Professional Documents
Culture Documents
goods are supplied by conserving forests (16). Cash can also allow deforestation; i.e., earlier and later cohort villages have similar PKH
households to substitute forest-intensive goods with market-produced exposure propensities, but their enrollment timing differs because
goods. The promise of regularly receiving cash could also help house- of factors not systematically related to deforestation. Another way
holds smooth their consumption over time, which could affect how to express this idea is that, on the basis of our understanding of how
they use forests (e.g., make them less likely to resort to deforestation the program was rolled out, we assume that, across the country, there
when they experience negative shocks). CCTs can also affect de- are villages that are similar in terms of their poverty, infrastructure,
forestation through the conditionality obligations themselves or and other potential confounding variables, but some were enrolled
by indirectly encouraging compliance with forest-use laws through early and some enrolled late. Consistent with this claim, earlier and
reciprocity or greater government scrutiny [for a complementary later exposed villages are relatively homogeneous in terms of ob-
view of how CCTs can affect environmental outcomes, see (27)]. servable baseline characteristics (table S1), including pre-PKH de-
The net effect of CCTs is thus an empirical question, to which forestation trends (fig. S2). Using only the sample of villages that are
the answer requires a design that facilitates unbiased estimation of eventually exposed to the PKH helps control for unobservable factors
counterfactual rates of deforestation in the absence of the CCTs. Such that determine exposure to PKH and are held in common by all PKH
estimation is challenging when the cash transfers are not randomized villages. Using panel data on annual forest cover loss further helps
across villages. Care must be taken to control for confounding fac- control for time-invariant confounders, both observable and un-
tors that affect both participation and deforestation (Fig. 1). In observable, as well as factors that change over time but not across
Indonesia, for example, PKH and non-PKH villages differ in im- villages (e.g., national economic and political events). For further con-
portant economic and environmental attributes (table S1), including trol, and to increase the precision of the estimates, we also control for
their pre-PKH deforestation trends. changes in precipitation and temperature over time. Because infer-
To estimate PKH’s causal effect on deforestation, we combined ring causality from correlations is challenging in nonexperimental
geo-referenced administrative data, remotely sensed panel data on contexts, we also subject our design to a battery of robustness checks.
annual forest cover loss (21), a field-verified understanding of how
PKH was phased in over time, and statistical methods to estimate
counterfactual deforestation rates in the absence of PKH (Materials RESULTS
and Methods). The way in which PKH was phased in over time can On average, exposure to PKH is associated with a 30% reduction in
be exploited to credibly estimate its causal effects on deforestation forest cover loss [95% confidence interval (CI), 10 to 50%] (Fig. 2),
in participating villages. PKH targeting was a function of time- a conclusion that is robust to how "forest" is defined (table S4) and
invariant subdistrict and household characteristics and of pressure the administrative level at which we conduct the analysis (i.e.,
exerted by the central government on PKH administrators to roll villages or the subdistricts into which groups of villages are nested;
out the program evenly across provinces within the annual budget table S10). If we were to use the count of pixels of forest lost rather
constraint [see the Supplementary Materials (SM)]. This pressure than the area in hectares lost (table S4), exposure to PKH is associated
created some variation in exposure to PKH that was unrelated to with a 16% reduction in forest cover loss (95% CI, 9 to 23%).
DISCUSSION
Estimating causal effects from observational data is challenging.
Nevertheless, the analysis suggests that the estimated negative
correlation between PKH and forest loss can plausibly be interpreted
as reflecting a causal relationship. To provide evidence on which
mechanisms were operative (Fig. 1), we use experimental and
nonexperimental data and predictions from theory (Materials and
Methods). We use data from a 2007 randomized pilot of the PKH,
which was largely in agricultural communities with little or no forest,
to assess the likelihood that the conditionality mechanisms were
operative. In the evaluation of that pilot, no effects on child labor
(agricultural or otherwise) were detected. Moreover, the detectable
health effects were small in the third year and waned by the sixth year,
a pattern that does not match the pattern of the PKH treatment
Fig. 4. Assessing rival explanations for estimated PKH impacts. If the estimated effect, which appears immediately in the first year and persists over
impact in Fig. 2 reflects the true PKH impact, rather than hidden biases in our time (see SM for more details). Thus, although the labor and health
design, we would expect that (i) the estimates from the flexible time trend (village responses of households in the pilot may have differed from the re-
linear time trend and province by year dummy variables), the matched design, and sponses of households in our sample, the evidence does not support
lagged cover loss design imply an impact similar to the estimated impact in Fig. 2, the hypothesis that PKH affected deforestation through condition-
and (ii) the estimate from the sensitivity test, which is a stress test that assesses
ality obligations.
how much our estimate would change if there were a powerful unobserved
In contrast, the evidence from our sample of 7468 villages more
confounder in our design, implies a deforestation-reducing effect of the PKH, albeit
smaller than the estimate in Fig. 2. Estimated percent reduction is calculated by
strongly supports two other channels: consumption smoothing, whereby
dividing the estimated reduction in forest cover loss by the (weighted) average cash substitutes for deforestation as a form of insurance for poor
annual forest cover loss in the villages that provide the estimate of counterfactual households, and consumption substitution, whereby market-purchased
loss. Black bars depict the estimated average effects. Whisker lines depict 95% CIs. goods substitute for deforestation-sourced consumption goods. As
SE estimates are clustered at the subdistrict level. in other nations, deforestation in Indonesia serves as an investment
in insurance (or a coping strategy) against weather shocks: when the assumptions about the persistence of PKH’s effects on deforestation,
rice season rains are delayed, poor farmers anticipate lower yields the economic value from reduced carbon emissions alone could cover
and deforest more land to compensate for the higher probability of the PKH implementation costs (Table 1).
a negative consumption shock. PKH is most effective at reducing Our study does not, however, imply that conservation funds should
deforestation in years when a village experiences negative rainfall be redirected to poverty alleviation. Drawing such a conclusion re-
shocks (table S14), a pattern consistent with transfers acting as a quires analyses, similar to Table 1, for a larger set of environmental
consumption smoothing mechanism. Moreover, PKH is most ef- and social values across a range of environmental and anti-poverty
fective at reducing deforestation in communities with better access interventions. Our study, with its emphasis on better design for
to markets (table S14), a moderating effect that is consistent with causal inference and on elucidating multiple mechanistic pathways,
recipients using PKH cash to substitute deforestation-sourced goods offers one exemplar for future research in that direction.
with market-purchased goods, but not consistent with changes in
demand for forest-intensive products or with changes in productive
investments. MATERIALS AND METHODS
Using the same data, we find weak evidence for the reciprocity To ensure sufficient detail to evaluate and replicate our results, we
channel: The likelihood of citizens voting in favor of an incumbent provide a detailed SM section. All data and code used in this study
candidate in the 2008–2012 period increases with the proportion of can be found at https://osf.io/uzaxg/. Here, we provide an outline of
a district’s villages that is exposed to PKH (table S15). That relation- the key aspects of Materials and Methods.
ship, however, does not necessarily imply that villagers were more
likely to obey government regulations. With regard to villagers Determination of PKH assignment across villages
obeying government regulations via the scrutiny channel, we do not (section S1)
have empirical evidence for or against the scrutiny channel. Never- The key task in any observational design is to uncover the treatment
sample consists of 15 provinces, distributed across Sumatera, Traditional designs and estimators (section S3)
Kalimantan, Bali, and West Nusa Tenggara Islands, representing We implement (table S6) (i) a “before-after” design, which uses
53% of Indonesia’s forests with 75% or more canopy cover in 2000 longitudinal data to infer the effect of PKH by detecting a shift
and accounting for over 80% of the forest cover loss between 2000 in deforestation between pre-PKH and post-PKH periods; (ii) a
and 2012 (fig. S1). After applying inclusion and exclusion criteria “with-without” design, which seeks to infer the effect of PKH by
for villages, we had a sample of 26,574 villages, of which 7468 villages looking at post-PKH cross-sectional differences in deforestation
received the PKH between 2008 and 2012, and 19,106 villages had between PKH and non-PKH villages; and (iii) a sophisticated
never received the PKH by the end of 2012. The number of villages version of the “before-after-control-impact” design, also called a
that are newly exposed to PKH each year is 1251 in 2008, 307 in DD design, which seeks to combine the designs of (i) and (ii) to in-
2009, 633 in 2010, 1624 in 2011, and 3653 in 2012. fer the effect of PKH by contrasting the trends in deforestation for
PKH and all non-PKH villages; we call this design the fixed-effects
Data (section S2c) panel data design using the full sample of villages. We implemented
Our outcome variable is forest cover loss, for which our measures these estimators using Stata 16 MP.
come from Hansen et al. (21). To avoid including losses from non-
native forests (e.g., rubber plantations), we defined “forests” as pixels Heterogeneity in PKH impact (section S4)
with tree cover canopy density greater than 75%, with a robustness To test for the heterogeneity of treatment effects, we interact the
check using a 30% threshold (table S4). To designate forest in our PKH variable with characteristics that are believed to be moderators
sample as “primary,” we use polygons of Indonesia’s primary forest of forest use (tables S7 and S8). Other tests of heterogeneity are used
in 2000 from Global Forest Watch (GFW), which are based on data in the mechanism analysis.
from Margono et al. (20). To calculate regrowth, we use Vegetation
Continuous Fields data from Song et al. (32). Our treatment vari- Rival explanation robustness checks (section S5)
intensity of exposure to neighboring PKH villages, we estimate the 14. A. Balmford, J. L. Moore, T. Brooks, N. Burgess, L. A. Hansen, P. Williams, C. Rahbek,
Conservation conflicts across africa. Science 291, 2616–2619 (2001).
localized spillover effects on forest cover loss (table S13).
15. S. Dasgupta, B. Laplante, H. Wang, D. Wheeler, Confronting the Environmental Kuznets
Curve. J. Econ. Perspect. 16, 147–168 (2002).
Mechanisms (section S7) 16. A. D. Foster, M. R. Rosenzweig, Economic growth and the rise of forests. Q. J. Econ. 118,
To shed light on mechanisms, we combine theory, the results from 601–637 (2003).
evaluations of randomized 2007 pilot PKH program (25), and addi- 17. D. S. Wilkie, M. Starkey, K. Abernethy, E. N. Effa, P. Telfer, R. Godoy, Role of prices
and wealth in consumer demand for bushmeat in Gabon, Central Africa. Conserv. Biol. 19,
tional analyses with our nonexperimental data. We explore the 268–274 (2005).
implications for each mechanism listed in Fig. 1, and we test those 18. J. Alix-Garcia, C. McIntosh, K. R. E. Sims, J. R. Welch, The ecological footprint of poverty
implications against our observational and experimental data (tables alleviation: Evidence from Mexico’s oportunidades program. Rev. Econ. Stat. 95, 417–435
S14 and S15). (2013).
19. N. L. Harris, E. Goldman, C. Gabris, J. Nordling, S. Minnemeyer, S. Ansari, M. Lippmann,
L. Bennett, M. Raad, M. Hansen, P. Potapov, Using spatial statistics to identify emerging
Benefit-cost analyses (section S8) hot spots of forest loss. Environ. Res. Lett. 12, 24012 (2017).
To calculate the metric tons of CO2 equivalent emissions per hectare 20. B. A. Margono, P. V. Potapov, S. Turubanova, F. Stolle, M. C. Hansen, Primary forest cover
in the PKH communities’ forests, we use GFW data. We estimate loss in Indonesia over 2000–2012. Nat. Clim. Chang. 4, 730–735 (2014).
the average aboveground biomass (AGBM) from provinces in our 21. M. C. Hansen, P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina,
D. Thau, S. V. Stehman, S. J. Goetz, T. R. Loveland, A. Kommareddy, A. Egorov, L. Chini,
sample with canopy cover greater than 30% in 2000 and then C. O. Justice, J. R. G. Townshend, High-resolution global maps of 21st-century forest cover
multiply that value by 0.5 to obtain carbon per ha, and then by 3.67 change. Science 342, 850–853 (2013).
to convert it to CO2 per ha. We conservatively estimate that 75% 22. Emissions Summary for Indonesia (UNCC Secretariat), https://unfccc.int/files/ghg_data/
of the AGBM is lost soon after clearing in the form of greenhouse ghg_data_unfccc/ghg_profiles/application/pdf/idn_ghg_profile.pdf.
23. N. Johnson, Why are we still losing so much rainforest in Indonesia? A hypothesis Grist
gas emissions, and the remaining 25% is permanently sequestered
(2016); https://grist.org/business-technology/why-are-we-still-losing-so-much-
in wood products. We use government and multilateral donor data
43. A. Diamond, J. S. Sekhon, Genetic matching for estimating causal effects: A general 55. T. Santika, E. Meijaard, S. Budiharta, E. A. Lawa, A. Kusworo, J. A. Hutabarat,
multivariate matching method for achieving balance in observational studies. T. P. Indrawan, M. Struebig, S. Raharjo, I. Huda, Sulhani, A. D. Ekaputri, S. Trison,
Rev. Econ. Stat. 95, 932–945 (2012). M. Stigner, K. A. Wilson, Community forest management in Indonesia: Avoided
44. P. J. Ferraro, J. J. Miranda, Panel data designs and estimators as substitutes deforestation in the context of anthropogenic and climate complexities.
for randomized controlled trials in the evaluation of public programs. J. Assoc. Environ. Glob. Environ. Chang. 46, 60–71 (2017).
Resour. Econ. 4, 281–317 (2017). 56. J. Labonne, The local electoral impacts of conditional cash transfers: Evidence from a field
45. S. Jayachandran, J. de Laat, E. F. Lambin, C. Y. Stanton, R. Audy, N. E. Thomas, Cash for experiment. J. Dev. Econ. 104, 73–88 (2013).
carbon: A randomized trial of payments for ecosystem services to reduce deforestation. 57. A. L. De La O, Do conditional cash transfers affect electoral behavior? evidence
Science 357, 267–273 (2017). from a randomized experiment in Mexico. Am. J. Pol. Sci. 57, 1–14 (2013).
46. P. R. Rosenbaum, Observational Studies (Springer, 2002). 58. M. Manacorda, E. Miguel, A. Vigorito, Government transfers and political support.
47. D. E. Ho, K. Imai, G. King, E. A. Stuart, Matching as nonparametric preprocessing Am. Econ. J. Appl. Econ. 3, 1–28 (2011).
for reducing model dependence in parametric causal inference. Polit. Anal. 15, 199–236 59. O. Attanasio, S. Polania-Reyes, L. Pellerano, Building social capital: Conditional cash
(2007). transfers and cooperation. J. Econ. Behav. Organ. 118, 22–39 (2015).
48. J. D. Angrist, J.-S. Pischke, Mostly Harmless Econometrics: An Empiricist’s Companion
(Princeton Univ. Press, 2009). Acknowledgments: P.J.F. and R.S. thank J. Alix-Garcia, A. Balmford, P. Kareiva, C. McIntosh,
49. T. J. VanderWeele, O. A. Arah, Bias formulas for sensitivity analysis of unmeasured P. Stephan, B. Meiselman, K. Sims, C. Weigel, and S. Wunder for helpful comments on earlier
confounding for general outcomes, treatments, and confounders. Epidemiology 22, versions and V. Yulaswati, BAPPENAS, for PKH data. Funding: R.S. gratefully acknowledges
42–52 (2011). financial support from the UNCE project (UNCE/HUM/035) and AYSPS Dissertation Fellowship.
50. P. H. Verburg, E. C. Ellis, A. Letourneau, A global assessment of market accessibility and Author contributions: P.J.F. and R.S. designed the study, interpreted the results, and wrote
market influence for global environmental change studies. Environ. Res. Lett. 6, 034019 the final manuscript. R.S. supervised data collection and data analysis. Competing interests:
(2011). The authors declare that they have no competing interests. Data and materials availability:
51. S. Bazzi, A. Gaduh, A. D. Rothenberg, M. Wong, Skill Transferability, Migration, and All data needed to evaluate the conclusions in the paper are present in the paper and/or the
Development: Evidence from Population Resettlement in Indonesia. Am. Econ. Rev. 106, Supplementary Materials. Additional data related to this paper may be requested from the
2568–2698 (2001). authors. All data and code used in the analysis are available at https://osf.io/uzaxg/.
52. S. K. Pattanayak, E. O. Sills, Do tropical forests provide natural insurance? Submitted 14 August 2019
Science Advances (ISSN 2375-2548) is published by the American Association for the Advancement of Science. 1200 New York Avenue
NW, Washington, DC 20005. The title Science Advances is a registered trademark of AAAS.
Copyright © 2020 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science. No claim
to original U.S. Government Works. Distributed under a Creative Commons Attribution NonCommercial License 4.0 (CC BY-NC).