowitz portfolio optimization. Ann. Oper. Res. 95, 229250. http://dx.doi.org/
10.1023/A:1018974712925.
Restrepo, H.I., 2010. Estimacin del rendimiento forestal con informacin espacialmente
explcita y anlisis nanciero de plantaciones forestales en Colombia. Universidad
Nacional de Colombia, Sede Medelln.
Restrepo, C., Alviar, M., 2010. Tasa de descuento y rotacin forestal: el caso del Eucalyptus
saligna. Lect. Econ. 73, 149164.
Restrepo, H.I., Orrego, S.A., del Valle, J.I., Salazar, J.C., 2012. Rendimiento, turno ptimo for-
estal y rentabilidad de plantaciones forestales de Tectona grandis y Pinus patula en
Colombia. Interciencia 37, 1420.
Rodrguez, L., Diaz-Balteiro, L., 2006. Rgimen ptimo para plantaciones de eucaliptos en
Brasil: un anlisis no determinista. Interciencia 31, 739744.
Romero, C., 1997. Economa de los recursos ambientales y naturales.
Samuelson, P.A., 1995. Economics of forestry in an evolving society. J. For. Econ. 1,
115150.
Thapa, H.B., Gautam, S.K., 2005. Growth performance of Tectona grandis in the western
Terai of Nepal. Banko Janakari 15, 612.
37
Torres, D.A., 2004. Modelacin del crecimiento y produccin en volumen y biomasa de la
teca. Universidad Nacional de Colombia, Sede Medelln, Medelln.
Torres, D.A., del Valle, J.I., 2007. Growth and yield modelling of Acacia mangium in
Colombia. New For. 34, 293305. http://dx.doi.org/10.1007/s11056-007-9056-5.
Torres, D.A., del Valle, J.I., Restrepo, G., 2012. Site index for teak in Colombia. J. For. Res. 23,
405411. http://dx.doi.org/10.1007/s11676-012-0277-x.
Vanclay, J., 1992. Assessing site productivity in tropical moist forest: a review. For. Ecol.
Manag. 54, 257287. http://dx.doi.org/10.1016/0378-1127(92)90017-4.
Walters, D.K., Gregoire, T.G., Burkhart, H.E., 1989. Consistent estimation of site index
curves tted to temporary plot data. Biometrics 45, 2333.
Wang, Y., Raulier, F., Ung, C.H., 2005. Evaluation of spatial predictions of site index obtain-
ed by parametric and nonparametric methods a case study of lodgepole pine pro-
ductivity. For. Ecol. Manag. 214, 210211. http://dx.doi.org/10.1016/j.foreco.2005.04.
025.
Wang, Y., LeMay, V.M., Baker, T.G., 2007. Modelling and prediction of dominant height
and site index of Eucalyptus globulus plantations using a nonlinear mixed-effects
model approach. Can. J. For. Res. 37, 13901403. http://dx.doi.org/10.1139/X06-282.
Zanin, D., 2005. Feasibility of Teak Production for Smallholders in Eastern Panam. Mich-
igan Technological University.
Zapata, M., 2007. Modelos de sitio para plantaciones tropicales. Caso de estudio:
aplicacin de modelos jerrquicos bayesianos. XVII Simposio Nacional de Estadstica,
Cali, Colombia.
Zeide, B., 1993. Analysis of growth equations. For. Sci. 39, 594617.