The Impact of Chinese Development Finance on FDI and Forest Cover
Evidence from Developing Asia, 2000 – 2017
June 2024 · Soumita Roy · 10 min read
Geneva Graduate Institute (IHEID)
In brief: China's financial footprint in developing Asia grew nearly threefold between 2000 and 2015, with commitments surging from under $100 billion to roughly $280 billion. This project asks a deceptively simple question: what did all that money actually do? Using a project-level dataset of 794 Chinese financing commitments across 31 Asian economies, I trace the effects on two outcomes that rarely sit in the same regression: foreign direct investment inflow and forest cover loss. The results point to a robust positive association between Chinese financing and FDI, driven primarily by commercially oriented flows into infrastructure and production. On deforestation, the evidence is more ambiguous: no statistically significant link emerges, suggesting the environmental story is more nuanced than the headlines might imply.
The question
Under the Belt and Road Initiative, China pledged over one trillion dollars in infrastructure investments spanning more than 60 countries. In Southeast Asia alone, Chinese finance accounted for roughly 24% of total development finance needs between 2009 and 2015. Yet the academic literature has overwhelmingly concentrated on Africa, leaving Asia's experience surprisingly underexplored.
I set out to fill that gap by investigating three questions. First, does Chinese development finance attract additional foreign direct investment from the rest of the world, or does it crowd it out? Second, does the accompanying infrastructure boom come at the cost of forest cover? And third, do these effects vary depending on whether the money flows into roads and power plants (commercially oriented "OOF-like" flows) or into schools and hospitals (concessional "ODA-like" aid)?
The motivation is not purely academic. Nearly half of all Chinese financing commitments in Asia went to physical infrastructure: transport corridors, energy generation, telecommunications. These are precisely the kinds of projects that can simultaneously attract investors and flatten forests. Understanding this trade-off matters for the 31 countries in the sample, from Pakistan and Myanmar to Cambodia and Laos, all of which sit at the intersection of China's strategic ambitions and their own development needs.
Two channels, opposite directions
The theoretical case is genuinely ambiguous, which is what makes the empirics interesting. On the FDI side, two competing mechanisms operate simultaneously. A catalytic channel works through improved infrastructure: better roads, ports and power grids lower operating costs, raise productivity and make a country more attractive to outside investors. Work by Clemens et al. (2012), Dreher and Langlotz (2020) and Dreher, Fuchs, Parks et al. (2021) documents versions of this effect across different settings.
Running in the opposite direction is a deterrence channel. Large-scale projects, particularly those with questionable economic returns, can pile up debt, squeeze government budgets and create the kind of macroeconomic instability that sends foreign investors elsewhere. Currency risk, fiscal overstretch and the spectre of "debt-trap diplomacy" all weigh on the other side of the ledger.
The forest cover story has a parallel structure. Infrastructure projects, above all roads and dams, open up previously inaccessible areas to logging and agricultural expansion. But the same projects can also raise agricultural productivity on existing land, reducing pressure to clear new forest. Electrification can displace wood fuel. The net effect is an empirical question.
Assembling the evidence
The backbone of the analysis is AidData's Global Chinese Development Finance Dataset (Version 2), which tracks 794 individual financing commitments to 31 developing Asian countries between 2000 and 2015. The dataset was compiled using the Tracking Underreported Financial Flows (TUFF) methodology, drawing on Chinese government sources, recipient-country finance ministries, local-language news reports and NGO case studies, all of which help to overcome the opacity of China's official reporting.
Each project carries a geoJSON file with precise coordinates. I imported these into ArcGIS to produce spatial joins linking each project to administrative boundaries, enabling country-year aggregation. The financing data distinguish three types: ODA-like flows (concessional, development-oriented), OOF-like flows (commercially driven, less concessional) and "Vague" flows that lack sufficient detail for classification. I also reclassified the standard OECD-DAC sector codes into seven broader categories tailored to the research questions.
FDI inward flow data come from UNCTAD's World Investment Report for 2002 to 2017, a window that introduces a two-year lag to capture the gestation period of financing effects. Forest cover loss is tracked using Global Forest Watch's 30-metre satellite imagery from the GLAD lab at the University of Maryland, supplemented with Python-based web scraping for additional metadata. I normalise forest loss against the FAO's baseline forest area in 2000 to make cross-country comparisons meaningful.
Several supporting datasets round out the picture. UNGA voting alignment data from the Harvard Dataverse proxy for the political likelihood of receiving Chinese finance. Historical aid-receipt probabilities, constructed from Dreher, Minasyan and Nunnenkamp (2015), capture aid inertia across 1970 to 1999. China's net foreign exchange reserves, from the World Bank's WDI, provide a time-varying supply-side instrument. Population and GDP controls come from UNCTAD.
Getting at causality (and the limits of doing so)
The starting point is an OLS panel regression with country and year fixed effects. The outcome variables include FDI per capita, FDI as a share of GDP, log FDI inflow, log forest cover loss (in hectares) and forest loss as a percentage of the 2000 baseline. The key explanatory variable is two-year lagged per capita Chinese official financing commitments, a lag chosen to capture the gestation period between financial commitment and observable impact. Lagged log population and log GDP serve as controls.
Where Yi,t is the outcome for country i in year t, OFi,t−2 is two-year lagged per capita Chinese official financing, and ηi, ηt are country and year fixed effects.
The core endogeneity concern is straightforward: China may direct financing towards countries that are already experiencing economic growth (or decline), creating reverse causality. To address this, I employ an instrumental variable strategy following Bluhm et al. (2020), Dreher, Fuchs, Parks et al. (2021) and Andersen, Johannesen and Rijkers (2022).
The instrument interacts two components. The first is the three-year lagged, detrended change in China's net foreign exchange reserves, which captures Beijing's time-varying capacity to fund overseas projects. The second is a country-specific measure of the likelihood of receiving Chinese finance, proxied either by UNGA voting alignment with China or by the historical probability of receiving Chinese aid (the share of years between 1970 and 1999 in which a country received at least one Chinese project). The intuition is that when China's reserves surge, the windfall disproportionately benefits countries that are politically aligned or historically connected.
OF̂i,t−2 = γ1(Reservest−3 × pCHN,i) + γ2 log(popi,t−1) + γ3 log(GDPi,t−1) + μi + μt + ui,t−2
(3) Second stage:
Yi,t = δ1 OF̂i,t−2 + δ2 log(popi,t−1) + δ3 log(GDPi,t−1) + λi + λt + υi,t
Where Reservest−3 is the detrended change in China's foreign exchange reserves, and pCHN,i proxies a country's likelihood of receiving Chinese aid. The interaction is the excluded instrument.
To unpack which types of financing drive the aggregate results, I run a multivariate OLS regression that replaces the single CDF variable with three sectoral components: physical infrastructure, social infrastructure and production/services. This decomposition cannot claim the causal credentials of the IV approach, but it provides suggestive evidence on how sectoral allocation patterns shape outcomes.
What the numbers say: FDI
The OLS results paint a consistent picture. For every additional million dollars per capita of Chinese official financing committed two years earlier, FDI per capita rises by $0.057 million, statistically significant at the 1% level. FDI as a share of GDP increases by 0.001 percentage points (significant at 5%), and the log of FDI inflow also shows a positive, statistically significant coefficient. These magnitudes are in line with Dreher, Fuchs, Parks et al. (2021) and Clements and Milner-Gulland (2015), who document similar catalytic effects in other contexts.
Digging into the type of financing reveals a striking asymmetry. Other Official Flows (the commercially oriented category) drive essentially the entire effect. Each additional million dollars per capita of OOF raises FDI per capita by $0.058 million at the 1% level. ODA-like flows, by contrast, show no significant impact on any FDI metric. This is not entirely surprising: OOF tends to fund economic infrastructure and commercially viable projects that directly improve the investment climate, whereas ODA flows into social sectors like health and education, which operate on a longer and less investor-visible horizon.
The IV results are directionally consistent but tell a more cautious story. Using UNGA voting alignment as the demand-side proxy, the coefficient on FDI per capita rises to 0.28 (significant at 10%), and log FDI inflow shows a significant effect. The IV estimates being larger than OLS is textbook behaviour when OLS is biased downward by endogeneity. But here is the catch: the first-stage F-statistics hover around 3.4 to 3.6, well below the conventional threshold of 10. With the historical-probability proxy, F-statistics drop further to around 1.9, and none of the second-stage coefficients retain significance.
This is an honest limitation. The OLS evidence for a positive association is robust and consistent across specifications, but the IV approach cannot confirm causality. Stronger instruments, perhaps exploiting quasi-natural experiments arising from policy changes or exogenous shocks to Chinese lending capacity, would be needed to close the gap.
What the numbers say: forest cover
The environmental results are less dramatic but no less interesting for that. Neither OLS nor IV estimation produces a statistically significant effect of Chinese financing on forest cover loss, whether measured in absolute hectares or as a percentage of the 2000 baseline. The OLS point estimate is slightly negative (suggesting, if anything, a marginal reduction in forest loss), while the IV estimate is slightly positive. Both are comfortably within the bounds of statistical noise.
This null result does not mean that Chinese-funded projects have no environmental footprint. It may instead reflect the aggregation problem: at the country-year level, localised deforestation around a road or dam project could be offset by conservation gains elsewhere, or by improvements in agricultural productivity that reduce the need to clear forest. The environmental impact is likely heterogeneous, context-dependent and difficult to capture at this level of analysis.
Unpacking by sector
The sectoral decomposition adds useful texture. Physical infrastructure financing significantly raises FDI as a share of GDP (by 0.0044 percentage points for each additional million dollars per capita, significant at the 1% level), while production and services financing significantly boosts FDI per capita (by $0.053 million, also at the 1% level). Social infrastructure, despite a large point estimate, is not statistically significant.
These results make intuitive sense. Roads, ports and power plants lower the cost of doing business and signal government commitment to development, both of which matter to foreign investors. Production-sector investments in mining, agriculture and construction create direct economic activity that attracts further capital. Social investments, while valuable in their own right, operate on a timeline and through channels that are less immediately visible to the FDI calculus.
What it all means
Chinese development finance is positively associated with FDI inflows across Asian developing countries. The effect is robust across OLS specifications and directionally confirmed by IV estimates, though causal claims are limited by weak instruments.
The FDI effect is driven almost entirely by commercially oriented "OOF-like" flows. Concessional "ODA-like" financing shows no significant relationship with FDI, suggesting that the investment-attraction channel operates through economic infrastructure rather than social development aid.
Physical infrastructure and production-sector investments are the key drivers. Infrastructure raises FDI relative to GDP; production and services raise it in absolute per-capita terms. Social infrastructure shows no discernible FDI impact.
No statistically significant link between Chinese financing and deforestation is observed at the country-year level. The null result likely reflects aggregation masking localised effects rather than a genuine absence of environmental impact.
Looking ahead
For the 31 countries in the sample, the practical takeaway is relatively clear: Chinese financing, when directed at economic infrastructure and production, is associated with a more attractive investment climate. This is not a trivial result for governments trying to balance competing demands on limited fiscal space. But the absence of clear environmental effects at the aggregate level should not be read as an all-clear. It is more likely a limitation of the data and the level of analysis.
Future work could exploit more granular spatial data, linking individual project locations to localised forest cover change rather than relying on country-year aggregates. Quasi-natural experiments arising from sudden shifts in Chinese lending policy or political realignments would also help sharpen the causal story. The weak-instruments problem is a genuine constraint, and overcoming it will likely require creative identification strategies that go beyond the standard reserves-times-alignment interaction.
The broader lesson is that development finance is not monolithic. Its effects depend on what it funds, how it is structured and where it lands. Treating Chinese financing as a single phenomenon risks missing the variation that matters most for policy.