Spatial and temporal patterns of land clearing during policy change

Effectiveness of regulatory policy in curbing deforestation in a biodiversity hotspot

“…The ATTs for this pre-intervention period revealed remnant deforestation rates significantly higher than the counterfactual (ATT 1999,FO =4.46%, ATT 1999,FL =3.29%). Two explanations for this preferential clearing of remnant trees compared to similar non-remnant trees in this period are: (1) that this period shows 'pre-emptive' clearing of remnant trees prior to the Act, evidence of which has been reported in previous analyses (Simmons et al 2018a(Simmons et al , 2018b and landholder testimonials (Productivity Commission 2004, Senate Inquiry 2010), or (2) that this is a social preference for clearing of remnant trees versus similar areas of non-remnant trees. We developed two assumption scenarios from these: (1) Pre-emptive Clearing scenario (ATT PC ), assuming that elevated pre-emptive clearing of remnant trees is caused entirely by the Act and that the true difference between our treated and counterfactual samples is zero (ATT 1999 =0%), and (2) Social Preference scenario (ATT SP ), assuming that there is a fixed social preference for clearing remnant over equivalent nonremnant, which is equal to a constant deforestation rate of SP FO =4.46% for 2000-2004 and SP FL = 3.29% for 2005-2016.…”

“…It is likely that this drought was largely responsible for the reduction of absolute deforestation rates during this period by diminishing the economic incentives to clear, which could explain why the VMA was only marginally effective in those years. Characteristics of landholders' clearing patterns during this period have reflected a preference for maximizing quality over quantity (Simmons et al 2018a), and this may explain the subsequent increase in the quantity of non-remnant deforestation following the end of the drought. These climate constraints could also have flow-on effects on other socio-economic drivers of deforestation, such as food prices and potential profitability, which can ultimately diminish the direct impact of the VMA (Marcos-Martinez et al 2017, Rhodes et al 2017, Simmons et al 2018b.…”

“…clearing for fences, weed control, necessary infrastructure) will be present in the deforestation data. This represents a simplification of the temporal and thematic coverage of the policy, which has undergone fluctuations in the level of restrictions placed on deforestation (Simmons et al 2018a), including provisions for 'high-value regrowth', and for remnant vegetation on leasehold lands. High-value regrowthregrowth that had not been cleared since 31 December 1989-also experienced some deforestation regulation during 2009-2013 from an amendment to the policy, but we excluded these trees from our definition of treatment status since they are not remnant trees.…”

“…For most of the high-impact years, the sensitivity analysis suggests that any potential confounding factor that was excluded from this study would need to exert an influence on the occurrence of remnant trees more than twice that of the other covariates in order to affect the impact estimates (Keele 2010). The years most sensitive to hidden bias coincide with the early years of the VMA timeline, where landholders were subjected to frequent policy uncertainty arising from multiple amendments, a moratorium on clearing permits, withdrawal of Federal Government support for a $150 million structural adjustment package, and implementation of a broad-scale clearing ban (Simmons et al 2018a). Frequent periods of policy uncertainty have been shown to increase deforestation in Queensland (Simmons et al 2018b), which could explain the sensitivity of impact estimates during this period.…”

“…This study provides the first robust quantitative analysis of the policy's ability to reduce deforestation of remnant trees, amidst ongoing intense debate surrounding the policy's effectiveness and concerns over its significant costs to local farmers and graziers (Senate Inquiry 2010, Reside et al 2017, Simmons et al 2018a. We illustrate a new application of causal inference techniques to a broad-reaching environmental statute over a temporally dynamic period, using covariate matching to determine how the policy has affected rates of remnant deforestation in the Brigalow Belt South (BBS) bioregion of Queensland, Australia, compared to two counterfactual scenarios-pre-emptive clearing (PC) and social preference (SP) scenariosrepresenting the lower and upper impact estimates, respectively, over time.…”