## _wp16106

## Source details

**Canonical URL:** [_wp16106](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16106.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16106.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16106.pdf.json)

---

### I. Introduction — role of China and hypotheses for import slowdown
- China accounted for one-third of global growth during 2000–14.
- Exports to China rose from 3 percent to 9 percent of world exports and from 9 percent to 22 percent of regional exports during 2000–14.
- China accounts for over a quarter of exports and imports of emerging and developing economies and is a top-ten trading partner for over 100 economies that constitute about 80 percent of world GDP.
- Concerns motivating the analysis:
  - China’s slowdown since 2013 and financial market turmoil amid weaker trade volumes.
  - Global trade growth sluggishness: in the first half of 2015, compounded by low commodity prices, world trade value declined.
  - China’s contribution to global trade slowdown was unusually large in 2015 (measured by import volume of goods), contrasting with its positive post-2008–09 role.
- Hypotheses suggested to explain weak import demand from China:
  - After investment-driven growth and a massive stimulus since 2008, overall domestic demand moderated, slowing import demand.
  - Structural transition toward consumption and services reduces import intensity since consumption is relatively less import-intensive than investment and exports.
  - Substitution of imports with domestic production (“onshoring”).
  - China’s exchange rate appreciated by about 15 percent in real effective terms between mid-2014 and mid-2015, eroding export price competitiveness (which reduces demand for imported intermediate inputs) despite making imports cheaper directly.

### II. Contributions and methodology highlights
- Data and measures used:
  - Merchandise goods imports from customs data and goods and services imports series from national accounts.
  - Study of processing versus nonprocessing merchandise imports using customs data.
  - Import-intensity adjusted demand (IAD): weighted average of private consumption, government consumption, investment, and exports with time-varying component-specific import intensity derived from input-output tables.
  - Explicit consideration of onshoring in empirical estimation.
  - Use of PPI-based real effective exchange rate (REER) in addition to CPI-based REER.
- Empirical strategy:
  - Counterfactual nonrebalancing scenarios used to quantify impact of rebalancing on import demand.
  - Baseline sample 2002:Q1–2015:Q3; alternate sample 2005:Q3–2015:Q3 and longer sample 1996:Q1–2015:Q3 for national accounts series.
  - Estimation uses year-over-year log differences (four-quarter changes) and Newey-West (1987) estimator; robustness checks include quarter-on-quarter seasonally adjusted X12 series.

### III. Stylized facts and recent developments (selected quantitative points)
- Current account and trade balance:
  - Current account surplus: peaked at 10.0 percent of GDP in 2007; declined to below 2 percent in 2011; reached 2.8 percent of GDP in 2015.
  - Trade surplus: 8.7 percent of GDP in 2007 → 2.4 percent in 2011 → 3.5 percent in 2015.
  - Over the last 4 years: services deficit widened by 1.3 percentage points of GDP; goods trade surplus increased by 2.4 percentage points of GDP; imports declined by 6.6 percentage points of GDP; exports declined by 4.2 percentage points of GDP.
- Nominal and real imports:
  - Nominal goods imports: more than 20 percent average annual growth between 2000 and 2011; contracted by 12.2 percent in 2015 (first contraction since 2009).
  - Import volume contracted by 0.7 percent in 2015 while import prices also declined sharply.
  - Real import growth averaged 11 percent (yoy) for 2006–13; slowed to around 3 percent over the last two years up to 2015:Q3.
  - Real imports contracted in 2015:Q1 by 3½ percent (yoy); recovered to about 2 percent and 3⅔ percent (yoy) in 2015:Q2 and 2015:Q3, respectively.
- Commodity versus noncommodity contributions in 2015:
  - Commodity imports ≈ one-third of total imports; about half of the import decline in 2015 due to fall in commodity imports.
  - Mineral fuels imports declined by 37 percent (yoy) in 2015.
  - Crude materials declined by 22 percent (yoy) in 2015.
  - Food imports increased by 5 percent (yoy); overall commodity imports declined by 27 percent.
  - Price effect alone would have reduced total commodity imports by about 33 percent (yoy) absent offsetting positive volume growth.
  - Food imports: 22 percent growth in volume terms in 2015.
  - Mineral fuels and crude materials import growth slowed to 8 percent and 1 percent (volume), respectively.
- Noncommodity imports in 2015:
  - Declined by 8 percent in 2015.
  - About 35 percent of the decline due to fall in non-commodity materials and another 35 percent due to fall in machinery and transportation equipment.
- Domestic demand components (growth deceleration magnitudes):
  - Investment decelerated by about 5¼ percentage points.
  - Exports decelerated by about 6 percentage points.
  - Consumption growth slowed by about 1¾ percentage points and remained faster than GDP growth.
- Real effective exchange rate:
  - Continued to appreciate; mid-2014 to mid-2015 appreciation about 15 percent.

### IV. Import intensity-adjusted demand (IAD) and onshoring
- IAD construction and properties:
  - IAD uses Input-Output database to track direct and indirect imported content; annual data up to 2011, quarterly interpolation to 2011, post-2011 weights held constant at 2011 values (following Bussiere and others (2011)).
  - Robustness checks with alternative interpolation and TiVA-based measures produce qualitatively similar results.
  - Estimated import contents: investment highest, then exports, then private consumption; government consumption lowest.
  - Import contents declining since mid-2000s, reflecting onshoring.
  - Although GDP growth moderated to about 7–7¼ percent in 2014–15, IAD growth was slower than GDP growth due to rebalancing, implying larger effect on imports than headline GDP slowdown.
- Processing trade and onshoring metrics:
  - Processing trade includes “processing and assembling” and “processing with imported materials.”
  - Processing trade share: less than 10 percent in early 1980s → about half of total trade until late 1990s → about a quarter of total imports and about a third of total exports in recent years.
  - Proxy for onshoring: ratio of processing imports to total exports: used to infer reliance on imported intermediate inputs. Historically 35‒40 percent until early 2000s; ≈20 percent of exports currently.
  - Table 1 (2014 customs regime proportions):
    - Ordinary Trade: Exports 51.4%, Imports 56.6%
    - Processing & assembling: Exports 3.9%, Imports 5.0%
    - Processing with imported materials: Exports 33.9%, Imports 21.8%
    - Customs warehousing trade: Exports 2.3%, Imports 5.1%
    - Entrepot trade by bonded area: Exports 4.7%, Imports 9.5%

### V. Empirical findings — aggregate and disaggregated import regressions
- Aggregate imports (national accounts):
  - Aggregate demand coefficients range from 1.5 to 1.8 depending on specification.
  - IAD-based models markedly improve goodness-of-fit relative to GDP models; example: IAD model (without GFC dummy) improves adjusted R-squared by more than 0.2 versus GDP model without GFC dummy.
  - IAD coefficients in multiple specifications: 1.974***, 1.895***, 1.965*** (t-statistics (8.649), (5.565), (7.766)); and 1.919***, 1.738***, 1.773*** (t-statistics (10.315), (9.104), (9.489)).
  - GDP coefficients example: 2.418***, 2.279***, 2.583*** (t-statistics (3.649), (3.038), (3.947)); and 1.598***, 1.433***, 1.493*** (t-statistics (3.532), (3.317), (3.192)).
  - Domestic demand, consumption, investment show positive and often significant elasticities:
    - Domestic demand examples: 1.509***, 1.528***; 1.401***, 1.434***.
    - Consumption: 0.736***, 0.743*** (t-statistics (3.012), (2.994)).
    - Investment: 0.619***, 0.638*** (t-statistics (4.625), (4.533)).
  - Onshoring positive and statistically significant across aggregate models: examples 0.919***, 0.907***, 0.921***, 1.299***, 1.182***, 1.243***, and another set 1.287***, 1.276***, 1.288***, 1.702***, 1.586***, 1.648***.
  - GFC dummy generally negative and often significant.
- Processing versus nonprocessing imports (customs-based regressions; Tables 7–8):
  - Processing imports:
    - IAD and GDP show strong positive elasticities when included: IAD 1.635***, 1.416***, 1.575*** (t-stats (8.416), (4.770), (7.967)); GDP 1.944***, 1.748***, 2.143*** (t-stats (3.461), (2.819), (4.118)).
    - Onshoring has large positive coefficients across models (examples): 1.840***, 1.741***, 1.961***, 2.399***, 2.324***, 2.794***, 2.151***, 2.195***, 2.144***, 2.181*** (with high t-statistics).
    - Exchange rate coefficients negative in eight out of 10 models but generally not statistically significant.
    - Fit statistics examples: Observations 39; R-squared values 0.863, 0.873, 0.871, 0.748, 0.818, 0.807, 0.900, 0.901, 0.904, 0.905; Adjusted R-squared 0.852, 0.858, 0.856, 0.727, 0.796, 0.784, 0.885, 0.886, 0.886, 0.887.
  - Nonprocessing imports:
    - Exports and external demand important: Exports coefficients 0.820***, 0.794***, 0.828***, 0.804*** (t-statistics (4.232), (4.393), (4.197), (4.325)).
    - Onshoring positive and highly significant: 2.796***, 2.651***, 3.100***, 3.126***, 2.936***, 3.632***, 3.654***, 3.672***, 3.653***, 3.674***.
    - Exchange rate coefficients positive in seven out of 10 models but not statistically significant.
    - Fit statistics examples: Observations 39; R-squared 0.622, 0.672, 0.654, 0.596, 0.675, 0.659, 0.861, 0.863, 0.861, 0.863; Adjusted R-squared 0.590, 0.633, 0.614, 0.562, 0.637, 0.619, 0.840, 0.842, 0.835, 0.838.
- Robustness and diagnostics:
  - Regression results stable across specifications; main explanatory variables statistically significant.
  - Variance inflation factors (VIFs) < 3 for benchmark regressions.
  - Results for goods-only series and customs data qualitatively similar: importance of domestic and external demand; ambiguous exchange rate effect; IAD superior fit; onshoring drag.
  - Processing imports dynamics mainly driven by export demand.

### VI. Quantification of contributions to the import slowdown (Section V)
- Aggregate slowdown magnitude:
  - Real import growth decelerated by about 7¾ percentage points: from 11 percent (until 2013) down to 3¼ percent in 2014:Q1–2015:Q3.
- Main contributors to slowdown (model estimates):
  - Weaker investment accounts for about 40–50 percent of the slowdown since 2014.
  - Weaker exports account for about 40 percent of the slowdown.
    - A regression of exports indicates about a quarter of export slowdown is due to stronger RMB, implying about 10 percent of import slowdown is an indirect effect via weaker exports.
- Onshoring and 2015 developments:
  - Share of imported inputs for exports ≈ 22½ percent in 2014; declined by about 2¾ percentage points in 2015, dragging import growth by about 2¼ percentage points.
  - Exports remained flat in 2015 (up to Q3), down from about 5 percent growth in 2014, contributing about 2½ percentage points to import slowdown.
- Rebalancing scenarios and estimated effects:
  - Narrow rebalancing (composition shift from investment to consumption over last two years):
    - Consumption grew faster than GDP by about 0.5 percentage points; investment grew slower than GDP by about 0.4 percentage points.
    - Model estimates imply about 0.1 percentage points of the 7¾ percentage point slowdown is due to this narrow rebalancing: +0.2 ppt higher imports from more consumption and -0.3 ppt lower imports from less investment → net -0.1 ppt (reported as about 0.1 ppt due to rebalancing).
    - Conclusion: narrowly defined rebalancing explains very little of the slowdown.
  - Broader definitions:
    - Defining benchmark pace as average growth rates in 2012–13 and attributing deviations in 2014–15 to rebalancing suggests about half of import slowdown could be attributable to continued rebalancing progress.
    - Alternative attribution using gap between GDP growth and IAD growth over last two years implies about 30 percent of import slowdown due to rebalancing (Table 2 column 3).
- Uncertainty:
  - Large uncertainty remains about the impact of rebalancing on imports due to difficulties in identifying the counterfactual nonrebalancing path.

### VII. Conclusions — synthesis of drivers
- Identified drivers of import slowdown:
  - Slower growth after unwinding of post-GFC stimulus.
  - Composition shift toward consumption and services (less import intensive).
  - Onshoring: substitution of imports with domestic production reducing import intensity.
  - Real effective appreciation (about 15 percent mid-2014 to mid-2015) eroding export competitiveness and indirectly lowering import demand for intermediate inputs.
- Key quantified findings:
  - Weaker investment accounts for about 40–50 percent of the import slowdown over the last two years.
  - Weaker exports account for about 40 percent of slowdown; about a quarter of export-related slowdown due to stronger exchange rate (≈10 percent of import slowdown indirect effect).
  - Onshoring has been a drag on import growth (baseline estimates: reduced import growth by about ¾–1 percentage points per year post-2002; post-2005 sample implies 1 to 1¾ percentage points per year), but has not been an additional accelerating drag over the last two years beyond its ongoing pace.
  - Depending on rebalancing definition, preliminary assessments suggest around half of recent import slowdown could be attributable to rebalancing away from investment and exports toward consumption, with substantial uncertainty.

### VIII. Appendices — alternative import-intensity measure and robustness
- APPENDIX II (TiVA-based alternative):
  - Uses OECD–WTO TiVA database and Hong and others (2016) construction for 1995, 2000, 2005, and 2008–2011; extends through 2014 via transformations using gross import end-use shares and interpolation/extrapolation.
  - Key finding: import intensity of investment declined by about 8 percentage points over the last decade after having increased by more than 10 percentage points between mid-1990s and mid-2000s. Import intensity for consumption relatively stable.
- APPENDIX III (seasonally-adjusted quarter-on-quarter regressions; 2002:Q1–2015:Q3):
  - Results reaffirm main conclusions: domestic and external demand matter; REER effects ambiguous; IAD improves fit; investment elasticity > consumption/export elasticities; onshoring drags on import growth.
  - Selected coefficients (Table A1, aggregate imports of goods and services):
    - IAD 1.847***; GDP 2.106*** 2.196*** 2.198***; Domestic demand 2.281*** 2.293***; Consumption 0.712** 0.717**; Investment 1.127***; Exports 0.623*** 0.624*** 0.567*** 0.577***.
    - Onshoring examples: 0.912*** 1.033*** 1.013*** 2.036*** 2.155*** 2.223*** 0.450 0.451 0.591** 0.609**.
    - Observations 55; Adjusted R-squared examples 0.643, 0.641, 0.640, 0.395, 0.426, 0.418, 0.708, 0.709, 0.686, 0.685.
  - Selected coefficients (Table A2, aggregate imports of goods):
    - IAD 1.540*** 1.551*** 1.506***; GDP 1.974*** 2.189*** 2.053***; Domestic demand 2.113*** 2.065***; Consumption 0.630** 0.585**; Investment 1.017*** 0.988***; Exports 0.564*** 0.566*** 0.509*** 0.516***.
    - Onshoring examples: 1.536*** 1.673*** 1.624*** 2.459*** 2.588*** 2.620*** 1.085*** 1.084*** 1.240*** 1.258***.
    - Observations 55; Adjusted R-squared examples 0.633, 0.618, 0.628, 0.497, 0.510, 0.508, 0.736, 0.739, 0.700, 0.701.
- APPENDIX IV (export slowdown decomposition; 2002:Q1–2015:Q3):
  - External demand coefficients: 2.884*** 2.889*** 3.142*** 3.007*** (t-stats (7.180), (4.163), (4.323), (8.004)).
  - CPI-based REER: -0.771*** -0.771*** (t-stats (-7.235), (-7.259)); PPI-based REER: -0.809*** -0.827*** (t-stats (-5.097), (-5.414)).
  - Observations 55; Adjusted R-squared 0.723, 0.751, 0.720, 0.755.

*Source: _wp16106 - References; content unit _wp16106 (IMF working paper).*

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

### _wp16106 - References

### I. Introduction: role of China and hypotheses for import slowdown
- China accounted for one-third of global growth during 2000–14.
- Exports to China rose from 3 percent to 9 percent of world exports and from 9 percent to 22 percent of regional exports over the same period.
- China accounts for over a quarter of exports and imports of emerging and developing economies and is a top-ten trading partner for over 100 economies that constitute about 80 percent of world GDP.
- Concerns motivating the analysis:
  - China’s slowdown since 2013 and financial market turmoil amid weaker trade volumes.
  - Global trade growth sluggishness: in the first half of 2015, compounded by low commodity prices, world trade value declined.
  - China’s contribution to global trade slowdown was unusually large in 2015 (measured by import volume of goods), contrasting with its positive post-2008–09 role.

- Hypotheses suggested to explain weak import demand from China:
  - After investment-driven growth and a massive stimulus since 2008, overall domestic demand moderated, slowing import demand.
  - Structural transition toward consumption and services reduces import intensity since consumption is relatively less import-intensive than investment and exports.
  - Substitution of imports with domestic production (“onshoring”).
  - China’s exchange rate appreciated by about 15 percent in real effective terms between mid-2014 and mid-2015, eroding export price competitiveness (which reduces demand for imported intermediate inputs) despite making imports cheaper directly.

### Contributions and methodology highlights
- Data and measures used:
  - Use of merchandise goods imports from customs data and goods and services imports series from national accounts to capture a comprehensive picture.
  - Study of processing versus nonprocessing merchandise imports using customs data.
  - Introduction of import-intensity adjusted demand (IAD): a weighted average of private consumption, government consumption, investment, and exports with time-varying component-specific import intensity derived from input-output tables.
  - Explicit consideration of onshoring in empirical estimation.
  - Use of PPI-based real effective exchange rate (REER) instead of the more common CPI-based REER.
- Aim to quantify impact of rebalancing on import demand by comparing slowdown to several counterfactual nonrebalancing scenarios.

### Key empirical findings and quantitative results
- Both domestic and external demand are important drivers of import dynamics, reflecting China’s role as an assembling hub in global value chains.
- Main contributors to the import slowdown since 2014:
  - Weaker investment (partly due to progress in rebalancing) accounts for about 40‒50 percent of the import slowdown.
  - Weaker exports account for about 40 percent of the slowdown; about a quarter of this export-related component is due to a stronger exchange rate.
- Onshoring:
  - Onshoring is a drag on import growth, but it has not been an additional drag over the last two years because it continues to slow import growth at a constant pace.
- Uncertainty:
  - Large uncertainty remains about the impact of rebalancing on imports due to difficulties in identifying the counterfactual nonrebalancing path.

### II. Recent developments and stylized facts (selected quantitative points)
- Current account and trade balance:
  - China’s current account surplus peaked at 10.0 percent of GDP in 2007.
  - It declined to below 2 percent in 2011 and reached 2.8 percent of GDP in 2015.
  - Trade surplus narrowed from 8.7 percent of GDP in 2007 to 2.4 percent in 2011, then increased to 3.5 percent in 2015.
  - Over the last 4 years, services deficit widened by 1.3 percentage points of GDP, goods trade surplus increased by 2.4 percentage points of GDP, and imports declined by 6.6 percentage points of GDP while exports declined by 4.2 percentage points of GDP.
- Nominal goods imports:
  - More than 20 percent average annual growth between 2000 and 2011.
  - Contracted by 12.2 percent in 2015 (first contraction since 2009).
  - Import volume contracted by 0.7 percent in 2015 while import prices also declined sharply.
- Commodity versus noncommodity contributions in 2015:
  - Commodity imports account for about one-third of total imports; about half of the import decline in 2015 was due to fall in commodity imports.
  - Imports of mineral fuels declined by 37 percent (yoy) in 2015.
  - Imports of crude materials declined by 22 percent (yoy) in 2015.
  - Food imports increased by 5 percent (yoy), yet overall commodity imports declined by 27 percent.
  - The price effect alone would have reduced total commodity imports by about 33 percent (yoy) absent offsetting positive volume growth.
  - Food imports: 22 percent growth in volume terms in 2015.
  - Mineral fuels and crude materials import growth slowed to 8 percent and 1 percent (volume), respectively (footnote context).
- Noncommodity imports in 2015:
  - Declined by 8 percent in 2015.
  - About 35 percent of the decline due to fall in non-commodity materials and another 35 percent due to fall in machinery and transportation equipment.
- Real import growth:
  - 11 percent (yoy) average for 2006–13.
  - Slowed to around 3 percent over the last two years up to 2015:Q3.
  - Real imports contracted in 2015:Q1 by 3½ percent (yoy).
  - Real import growth recovered to about 2 percent and 3⅔ percent (yoy) in Q2 and Q3 of 2015, respectively.
- Domestic demand components (growth deceleration magnitudes):
  - Investment decelerated by about 5¼ percentage points.
  - Exports decelerated by about 6 percentage points.
  - Consumption growth slowed by about 1¾ percentage points and remained faster than GDP growth.
- Real effective exchange rate:
  - Continued to appreciate, at a faster pace in recent years (mid-2014 to mid-2015 appreciation quantified above as about 15 percent).

### Literature context (selected numeric elasticities and findings)
- Kwack and others (2007): multilateral import price elasticity 0.50 and income elasticity 1.57 (1984–2003 panel).
- Thorbecke (2006): exchange rate elasticity for Chinese imports from the United States ranges from 0.42 to 2.04; income elasticity between 1.05 and 1.21 (1988–2005 bilateral U.S.-China data).
- Mann and Plück (2007): Chinese imports from the United States income elasticity 0.74 for capital goods and 2.25 for consumer goods (1980–2004).
- Thorbecke and Smith (2012): a 10 percent RMB appreciation would increase imports for processing and ordinary imports by three to four percent (panel including imports from 25 countries), inferring import substitution potential as China climbs the value chain.
- Cheung, Chinn, and Qian (2012): aggregate imports rise in response to an RMB depreciation and decline with Chinese GDP after accounting for imports incorporated into exports.

### Structure of the paper (sections referenced)
- Section II: Stylized facts and discussion on progress in rebalancing.
- Section III: Chinese processing trade and onshoring developments.
- Section IV: Empirical estimations of Chinese import dynamics.
- Section V: Relative contribution of various factors to recent import slowdown based on empirical results.
- Section VI: Conclusions.

*Source: _wp16106 - References*

### 17.6 percent, respectively, while refined petroleum products remained flat in 2015. Crude material imports has been

### _wp16106 - 17.6 percent, respectively, while refined petroleum products remained flat in 2015. Crude material imports has been

### Summary findings
- Refined petroleum products remained flat in 2015.
- Crude material imports were mixed:
  - Iron ore up by 2.2 percent.
  - Copper up by 12.3 percent.
  - Aluminum down by 18.5 percent.
- Real Import Growth (yoy, in percent) is plotted for quarters:
  - 06Q1, 07Q1, 08Q1, 09Q1, 10Q1, 11Q1, 12Q1, 13Q1, 14Q1, 15Q1.
- Historical average or benchmark reported as:
  - 11.0 percent (06Q1-13Q4).

### Data notes and references
- Footnote indicator: 4 See Appendix I for more discussion on national account data that we use in this paper.
- Chart axis ticks or labels observed:
  - -30, -20, -10, 0, 10, 20, 30, 40.

*Source: _wp16106 - 17.6 percent, respectively, while refined petroleum products remained flat in 2015. Crude material imports has been*

### 3.3 p ercen t

### 3.3 p ercen t

### Rebalancing: external vs internal
- External rebalancing progress:
  - Export-to-GDP ratio in nominal terms: peaked at 36 percent in 2006, declined to 27 percent in 2011, and further to 21½ percent in 2015.
  - In real terms, the export-to-GDP ratio declined by about 5¾ percentage points between 2007 and 2015.
  - Real exports growth: more than 15 percent every year on average between 1990 and 2011; moderated to about 5½ percent per year on average over the last 4 years (including only about 1 percent in 2015).
  - Contribution of exports to real GDP growth: about 1½ percent over the last four years, compared to about 4¼ percent in the previous decade.
- Internal rebalancing (investment toward consumption) has been slower:
  - Nominal share changes between 2011 and 2015: investment in GDP declined by about 4 percentage points; consumption increased by about 3 percentage points.
  - Real terms between 2011 and 2015: share of consumption in real GDP increased by about 1¼ percentage points; share of investment declined/remained by ½ percentage point.
  - Price developments drove much of nominal internal rebalancing: CPI increased by about 9 percent while PPI fell by about 10 percent over the period referenced.
  - Contrast with 2007–2011: investment-to-GDP ratio increased by about 6½ percentage points in both real and nominal terms on the back of large stimulus after the global financial crisis.

### Import intensity-adjusted demand (IAD) and import implications
- Definition and construction:
  - IAD is a weighted average of private consumption, government consumption, investment, and exports using the share of import content in each component as weights.
  - Chinese IAD constructed based on the Input-Output database, tracking direct and indirect (embedded) imports via the import matrix and domestic matrix.
  - Annual data available up to 2011; quarterly weights constructed by linear interpolation up to 2011; for the period after 2011, weights assumed the same as in 2011 following Bussiere and others (2011).
  - Robustness checks: alternative interpolation methods (constant weights; time-varying weights from U.N. Comtrade) yield qualitatively similar import regression results.
- Empirical observations:
  - Estimated import contents: investment has the highest import content, followed by exports and private consumption; government consumption has the lowest import content.
  - Import contents have been on a somewhat declining trend since mid-2000s, reflecting onshoring.
  - Although overall GDP growth moderated to about 7–7¼ percent in 2014–15, IAD growth was slower than GDP growth due to rebalancing from investment and exports to consumption, implying a larger effect on imports than headline GDP slowdown would suggest.

### China’s processing trade and onshoring
- Customs regime classification:
  - Processing trade comprises “processing and assembling” and “processing with imported materials.”
  - “Processing and assembling”: foreign firm supplies components, retains ownership/control, Chinese firms import raw materials for free and send value-added products back to the same firm; imports under this regime are 100 percent duty free.
  - “Processing with imported materials”: Chinese assembly plant imports components and retains control, firms must pay for imported intermediate inputs and may obtain full duty rebate after exporting processed/final goods.
  - Processing trade is defined as the sum of these two regimes for analysis.
- Historical and recent shares:
  - Processing trade accounted for less than 10 percent of total trade in early 1980s; increased through 1980s and 1990s, with processing imports and exports accounting for about a half of total trade until late 1990s.
  - Since early 2000s, share of processing trade in both imports and exports has declined; processing trade now accounts for about a quarter of total imports and about a third of total exports.
  - Proxy for onshoring: ratio of processing imports to total exports. China used to rely on about 35‒40 percent of its exports on imported intermediate inputs until early 2000s; processing imports account for only about 20 percent of exports currently.
- Table 1 (2014 customs regime proportions, trade value):
  - Ordinary Trade: Exports 51.4%, Imports 56.6%
  - Processing & assembling: Exports 3.9%, Imports 5.0%
  - Processing with imported materials: Exports 33.9%, Imports 21.8%
  - Customs warehousing trade: Exports 2.3%, Imports 5.1%
  - Entrepot trade by bonded area: Exports 4.7%, Imports 9.5%
  - Other minor regimes reported as in table (values preserved above).

### Empirical estimation: specification and estimation approach
- Model specification (growth form, Newey-West estimator):
  - The empirical model uses year-over-year (yoy) log differences over four quarters for real imports (M), various demand aggregates (Y), and REER.
  - A proxy variable for onshoring (ratio of processing imports to total exports) is included.
  - Newey-West (1987) estimator applied to address potential serial correlation.
  - Alternative estimation using quarter-over-quarter growth of seasonally adjusted series produced similar results.
- Data and sample periods:
  - Baseline sample period: 2002:Q1–2015:Q3 to capture post-WTO structural shift.
  - Shorter sample period: 2005:Q3–2015:Q3 to reflect July 2005 exchange rate reform and availability of customs data for disaggregated merchandise imports.
  - Series considered: aggregate imports of goods and services (national accounts), goods only (national accounts), and customs data disaggregated into processing and nonprocessing imports.
  - REER measures used: CPI-based REER and PPI-based REER.

### Empirical results: aggregate imports (national accounts) — key findings and coefficients
- Demand variables:
  - All combinations of demand variables are positively associated with import demand with statistically significant coefficients.
  - Aggregate demand coefficients (range): 1.5 to 1.8 depending on specification.
  - Disaggregate demand: investment elasticity higher than exports and consumption; import intensity ranking consistent with Input-Output estimates (investment highest, then exports, then consumption).
- REER effects:
  - Coefficient estimates on CPI-based and PPI-based REER are counterintuitive: negative in some models and not statistically significant in others.
  - Ambiguous net effect attributed to offsetting impacts of RMB appreciation on value-added component of exports and imported intermediate input costs.
- IAD performance:
  - IAD improves goodness-of-fit markedly relative to conventional GDP models.
  - Example: model with IAD (without GFC dummy) improves adjusted R-squared by more than 0.2 compared to GDP model without GFC dummy.
  - IAD-based model performs well in explaining trade collapse during the global financial crisis, though GFC dummy remains statistically significant in some models.
- Onshoring effect magnitude:
  - Onshoring has reduced import growth by about ¾–1 percentage points per year in the baseline (post-2002) sample, given that the share of imported inputs for exports has been declining by about 1 percentage point every year.
  - For the shorter post-2005 sample, onshoring effect is larger: estimation implies onshoring reduced import growth by about 1 to 1¾ percentage points per year.

### Robustness, diagnostics, and additional findings
- Stability and multicollinearity:
  - Regression results are stable across specifications; main explanatory variables are statistically significant.
  - Variance inflation factors (VIFs) for benchmark regressions are lower than three for each explanatory variable, suggesting limited multicollinearity concerns (rule of thumb threshold 10).
- Results for goods-only series and customs data:
  - Tables 2 and 3 (aggregate imports of goods and services; goods only) show qualitatively similar results: importance of domestic and external demand, ambiguous exchange rate effect, superior fit of IAD, and onshoring drag on import growth.
  - For post-2005 period and customs data, results remain broadly similar; onshoring effect appears to be a bigger drag for merchandise goods imports while investment elasticity is somewhat lower for customs-based goods imports.
- Processing vs non-processing imports:
  - Different dynamics observed between processing and non-processing imports (detailed coefficient comparisons in subsequent tables referenced).
  - Export demand is the main factor driving processing goods import dynamics.

*Source: IMF working paper content unit _wp16106 - 3.3 p ercen t*

### 0.4 percent increase in processing import demand, while domestic demand is not statistically

### _wp16106 - 0.4 percent increase in processing import demand, while domestic demand is not statistically significant factor.

### Regression results: Imports of processing and nonprocessing goods (Tables 7–8)
- Processing imports:
  - Coefficient estimates on exchange rate are negative in eight out of 10 models but are not statistically significant.
  - Onshoring has a large positive coefficient in all models (examples):
    - Onshoring: 1.840***, 1.741***, 1.961***, 2.399***, 2.324***, 2.794***, 2.151***, 2.195***, 2.144***, 2.181*** (t-statistics shown in parentheses).
  - GFC dummy shows negative and in many models statistically significant coefficients (examples):
    - GFC dummy: -0.031**, -0.039***, -0.089***, -0.106*** (other columns less significant).
  - IAD and GDP show strong positive elasticities in models where included:
    - IAD: 1.635***, 1.416***, 1.575*** (t-statistics: (8.416), (4.770), (7.967)).
    - GDP: 1.944***, 1.748***, 2.143*** (t-statistics: (3.461), (2.819), (4.118)).
  - Fit statistics (examples):
    - Observations: 39 across models.
    - R-squared: 0.863, 0.873, 0.871, 0.748, 0.818, 0.807, 0.900, 0.901, 0.904, 0.905.
    - Adjusted R-squared: 0.852, 0.858, 0.856, 0.727, 0.796, 0.784, 0.885, 0.886, 0.886, 0.887.

- Nonprocessing imports:
  - Exports and external demand are important factors:
    - Exports: 0.820***, 0.794***, 0.828***, 0.804*** (t-statistics: (4.232), (4.393), (4.197), (4.325)).
  - Exchange rate coefficients are positive in seven out of 10 models but not statistically significant.
  - Onshoring again positive and highly significant across models:
    - Onshoring: 2.796***, 2.651***, 3.100***, 3.126***, 2.936***, 3.632***, 3.654***, 3.672***, 3.653***, 3.674*** (t-statistics shown).
  - Fit statistics (examples):
    - Observations: 39 across models.
    - R-squared: 0.622, 0.672, 0.654, 0.596, 0.675, 0.659, 0.861, 0.863, 0.861, 0.863.
    - Adjusted R-squared: 0.590, 0.633, 0.614, 0.562, 0.637, 0.619, 0.840, 0.842, 0.835, 0.838.

### Aggregate imports (Tables 9–10; national accounts, 1996:Q1–2015:Q3)
- Sample and data notes:
  - Longer sample period since 1996:Q1 used for import series from national accounts due to data availability.
  - Pre-1995 period truncated to avoid concerns about exchange rate; Chinese REER depreciated sharply by about 30 percent in January 1994.
- Key coefficient patterns:
  - IAD and GDP show strong positive elasticities in aggregate models:
    - Example IAD: 1.974***, 1.895***, 1.965*** (t-statistics (8.649), (5.565), (7.766)); and 1.919***, 1.738***, 1.773*** (t-statistics (10.315), (9.104), (9.489)).
    - GDP: 2.418***, 2.279***, 2.583*** (t-statistics (3.649), (3.038), (3.947)); and 1.598***, 1.433***, 1.493*** (t-statistics (3.532), (3.317), (3.192)).
  - Domestic demand, consumption, and investment often positive and significant in many specifications:
    - Domestic demand examples: 1.509***, 1.528***; 1.401***, 1.434***.
    - Consumption examples: 0.736***, 0.743*** (t-statistics (3.012), (2.994)).
    - Investment examples: 0.619***, 0.638*** (t-statistics (4.625), (4.533)).
  - Onshoring positive and statistically significant across aggregate models:
    - Examples: 0.919***, 0.907***, 0.921***, 1.299***, 1.182***, 1.243***, 0.852***, 0.880***, 0.888***, 0.920***.
    - Another set: 1.287***, 1.276***, 1.288***, 1.702***, 1.586***, 1.648***, 1.350***, 1.379***, 1.385***, 1.417***.
  - GFC dummy generally negative and often significant in aggregate models.
- Fit statistics (examples):
  - Observations: 79 across models.
  - Adjusted R-squared examples: 0.803, 0.818, 0.817, 0.504, 0.623, 0.601, 0.829, 0.825, 0.826, 0.823; and 0.812, 0.818, 0.819, 0.524, 0.617, 0.600, 0.822, 0.820, 0.818, 0.816.

### Contribution to the import slowdown (Section V)
- Aggregate slowdown magnitude:
  - Real import growth decelerated by about 7¾ percentage points: from 11 percent until 2013 down to 3¼ percent in 2014:Q1–2015:Q3.
- Main contributors to slowdown (model estimates, Table 2 referenced):
  - Weaker investment accounts for about 40–50 percent of the slowdown since 2014.
  - Weaker exports account for about 40 percent of the slowdown.
  - A separate regression of exports indicates about a quarter of export slowdown is due to stronger RMB, implying about 10 percent of import slowdown is an indirect effect via weaker exports.
- Role of onshoring and 2015 dynamics:
  - Onshoring reduced the share of imported inputs for production/re-exports:
    - Share was about 22½ percent in 2014, then declined by about 2¾ percentage points in 2015, dragging import growth by about 2¼ percentage points.
  - Exports remained flat in 2015 (up to Q3), down from about 5 percent growth in 2014, contributing about 2½ percentage points to import slowdown.

### Rebalancing effect (definitions and scenarios)
- Narrow rebalancing (composition shift from investment to consumption over last two years):
  - Consumption grew faster than GDP by about 0.5 percentage points.
  - Investment grew slower than GDP by about 0.4 percentage points.
  - Model estimates (Table 2, column 10) imply about 0.1 percentage points of the 7¾ percentage point slowdown is due to this narrow rebalancing:
    - +0.2 ppt higher imports from more consumption and -0.3 ppt lower imports from less investment → net -0.1 ppt (reported as about 0.1 ppt due to rebalancing).
  - Conclusion: Impact of narrowly defined rebalancing is very limited.
- Broader definitions:
  - Defining benchmark pace as average growth rates in 2012–13 and attributing deviations in 2014–15 to rebalancing suggests about half of import slowdown could be attributable to continued rebalancing progress.
  - Alternative: Gap between GDP growth and IAD growth over last two years attributed to rebalancing. Under this scenario (Table 2 column 3), about 30 percent of import slowdown is due to rebalancing.

### Conclusion (Section VI) — factors contributing to import slowdown
- Identified drivers:
  - Slower growth: moderating domestic demand following unwinding of massive post-GFC stimulus.
  - Composition shift in import demand: economy moving toward consumption and services (less import intensive) from investment and exports.
  - Onshoring: substitution of imports with domestic production, combined with real effective appreciation eroding price competitiveness of Chinese exports.
- Key quantified findings:
  - Weaker investment has been the main factor accounting for about 40-50 percent of the import slowdown over the last two years.
  - Weaker exports also account for about 40 percent of slowdown, of which about a quarter is due to stronger exchange rate eroding competitiveness in export sector.
  - Substitution of imported intermediate inputs with domestic production has not been an additional drag over this period but continues to slow import growth at a roughly constant pace.
  - Large uncertainty surrounds the impact of rebalancing; preliminary assessments across rebalancing definitions suggest around half of recent import slowdown could be attributable to ongoing rebalancing away from investment and exports toward consumption.

### Appendix I (trade data from national accounts) — data construction and comparability
- IMF WEO-based estimates produce real GDP by expenditure with breakdowns into private/public and goods/services components by:
  - Using balance of payments and customs trade data, anchored to official real GDP and contributions by final consumption, investment, and net exports.
  - Deflators: export and import prices from customs for goods; U.S. CPI as deflator for services; CPI for consumption; PPI for investment.
- Differences between customs and balance of payments:
  - Customs data are census-based; balance of payments adjusts to national accounts concepts, supplementing coverage and eliminating duplications.
  - In China, nominal imports from customs used to exceed balance of payments by more than 15 percent before the global financial crisis and by about 10 percent more recently.
  - Growth rates of imports from customs and BOP show very similar developments since 2000 (chart evidence in the source).

*Source: _wp16106 - 0.4 percent increase in processing import demand, while domestic demand is not statistically significant factor (PDF chapter/section).*

### APPENDIX II. ALTERNATIVE MEASURE OF IMPORT INTENSITY

### APPENDIX II. ALTERNATIVE MEASURE OF IMPORT INTENSITY

### Methodology (TiVA-based alternative)
- Alternative source: Trade in Value-Added (TiVA) database by the OECD–WTO.
- Construction approach (Hong and others, 2016):
  - For years 1995, 2000, 2005, and 2008–2011: import intensity is measured as domestic value added of trade partners ultimately absorbed by China for consumption or investment in percent of Chinese consumption or investment respectively.
  - For years through 2014: calculate the share of gross imports used for consumption or investment by multiplying the share of gross imports related to consumption and investment (based on end use categories) in total Chinese gross imports with imports from Chinese national accounts side.
  - Interpolation/extrapolation: growth rates of the latter estimated series are used to interpolate interim years (1996–99, 2001–04, 2006–07) and to extrapolate outer years (2012–14).

### Key findings on import intensity
- The alternative TiVA-based measure shows:
  - Import intensity has been on a declining trend for investment over the past decade while relatively stable for consumption.
  - The import intensity of investment, which increased by more than 10 percentage points between mid-1990s and mid-2000s, has declined by about 8 percentage points over the last decade.
- Chart label reproduced: "China's import intensity (% ) (consumption vs. investment)" with time series spanning 1995–2014.

### Appendix III — Estimation with seasonally-adjusted series (robustness)
- Estimation approach:
  - Re-run regressions using quarter-on-quarter growth rates.
  - All series seasonally adjusted using X12.
  - Sample: post-WTO period (2002:Q1 to 2015:Q3).
  - Dependent variables: imports of goods and services (Table A1) and imports of goods only (Table A2) from national accounts.
- Overall qualitative conclusions (from text summary):
  - Both domestic and external demands are important factors driving import dynamics.
  - Coefficient estimates on real effective exchange rates are either opposite signs or statistically not significant.
  - Models with IAD series improve the goodness-of-fit compared to the ones with conventional GDP series.
  - Estimated import elasticity for investment is higher than those for consumption and exports.
  - Onshoring drags on import growth.

### Selected quantitative results from Table A1 (Aggregate imports of goods and services; 2002:Q1–2015:Q3)
- Model coefficients (selected, with significance):
  - IAD 1.847*** (t-stats ( 4.137)(4.735)( 4.526) reported across models)
  - GDP 2.106*** 2.196*** 2.198*** (t-stats ( 3.360)( 3.162)( 3.204))
  - Domestic demand 2.281*** 2.293*** (t-stats ( 5.686)( 5.794))
  - Consumption 0.712** 0.717** (t-stats (2.343)( 2.469))
  - Investment 1.127*** 1.127*** (t-stats (5.165)( 5.266))
  - Exports 0.623*** 0.624*** 0.567*** 0.577*** (t-stats ( 5.960)( 6.221)(5.809)( 6.010))
  - CPI-based REER coefficients reported as 0.053- 0.2690.034- 0.079 (t-stat ranges (0.239)( -0.981)( 0.164)( -0.378))
  - PPI-based REER coefficients reported as - 0.040- 0.025- 0.277-0.2030.052- 0.028 (t-stat ranges ( -0.122)(-0.082)( -0.612)( -0.569)( 0.230)(-0.119))
  - Onshoring coefficients: 0.912*** 1.033*** 1.013*** 2.036*** 2.155*** 2.223*** 0.450 0.451 0.591** 0.609** (t-stats ( 2.896)(3.899)( 3.991)( 5.928)( 4.683)( 4.890)( 1.431)( 1.433)(2.064)( 2.109))
  - GFC dummy coefficients: -0.014 -0.014 -0.038 -0.037 -0.015 -0.016 -0.023 -0.023 (t-stats ( -0.346)(-0.356)( -0.979)( -0.983)( -0.578)( -0.592)( -0.808)(-0.787))
  - Constant terms include values like -0.014 -0.012 -0.011 -0.011 -0.009 -0.010 -0.039*** -0.039*** -0.027** -0.028** (t-stats ( -0.945)( -1.090)(-0.975)( -0.639)( -0.542)( -0.585)( -3.018)( -3.123)( -2.128)(-2.226))
- Sample and fit:
  - Observations 55 (repeated across models).
  - Adjusted R-squared values: 0.643 0.641 0.640 0.395 0.426 0.418 0.708 0.709 0.686 0.685.
- Notes: Robust t-statistics in parentheses. Significance markers: *** p<0.01, ** p<0.05, * p<0.1.

### Selected quantitative results from Table A2 (Aggregate imports of goods; 2002:Q1–2015:Q3)
- Model coefficients (selected, with significance):
  - IAD 1.540*** 1.551*** 1.506*** (t-stats ( 3.165)( 3.524)( 3.442))
  - GDP 1.974*** 2.189*** 2.053*** (t-stats ( 3.243)( 3.238)( 3.073))
  - Domestic demand 2.113*** 2.065*** (t-stats ( 5.675)( 5.752))
  - Consumption 0.630** 0.585** (t-stats ( 2.096)( 2.120))
  - Investment 1.017*** 0.988*** (t-stats ( 4.439)( 4.301))
  - Exports 0.564*** 0.566*** 0.509*** 0.516*** (t-stats ( 5.715)( 6.011)( 5.163)( 5.407))
  - CPI-based REER coefficients reported as -0.143 -0.391 -0.188 -0.287 (t-stat ranges ( -0.561)( -1.428)( -0.931)( -1.346))
  - PPI-based REER coefficients reported as -0.313 -0.300 -0.488 -0.424 -0.241 -0.321 (t-stat ranges ( -0.870)( -0.879)( -1.100)( -1.188)( -1.033)( -1.286))
  - Onshoring coefficients: 1.536*** 1.673*** 1.624*** 2.459*** 2.588*** 2.620*** 1.085*** 1.084*** 1.240*** 1.258*** (t-stats (3.696)( 4.007)( 4.186)( 8.511)( 6.207)( 6.510)( 3.917)( 4.041)( 4.340)( 4.438))
  - GFC dummy coefficients: -0.015 -0.012 -0.035 -0.032 -0.012 -0.009 -0.019 -0.016 (t-stats ( -0.339)( -0.310)( -0.861)( -0.853)( -0.479)( -0.396)( -0.676)( -0.583))
  - Constant terms include values like -0.005 -0.004 -0.003 -0.008 -0.009 -0.007 -0.033*** -0.033*** -0.021 -0.020 (t-stats ( -0.334)( -0.302)( -0.252)( -0.485)( -0.569)( -0.437)( -2.707)( -2.749)( -1.632)( -1.618))
- Sample and fit:
  - Observations 55.
  - Adjusted R-squared values: 0.633 0.618 0.628 0.497 0.510 0.508 0.736 0.739 0.700 0.701.
- Notes: Robust t-statistics in parentheses. Significance markers: *** p<0.01, ** p<0.05, * p<0.1.

### Appendix IV — Contribution to the export slowdown (aggregate exports of goods and services; 2002:Q1–2015:Q3)
- Selected regression coefficients (Models 1–4):
  - External demand: 2.884*** 2.889*** 3.142*** 3.007*** (t-stats ( 7.180)( 4.163)(4.323)( 8.004))
  - CPI-based REER: -0.771*** -0.771*** (t-stats (-7.235)( -7.259))
  - PPI-based REER: -0.809*** -0.827*** (t-stats ( -5.097)(-5.414))
  - GFC: 0.011 0.027 (t-stats (-0.283)(0.696))
  - Constant: 0.018 0.026 0.007 0.021 (t-stats ( 1.116)( 0.947)(0.261)( 1.415))
- Sample and fit:
  - Observations 55.
  - Adjusted R-squared values: 0.723 0.751 0.720 0.755.
- Figures: Contributions to export growth (yoy, in percent) decomposed into Residual, GFC, REER, External demand, Constant, and Exports for periods 06Q1-13Q4 and 14Q1-15Q3; and period 2013–2015 (up to Q3) shown similarly.

*Source: APPENDIX II. ALTERNATIVE MEASURE OF IMPORT INTENSITY (from the supplied IMF document content)._

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16106.pdf_