Projects

Panel econometrics / China's economy

Bank credit and private-state investment

Provincial evidence from China's post-crisis credit expansion.

When bank lending grows, does state-owned-economy fixed-asset investment respond more strongly than private-enterprise investment?

The paper tests a prediction from the literature on China's financial system: state firms have better access to formal credit, while private firms rely more on internal funds. A 2007-2012 province-year panel stacks state and private investment as two observations per province-year.

The baseline difference is large, but the identification tests are uncomfortable. I report both.

01 / Model

One province, two ownership sectors

The interaction coefficient measures how much less, or more, private investment covaries with bank-loan growth than state investment in the same national period.

C is provincial bank-loan growth and G contains macro controls added sequentially. Province, year, and ownership fixed effects absorb persistent regional, national, and sector differences.

02 / Theory

Credit allocation and internal funds

Song, Storesletten, and Zilibotti motivate a private sector with productive firms but tighter formal-credit constraints. Allen, Qian, and Qian document private growth outside strong formal finance, while Cong and co-authors connect stimulus credit to state-linked firms.

If that mechanism appears in provincial aggregates, state investment should have a larger contemporaneous association with bank-loan growth. Private investment may instead move more with retained profits.

0.765State loan-growth sensitivity, baseline + GDP
-0.673Private differential, p = 0.0037
0.093Implied private sensitivity, p = 0.559
30Provinces in the 2007-2012 baseline

03 / Main estimates

The private-state gap survives controls

SpecificationState sensitivityPrivate differentialImplied privateN
Base0.730***-0.673***0.057360
+ GDP0.765***-0.673***0.093360
+ Deposits0.794***-0.673***0.122360
+ Fiscal expenditure0.747***-0.598***0.149336
+ Government size0.521*-0.598***-0.077336

The state coefficient remains positive as controls enter. The private interaction is stable and negative, leaving the implied private sensitivity small and statistically insignificant. Standard errors are clustered by province.

04 / Diagnostics

The association is not the identification

SpecificationPrivate differentialp-value
Baseline + GDP growth-0.6730.0037
Full sequential controls-0.5980.0064
Ownership-specific year FE-0.0020.996
Lead-loan placebo-0.783<0.001

The first rows use contemporaneous bank-loan growth. The placebo uses future loan growth. Its significance and the null ownership-year specification rule out a strong causal interpretation.

05 / Mechanism

Private investment follows profits more closely

In a separate mechanism regression, the private differential on lagged industrial-profit growth is +0.095 (p = 0.026). That is consistent with private investment depending more on internal funds than state investment does.

Taken together, the results fit a state-biased credit-investment pattern during the post-crisis expansion. They do not isolate an exogenous credit-supply shock, and the baseline pattern appears tied to the stimulus years.

06 / Tools

Tools and sources

PythonpandasNumPystatsmodelsOpenPyXLPanel fixed effectsCNKILaTeXBibTeXpytest

The public repository includes code, paper source, bibliography, and reviewed aggregate outputs. Source workbooks, observation-level panels, and copyrighted literature PDFs are not redistributed.