Short Summary: This paper identifies productivity gains from trade by studying firms' manipulation behavior of in response to regulatory policies on international trade in China. Bunching estimates show that participation in international trade increases firm productivity. The productivity gains increase over time, indicating dynamic learning from trading. Further exploration shows no effects on R&D investment, product rationalization and markup. Young firms and nonstate-owned firms (non-SOEs) gain more from participating in trade.
Summary: This paper develops a causal inference methodology under the setting where agents face non-linear incentives at a policy kink and can adjust their assignment variable values around it, i.e., kinked bunching. While existing literature largely focuses on agents' assignment variable, we study the impact on other outcomes of interest due to the kinked policy and agents' manipulation. The estimator is reduced-form, and accounts for the interior responses under kinked bunching. Applying the proposed approach, we show how increase in co-payment rates lead to compressed outpatient medical treatment for urban employed and rural residents in China.
Summary: This study examines how the timing of benefits – immediate versus deferred – affects R&D investment. Using a difference-in-differences-in-differences (DDD) design, we find that immediate credits are over three times more effective in stimulating R&D than deferred credits, a gap amplified for liquidity-constrained firms. This result underscores that the timing of incentives can be more influential than the credit rate itself. Methodologically, we develop a practical test and correction for “selection on gains” in difference-in-differences with treatment intensity, validating our design and providing a reusable tool for empirical corporate finance. [Draft]
Summary: Ride-hailing platforms use rank-based dispatch, rewarding peak-period driving with lean-period priority. We study two consequences. Involution: ranking increases acceptance, driving hours and mileage, while lowering unit pay. Collusion: because rank is platform-specific, it becomes reputation capital that discourages multihoming, partitions the driver pool, and softens wage competition. We develop a dynamic model of rank-based dispatchto formally derive it. Using driver surveys, administrative transaction data, municipal monitoring reports, and a staggered quasi-experiment around a platform's ranking intensification, we find that ranking incentives raise driving mileage and hours, reduce hourly wages and revenue per kilometer, encourage acceptance under fatigue, and induce loyalty to a primary platform. Agent-based simulations show these effects are robust to bounded rationality. These findings support targeted regulation of rank-based incentives to address safety concerns and tacit labor-market collusion.
Summary: We study how payment protection reallocates financing within buyer-dominated supply chains. Using China's 2020 Regulation on Securing Payments to SMEs and its 2025 revision, we estimate DiD models on quarterly listed-firm data, comparing SMEs with near-threshold non-SME firms. Both reforms improve SMEs' receivables recovery. The 2020 regulation generates larger, broader deterioration in firm performance, while the 2025 revision produces smaller, less pervasive real-side pressure. Adverse effects concentrate among suppliers with high customer concentration and weak innovation-based differentiation. DiD causal forests and neural-network CATE estimates recover these heterogeneity dimensions. Payment protection improves SME suppliers' collection, but its real effects depend on buyer power, supplier differentiation, and the maturity of the enforcement regime. [Draft]
Summary: This paper uses matched order-level transactions and high-frequency GPS traces from all ride-hailing platforms in a large Chinese city. For each order arrival, we reconstruct the platform's local choice set by identifying eligible nearby drivers immediately before assignment, allowing us to compare the selected driver with rejected alternatives under identical order conditions. We find systematic evidence of "dispatch beyond proximity": conditional on pickup distance, and order fixed effects, drivers with higher prior activity are significantly more likely to receive orders, suggesting that platforms use allocation priority to manage labor supply. We then estimate the implied dispatch rule and evaluate counterfactual policies that place different weights on pickup proximity versus driver activity. These counterfactuals reveal a central operational trade-off: rewarding active drivers may strengthen supply-side incentives, but it increases passenger waiting, pickup distance, and driver deadheading relative to proximity-based and welfare-oriented benchmarks. The findings highlight order allocation as a powerful, nonprice lever in two-sided platform operations. [Draft]
Summary: We study whether the price become lower in response to the value-added tax cuts in China. We observe over-passthrough ain certain sectors and significant under-passthrough in other sectors. Market segmentation and product proliferation tend to be the underlying reasons. Further, we observe immediate response in some sectors and lagged response in other sectors, due to product cycle
Abstract: This paper explores a kinked tax policy in China that sets a wage deduction limit proportional to the firm-level average wage, creating incentives for firms to adjust their labor structure. We document significant bunching in the density distribution of firm-level average wages at the kink, as firms strategically reduce the ratio of skilled to unskilled labor to lower their taxable average wage. Based on a stylized employment model, we exploit this bunching behavior to infer the elasticity of substitution between skilled and unskilled labor—a key parameter in models utilizing efficient units of labor such as the CES production function. Our findings not only provide a novel estimator for this parameter but also shed light on the broader implications of tax policy on labor market decisions.
Abstract: On-the-job training is an important channel to human capital accumulation and economic growth. However, a lot of firms do not invest in such training in early 2000s China. Common concerns may involve employee turnover and their increased bargin power post training. In this study, we study the effect of employee training on firms' performance. We deal with endogeneity issue by exploitting a kinked policy, where firms' training expenses are deductible up to a fixed share of the labor cost. We find that a lot of firms adjust their training expenses upwards and bunch at the deduction limit. Using a novel causal estimator under kinked settings (Lu et al., 2023), we show that increase in training expenses lead to higher output, revenue and productivty. We also found evidence that some firms label administrative cost as training expenses to take adavantage of the tax deduction. Based on a stylized model of firms' training decisions and the causal estimates, we calibrate the parameters which governs the cost of relabelling and the productivity returns to training. Then, we conduct policy simulations and show that the tax deduction point should be raised from 2.5% to 8%. Different from China, most countries allow full deduction in training expenses but only for training related to skills in the current jobs. We show that under potential relabelling, a kinked tax deduction policy with a larger base is more efficient than full deduction with a narrower base.
Summary: This study proposes an alternative input–output based spatial structural decomposition analysis to elucidate the importance of domestic regional heterogeneity and inter-regional spillover effects in determining China's regional CO2 emissions growth.