## wp17273

## Source details

**Canonical URL:** [wp17273](https://www.imf.org/-/media/files/publications/wp/2017/wp17273.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2017/wp17273.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2017/wp17273.pdf.json)

---

### I. Introduction — context and objectives
- China average growth since 2001: 13.5 percent.
- Total social financing (TSF) more than doubled since 2001, reaching about 200 percent of GDP in 2016.
- Bank credit and non-financial corporate debt: 145 percent of GDP.
- Policy trade-off highlighted:
  - Rapid credit expansion raises financial-stability risks.
  - Reducing credit growth could drag on activity; fiscal stimulus may be required to meet growth targets.
- Paper objective: joint estimation of credit and fiscal multipliers in China to assess macroeconomic implications of a policy rebalancing from credit-led to fiscal-led stimulus.

### II. Identification strategy and conceptual framework
- Need to jointly estimate credit and fiscal multipliers because China actively uses credit growth as a policy tool and sets policy-driven credit targets.
- Instrumental strategy:
  - Use tenure of provincial party secretaries as source of exogenous variation in stimulus policies.
  - Interact provincial secretary tenure with credit and expenditure growth in all other provinces to construct two instruments (one for credit, one for expenditure).
  - Justification: timing of appointments is unrelated to local economic conditions; incentives to provide stimulus vary over tenure and peak at specific tenure moments.
- Potential instrument concerns addressed:
  - Aggregate macro shocks absorbed by year fixed effects.
  - Spillovers from other provinces could bias estimates downward; acknowledged explicitly.

### III. Data and empirical method
- Sample:
  - 31 provincial units (4 centrally administrated cities, 22 provinces, 5 autonomous regions).
  - Sample period: 2001-2015.
  - Main data source: CEIC China Premium Database.
- Key variable definitions:
  - Provincial fiscal stance: total expenditure.
  - Provincial credit: sum of bank loans in domestic and foreign currencies granted in the province.
  - Provincial GDP deflated to 2010 RMB using national CPI.
- Bank loans as proxy for aggregate credit:
  - Correlation between bank loans and TSF growth rates (as share of GDP): 0.93 (or 0.84 if the 2009 outlier year is excluded).
- Political-cycle data:
  - Manually collected provincial party secretary appointment terms and profiles.
- Econometric specification:
  - Two-year changes for Y, G, and CR used to capture dynamics and sluggish output response.
  - Province and year fixed effects (αi and γt).
  - Standard errors clustered at the province level.
  - Shocks to government expenditure and credit normalized by initial provincial GDP.
  - 2SLS first stage: regress provincial credit and expenditure growth on other provinces’ credit and expenditure growth interacted with provincial party secretary tenure.

### IV. Main empirical results — aggregate multipliers and time variation
- Aggregate multipliers (2001-2015):
  - Fiscal multiplier: 0.8
  - Credit multiplier: 0.2
- Headline IV (joint) estimate for 2001-2015:
  - Fiscal multiplier: 0.8.
  - Credit multiplier: 0.2.
- Comparison across specifications:
  - Multipliers estimated separately are larger than jointly estimated multipliers, implying credit and fiscal stimuli are used as complements and separate estimates suffer upward omission bias.
  - OLS estimates are somewhat lower than IV estimates, reflecting countercyclical fiscal and credit policies that bias OLS downward.
- Multipliers over subperiods (interpreted as “normal times”, excluding the GFC period):
  - Fiscal multiplier increased from 0.75 in 2001-2008 to 1.4 in 2010-2015.
  - Credit multiplier declined from 0.17 in 2001-2008 to zero (statistically insignificant) in 2010-2015.
- Additional reported regression coefficients (selected):
  - Table 3 (Full sample; two-year growth specifications):
    - Real Credit (OLS): 0.221*** [0.035]
    - Real Credit (IV): 0.283*** [0.039]
    - Real Expenditure (OLS): 1.009*** [0.186]
    - Real Expenditure (IV): 1.083*** [0.290]
    - Observations: 370 (varies by column); example R-squared values: 0.770, 0.760, 0.803
    - Instrument strength examples: Cragg-Donald Wald F: 105, 112.2, 51.65; Kleibergen-Paap rk Wald F: 27.10, 23.36, 14.75
  - Table 4 (Early and late subsamples):
    - 2001-2008:
      - Real Credit (OLS): 0.210*** [0.062]
      - Real Credit (IV): 0.172*** [0.064]
      - Real Expenditure (OLS): 0.696*** [0.196]
      - Real Expenditure (IV): 0.742*** [0.277]
      - Observations: 166 (OLS), 165 (IV)
    - 2010-2015:
      - Real Credit (OLS): 0.019 [0.168]
      - Real Credit (IV): 0.005 [0.095]
      - Real Expenditure (OLS): 0.878** [0.320]
      - Real Expenditure (IV): 1.395*** [0.272]
      - Observations: 110 (OLS and IV)
  - Table 5 (Weighted results):
    - Full sample:
      - Real Credit (OLS): 0.185*** [0.033]
      - Real Credit (IV): 0.180*** [0.036]
      - Real Expenditure (OLS): 1.050*** [0.269]
      - Real Expenditure (IV): 0.760** [0.301]
      - Observations: 359 (OLS and IV)
- Quoted additional statistic repeated in source:
  - 1.682 for expenditure, which are significant at the10 percent and 5 percent respectively.
  - 1.682 for expenditure, which are significant at the10 percent and 5 percent respectively.

### V. Sectoral effects and industry composition
- General pattern:
  - Both credit and fiscal stimuli boost construction and manufacturing, but not services.
  - Fiscal stimulus historically targeted infrastructure and related industries; rebalancing toward services would require fiscal rebalancing (for example, toward social expenditure).
- Panel C / Table 6 (sectoral contributions and constructed effects on industry growth):
  - Real Credit (2001-2015)
    - Sectoral multiplier contributions: Construction 0.015; Manufacturing 0.11; Services ex financial 0.046
    - Divided by sectoral share in GDP: Construction 0.075; Manufacturing 0.39; Services ex financial 0.37
    - Obtained effect on industry growth: Construction 0.20; Manufacturing 0.28; Services ex financial 0.12
  - Real Credit (2010-2015)
    - Sectoral multiplier contributions: Construction 0.003; Manufacturing 0.07; Services ex financial -0.098
    - Divided by sectoral share in GDP: Construction 0.078; Manufacturing 0.39; Services ex financial 0.37
    - Obtained effect on industry growth: Construction 0.04; Manufacturing 0.18; Services ex financial -0.27
  - Real Expenditure (2001-2015)
    - Sectoral multiplier contributions: Construction 0.055; Manufacturing 0.606; Services ex financial 0.138
    - Divided by sectoral share in GDP: Construction 0.075; Manufacturing 0.39; Services ex financial 0.37
    - Obtained effect on industry growth: Construction 0.73; Manufacturing 1.57; Services ex financial 0.37
  - Real Expenditure (2010-2015)
    - Sectoral multiplier contributions: Construction 0.073; Manufacturing 0.678; Services ex financial 0.219
    - Divided by sectoral share in GDP: Construction 0.078; Manufacturing 0.39; Services ex financial 0.37
    - Obtained effect on industry growth: Construction 0.94; Manufacturing 1.73; Services ex financial 0.59
- Table 6 (Panel A and B; sectoral regressions, selected coefficients):
  - Panel A (2001-2015; sectoral contributions):
    - Construction: Real Credit (IV) 0.015** [0.007]; Real Expenditure (IV) 0.055* [0.033]
    - Manufacturing: Real Credit (IV) 0.109** [0.044]; Real Expenditure (IV) 0.606*** [0.220]
    - Services ex financial: Real Credit (IV) 0.046** [0.022]; Real Expenditure (IV) 0.138 [0.099]
    - Observations: Construction 364; Manufacturing 360; Services ex financial 358
    - Instrument diagnostics examples: Cragg-Donald Wald F: 42.29, 42.48, 42.68; Kleibergen-Paap rk Wald F: 14.67, 13.88, 12.69
  - Panel B (2010-2015; sectoral contributions):
    - Construction: Real Credit (IV) 0.003 [0.011]; Real Expenditure (IV) 0.073** [0.028]
    - Manufacturing: Real Credit (IV) 0.070 [0.094]; Real Expenditure (IV) 0.678*** [0.238]
    - Services ex financial: Real Credit (IV) -0.098** [0.041]; Real Expenditure (IV) 0.219** [0.090]
    - Instrument diagnostics examples: Cragg-Donald Wald F: 13.45, 13.08, 11.99; corresponding Kleibergen-Paap rk Wald F values reported

### VI. Robustness, diagnostics, and limitations
- Instrument validity and strength:
  - Probited appointment timing: probability of appointment is not associated with contemporaneous or 1- or 2-year lagged GDP growth, per capita GDP growth, or employment growth.
  - Cragg-Donald Wald F-test for excluded instruments comfortably exceeds Stock and Yogo (2005) critical values in many reported tables; Kleibergen-Paap rk Wald F reported in tables (examples: 27.10, 23.36, 14.75).
  - Instrument weakness noted for the service sector in the post-crisis period: Cragg-Donald Wald F statistic is 11.99, compared to the Stock and Yogo (2005) 5 percent IV relative size value of 13.97, suggesting IV estimate for services may be somewhat biased towards OLS estimates.
- Temporal pattern of stimulus relative to tenure:
  - Provincial credit and expenditure growth peak toward the end of the first five-year term and again in the middle of the second five-year term of provincial party secretaries.
- Robustness checks:
  - GDP-weighted regressions produce similar credit-multiplier results, including zero credit multiplier in 2010-2015.
  - Weighted fiscal multipliers are somewhat larger, suggesting fiscal stimulus may be more effective in more developed provinces.
- Limitations and potential biases:
  - Potential downward bias from interprovincial demand spillovers.
  - Nonlinear financial-stability responses to abrupt credit reversals not modeled; abrupt reversal in credit growth may affect financial stability via nonlinear responses not captured here.
  - “Open economy multiplier” approach absorbs aggregate endogenous macroeconomic responses (e.g., monetary policy reactions) into year fixed effects; such endogenous responses could make aggregate multipliers smaller than estimated provincial relative multipliers.
  - Nonlinearities (e.g., simultaneous stimulus across provinces) could imply aggregate multipliers are larger than estimates.
  - Multipliers may differ if underlying output growth rates substantially diverge from historic averages (e.g., in an economic slowdown).

### VII. Policy implications and recommendations
- Macro policy stance:
  - Slower credit growth is unlikely to disrupt output growth given the recent effective zero credit multiplier.
  - Fiscal policy may be effective in supporting macroeconomic adjustment given the recent fiscal multiplier of 1.4.
- Rebalancing toward services:
  - Historical fiscal stimulus has primarily boosted manufacturing more than services.
  - To contribute to rebalancing the Chinese economy toward services, new fiscal stimulus should be designed to target services more than historical fiscal stimulus did (e.g., services-related expenditure such as health, education).
- Design considerations:
  - Consider the differing expenditure responsibilities of central versus sub-national governments when designing stimulus to achieve aggregate objectives.
  - Account for potential monetary policy responses and nonlinear cross-province interactions when scaling up provincial estimates to aggregate policy design.
- Caveats for policymakers:
  - The fiscal multiplier estimated is an on-balance-sheet expenditure multiplier; off-balance-sheet spending (e.g., via local government funding vehicles financed with bank loans) likely has a lower multiplier.
  - Be cautious of nonlinear financial-stability risks from abrupt credit reversals not captured by the linear estimation.

*Source: IMF Working Paper wp17273 (excerpts and tables as provided).*

### References .............................................................................................................

### wp17273 - References

### I. Introduction — context and objectives
- China average growth since 2001: 13.5 percent.
- Total social financing (TSF) more than doubled since 2001, reaching about 200 percent of GDP in 2016.
- Bank credit and non-financial corporate debt: 145 percent of GDP.
- Policy trade-off highlighted:
  - Rapid credit expansion raises financial-stability risks.
  - Reducing credit growth could drag on activity; fiscal stimulus may be required to meet growth targets.
- Paper objective: joint estimation of credit and fiscal multipliers in China to assess macroeconomic implications of a policy rebalancing from credit-led to fiscal-led stimulus.

### II. Identification strategy and conceptual framework
- Need to jointly estimate credit and fiscal multipliers because China actively uses credit growth as a policy tool and sets policy-driven credit targets.
- Instrumental strategy:
  - Use tenure of provincial party secretaries as source of exogenous variation in stimulus policies.
  - Interact provincial secretary tenure with credit and expenditure growth in all other provinces to construct two instruments (one for credit, one for expenditure).
  - Justification: timing of appointments is unrelated to local economic conditions; incentives to provide stimulus vary over tenure and peak at specific tenure moments.
- Potential instrument concerns addressed:
  - Aggregate macro shocks absorbed by year fixed effects.
  - Spillovers from other provinces could bias estimates downward; acknowledged explicitly.

### III. Data and empirical method
- Sample:
  - 31 provincial units (4 centrally administrated cities, 22 provinces, 5 autonomous regions).
  - Sample period: 2001-2015.
  - Main data source: CEIC China Premium Database.
- Key variable definitions:
  - Provincial fiscal stance: total expenditure.
  - Provincial credit: sum of bank loans in domestic and foreign currencies granted in the province.
  - Provincial GDP deflated to 2010 RMB using national CPI.
- Bank loans as proxy for aggregate credit:
  - Correlation between bank loans and TSF growth rates (as share of GDP): 0.93 (or 0.84 if the 2009 outlier year is excluded).
- Political-cycle data:
  - Manually collected provincial party secretary appointment terms and profiles from specified Chinese politics and vitae websites.
- Econometric specification:
  - Two-year changes for Y, G, and CR used to capture dynamics and sluggish output response.
  - Province and year fixed effects (αi and γt).
  - Standard errors clustered at the province level.
  - Shocks to government expenditure and credit normalized by initial provincial GDP.
  - 2SLS first stage regresses provincial credit and expenditure growth on other provinces’ credit and expenditure growth interacted with provincial party secretary tenure.

### IV. Main empirical results
- Headline IV (joint) estimate for 2001-2015:
  - Fiscal multiplier: 0.8.
  - Credit multiplier: 0.2.
- Comparison across specifications:
  - Multipliers estimated separately are larger than jointly estimated multipliers, implying credit and fiscal stimuli are used as complements and separate estimates suffer upward omission bias.
  - OLS estimates are somewhat lower than IV estimates, reflecting countercyclical fiscal and credit policies that bias OLS downward.
- Multipliers over subperiods (interpreted as “normal times”, excluding the GFC period):
  - Fiscal multiplier increased from 0.75 in 2001-2008 to 1.4 in 2010-2015.
  - Credit multiplier declined from 0.17 in 2001-2008 to zero (statistically insignificant) in 2010-2015.
- Policy interpretation:
  - A reduction in credit growth is unlikely to disrupt output growth at present given the near-zero credit multiplier.
  - Fiscal policy may be effective in supporting macroeconomic adjustment, given the increased fiscal multiplier.
  - Caveat: abrupt reversal in credit growth may affect financial stability via nonlinear responses not captured by this analysis.
- Robustness:
  - GDP-weighted regressions (giving prominence to larger provinces) produce similar credit-multiplier results, including zero credit multiplier in 2010-2015.
  - Weighted fiscal multipliers are somewhat larger, suggesting fiscal stimulus may be more effective in more developed provinces.
- Industry composition effects:
  - Both credit and fiscal stimuli boost construction and manufacturing, but not services.
  - Fiscal stimulus historically targeted infrastructure and related industries; rebalancing toward services would require fiscal rebalancing (for example, toward social expenditure).

### V. Additional methodological and interpretative notes
- The fiscal multiplier estimated is an on-balance-sheet expenditure multiplier; off-balance-sheet spending (e.g., via local government funding vehicles financed with bank loans) likely has a lower multiplier.
- First-stage validity:
  - Appointment timing of provincial party secretaries tested via probit; probability of appointment is not associated with contemporaneous or 1- or 2-year lagged GDP growth, per capita GDP growth, or employment growth.
- Instrument strength:
  - Cragg-Donald Wald F-test for excluded instruments comfortably exceeds Stock and Yogo (2005) critical values (reported in study tables).
- Temporal pattern of stimulus relative to tenure:
  - Provincial credit and expenditure growth peak toward the end of the first five-year term and again in the middle of the second five-year term of provincial party secretaries.
- Limitations acknowledged:
  - Potential downward bias from interprovincial demand spillovers.
  - Nonlinear financial-stability responses to abrupt credit reversals not modeled.

*Italic source: wp17273 - References (PDF) — content extracted from the document provided.*

### 1.682 for expenditure, which are significant at the10 percent and 5 percent respectively.

### wp17273 - 1.682 for expenditure, which are significant at the10 percent and 5 percent respectively.

### Key empirical findings
- Aggregate multipliers (2001-2015):
  - Fiscal multiplier: 0.8
  - Credit multiplier: 0.2
- Time variation:
  - Credit multiplier has recently been effectively zero.
  - Fiscal multiplier has recently increased to 1.4.
- Sectoral effects (Panel C, inferred effects on industry growth; sectoral shares preserved as reported):
  - Real Credit (2001-2015)
    - Sectoral multiplier (contribution to overall output multiplier): Construction 0.015; Manufacturing 0.11; Services ex financial 0.046
    - Divided by sectoral share in GDP: Construction 0.075; Manufacturing 0.39; Services ex financial 0.37
    - Obtained effect on industry growth: Construction 0.20; Manufacturing 0.28; Services ex financial 0.12
  - Real Credit (2010-2015)
    - Sectoral multiplier: Construction 0.003; Manufacturing 0.07; Services ex financial -0.098
    - Divided by sectoral share in GDP: Construction 0.078; Manufacturing 0.39; Services ex financial 0.37
    - Obtained effect on industry growth: Construction 0.04; Manufacturing 0.18; Services ex financial -0.27
  - Real Expenditure (2001-2015)
    - Sectoral multiplier: Construction 0.055; Manufacturing 0.606; Services ex financial 0.138
    - Divided by sectoral share in GDP: Construction 0.075; Manufacturing 0.39; Services ex financial 0.37
    - Obtained effect on industry growth: Construction 0.73; Manufacturing 1.57; Services ex financial 0.37
  - Real Expenditure (2010-2015)
    - Sectoral multiplier: Construction 0.073; Manufacturing 0.678; Services ex financial 0.219
    - Divided by sectoral share in GDP: Construction 0.078; Manufacturing 0.39; Services ex financial 0.37
    - Obtained effect on industry growth: Construction 0.94; Manufacturing 1.73; Services ex financial 0.59

### Representative regression coefficients (selected tables)
- Table 3 (Full sample; two-year growth specifications):
  - Real Credit (OLS): 0.221*** [0.035]
  - Real Credit (IV): 0.283*** [0.039]
  - Real Expenditure (OLS): 1.009*** [0.186]
  - Real Expenditure (IV): 1.083*** [0.290]
  - Observations: 370 (varies by column); R-squared reported (e.g., 0.770, 0.760, 0.803)
  - Instrument strength: Cragg-Donald Wald F reported (e.g., 105, 112.2, 51.65); Kleibergen-Paap rk Wald F (e.g., 27.10, 23.36, 14.75)
- Table 4 (Early and late subsamples):
  - 2001-2008:
    - Real Credit (OLS): 0.210*** [0.062]
    - Real Credit (IV): 0.172*** [0.064]
    - Real Expenditure (OLS): 0.696*** [0.196]
    - Real Expenditure (IV): 0.742*** [0.277]
    - Observations: 166 (OLS), 165 (IV)
  - 2010-2015:
    - Real Credit (OLS): 0.019 [0.168]
    - Real Credit (IV): 0.005 [0.095]
    - Real Expenditure (OLS): 0.878** [0.320]
    - Real Expenditure (IV): 1.395*** [0.272]
    - Observations: 110 (OLS and IV)
  - Instrument diagnostics: Cragg-Donald Wald F values reported (e.g., 22.67, 15.68); Kleibergen-Paap rk Wald F (e.g., 14.53, 8.576)
- Table 5 (Weighted results):
  - Full sample:
    - Real Credit (OLS): 0.185*** [0.033]
    - Real Credit (IV): 0.180*** [0.036]
    - Real Expenditure (OLS): 1.050*** [0.269]
    - Real Expenditure (IV): 0.760** [0.301]
    - Observations: 359 (OLS and IV)
  - 2001-2008 and 2010-2015 weighted columns include analogous estimates and reported Kleibergen-Paap rk Wald F statistics (e.g., 24.92, 15.56, 8.120)

### Sectoral multipliers (Table 6, Panel A and B; sectoral contributions)
- Panel A (2001-2015; sectoral contributions):
  - Construction
    - Real Credit (OLS): 0.021*** [0.006]; (IV) 0.015** [0.007]
    - Real Expenditure (OLS): 0.084** [0.039]; (IV) 0.055* [0.033]
  - Manufacturing
    - Real Credit (OLS): 0.083** [0.033]; (IV) 0.109** [0.044]
    - Real Expenditure (OLS): 0.438*** [0.135]; (IV) 0.606*** [0.220]
  - Services ex financial
    - Real Credit (OLS): 0.062*** [0.018]; (IV) 0.046** [0.022]
    - Real Expenditure (OLS): 0.212* [0.109]; (IV) 0.138 [0.099]
  - Observations: Construction 364; Manufacturing 360; Services ex financial 358
  - Instrument diagnostics: Cragg-Donald Wald F (e.g., 42.29, 42.48, 42.68); Kleibergen-Paap rk Wald F (e.g., 14.67, 13.88, 12.69)
- Panel B (2010-2015; sectoral contributions):
  - Construction
    - Real Credit (OLS): 0.009 [0.011]; (IV) 0.003 [0.011]
    - Real Expenditure (OLS): 0.032 [0.032]; (IV) 0.073** [0.028]
  - Manufacturing
    - Real Credit (OLS): 0.079 [0.127]; (IV) 0.070 [0.094]
    - Real Expenditure (OLS): 0.361 [0.259]; (IV) 0.678*** [0.238]
  - Services ex financial
    - Real Credit (OLS): -0.067 [0.050]; (IV) -0.098** [0.041]
    - Real Expenditure (OLS): 0.133 [0.139]; (IV) 0.219** [0.090]
  - Observations and diagnostics reproduced as in Panel A (Cragg-Donald Wald F: 13.45, 13.08, 11.99; corresponding Kleibergen-Paap rk Wald F values)

### Identification, instrumentation, and robustness notes
- Instrumental strategy:
  - Instruments are based on provincial political cycles interacted with credit and expenditure growth in other provinces; intended to capture exogenous variation in provincial leaders’ incentives and the credit-vs-expenditure mix.
- Representativeness:
  - Provincial governments in China are primarily responsible for infrastructure, urban development, environment, water supply, community services, and in practice most spending in shared categories (education, health care, social welfare, public safety, local economic development) is done by sub-national governments (Shen et al, 2012). This distribution suggests province-level fiscal multipliers may be representative of aggregate fiscal multipliers.
- Limitations and potential biases:
  - “Open economy multiplier” approach absorbs aggregate endogenous macroeconomic responses (e.g., monetary policy reactions) into year fixed effects; such endogenous responses could make aggregate multipliers smaller than estimated provincial relative multipliers.
  - Nonlinearities (e.g., simultaneous stimulus across provinces) could imply aggregate multipliers are larger than estimates.
  - Multipliers may differ if underlying output growth rates substantially diverge from historic averages (e.g., in an economic slowdown).
  - Instrument weakness for the service sector in the post-crisis period: Cragg-Donald Wald F statistic is 11.99, compared to the Stock and Yogo (2005) 5 percent IV relative size value of 13.97, suggesting IV estimate for services may be somewhat biased towards OLS estimates (Bound, Jaeger, and Baker 1995).

### Policy implications and recommendations
- Macro policy stance:
  - Slower credit growth is unlikely to disrupt output growth given the recent effective zero credit multiplier.
  - Fiscal policy may be effective in supporting macroeconomic adjustment given the recent fiscal multiplier of 1.4.
- Rebalancing toward services:
  - Historical fiscal stimulus has primarily boosted manufacturing more than services.
  - To contribute to rebalancing the Chinese economy toward services, new fiscal stimulus should be designed to target services more than historical fiscal stimulus did (e.g., services-related expenditure such as health, education).
- Design considerations:
  - Consider the differing expenditure responsibilities of central versus sub-national governments when designing stimulus to achieve aggregate objectives.
  - Account for potential monetary policy responses and nonlinear cross-province interactions when scaling up provincial estimates to aggregate policy design.

*Source: IMF Working Paper wp17273 (excerpts and tables as provided).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2017/wp17273.pdf_