## ch3annex

## Source details

**Canonical URL:** [ch3annex](https://www.imf.org/-/media/files/publications/reo/apd/2023/october/english/ch3annex.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/reo/apd/2023/october/english/ch3annex.pdf.md)
- [Structured JSON version](/-/media/files/publications/reo/apd/2023/october/english/ch3annex.pdf.json)

---

### Export Restrictions and Quality Downgrading
- Recent export restrictions increasingly target access to high-quality inputs; example: preventing procurement of logic chips with design rules under certain nanometer thresholds.
- Quality estimation scope and method:
  - Quality differentials estimated for 200,000 products at the Harmonized System (HS) 10-digit level using US trade data for 2002–18.
  - Following Khandelwal (2010), quality is inferred by comparing market shares conditional on price: if two products with the same price have different market shares, the product with the higher market share is inferred to be higher quality.
  - Estimated quality normalized between 0 and 1, with 1 indicating the highest quality for that product.
- Key quantitative findings:
  - Median estimated Chinese product quality in the semiconductor sectors is about one-third lower than the median quality for the OECD.
  - In a hypothetical cutoff of China and the OECD from each other’s semiconductor products, assuming substitution to a good-quality (75th percentile) product from the rest of the world:
    - Trade-weighted drop in quality is of about 5 percent for China and zero for the OECD.
  - For environmental goods, trade-weighted average quality loss is of about 11 percent for China and as high as 5 percent for the OECD.
  - For all goods combined, both sides would see input quality drop by about 8 percent.
  - Note: firms with a high-skilled labor force may experience output-quality reductions of similar proportions.

### Data Sources and Classification (Eora and trade/)
- Eora world input-output table:
  - Covers 189 countries, 26 sectors, and the years 1990-2021.
  - Value-added indicators and GVC metrics constructed using Stata’s icio command (Belotti and others, 2020).
- Definition and assignment rules:
  - GVC and non-GVC goods are intermediate-input sales taken from the square intermediate-input trade matrix. Sales between GVC and non-GVC sectors are assigned to the purchasing sector’s classification (example: agricultural (non-GVC) sales to transport equipment (GVC) considered GVC).
  - Investment and consumption goods correspond to sales to final demand from the rectangular final demand matrix. Eora end uses considered investment: gross fixed capital formation, changes in inventories, acquisitions less disposals of valuables. Considered consumption: household final consumption, non-profit institutions serving households, government final consumption.
- Annex Table 3.1.1 (sector classification): Eora sectors are mapped to classifications including GVC, Tradable non-GVC, and Non-tradable (list provided in source).

### GIMF-GVC Model: Structure and Calibration
- Model overview:
  - GIMF is an annual, multi-region, micro-founded DSGE model used with 10 regions: the United States, EU+, the other advanced economies, China, India, Indonesia, Japan, Korea, other Southeast Asia, and the rest of the world.
  - Added a tradable sector related to global value chains (GVC sector) for the chapter.
  - Households: some are non-Ricardian, finitely lived overlapping generations (saving households); others are liquidity constrained and consume all income each period.
  - Firms: profit-maximizing, monopolistically competitive; produce non-tradable, tradable, and GVC sectors.
  - GVC sector specifics: GVC goods are used in final goods and as inputs in production of other GVC goods; production combines capital and labor (Cobb-Douglas) with imported and domestically sourced GVC goods; output split between final goods inputs and inputs into other GVC goods, domestically and abroad.
  - Trade: regions trade final goods, tradable and GVC intermediate goods; flows tracked bilaterally; barriers modeled as non-tariff barriers (NTB) which act like tariffs but generate no fiscal revenue.
  - Policy rules: monetary policy follows inflation targeting; fiscal policy targets debt-GDP ratio.

- Calibration highlights:
  - Each region calibrated using OECD Inter-Country Input-Output Database for 2018.
  - Intertemporal elasticity of substitution for consumption common across regions at 0.2.
  - Share of liquidity constrained households:
    - 25 percent for the United States, EU+, the other advanced economies, and China.
    - 50 percent for the remaining regions.
  - Regions’ bias parameters set to match steady-state ratios from OECD (2021).

- Elasticities (Annex Table 3.2.2 - Key Production and Trade Elasticities):
  - Elasticity between Capital-Labor / GVC / Domestic/Imported / Different Regions (as in source table):
    - Consumption: - ; 1.5 ; 1.5
    - Investment: - ; 1.5 ; 1.5
    - Tradables: - ; 1.5 ; 1.5
    - GVC Goods: 0.5 ; 0.8* ; 0.8
  - Note: * Elasticity between domestic and imported when using GVC goods in the production of final goods or of other GVC goods.
  - Additional model elasticities and assembly:
    - Demand for goods in the GVC sector assumed relatively inelastic at about 0.8.
    - Elasticities for final consumption and investment goods and tradable intermediate goods about 1.5.
    - Final goods combine nontradable goods and a tradable goods bundle with elasticity 0.5.
    - Tradable goods bundle assembled from tradable intermediate goods and GVC goods with elasticity 0.95.

### Re-shoring and Friend-shoring: Observations and Calibration
- Observed re-shoring dimension (OECD foreign sourcing as percent of total sourcing):
  - OECD countries increased share of foreign sourcing by about 3 percentage points for GVC, non-GVC and investment goods, and by about 1 percentage point for consumption goods (from 2000 to 2021 panels shown).
- Observed friend-shoring dimension (China’s share in OECD sourcing):
  - China’s share in OECD sourcing increased by about 5 percentage points for GVC and non-GVC intermediates.
  - China’s share increased by about 10 percentage points for investment and consumption goods.

### Upside Scenario with Rebalancing (China)
- Scenario design in GIMF:
  - Public investment in China reduced by 0.5 percentage points of GDP per year for 15 years (halving China’s public investment), offset by an increase in social protection for households (represented by lump-sum transfers).
  - Assumption: the public investment that is cut is of low quality and does not directly reduce productivity in the private sector.
- Effects:
  - Short-term: rebalancing provides a boost to growth in China by lifting consumption demand by more than the decline in public investment.
  - Medium/long-term: rebalancing is GDP neutral in China as capital accumulation slows.
  - Rest of the world: rebalancing results in very small yet negative spillovers.

### Quality Estimation: Econometric Approach and Implementation
- Estimating equation (as in Khandelwal (2010), updated):
  - Equation form presented in source: llol(s_cht) - llol(s_0t) = λ1,ch + λ2,t + α p_cht + σ lls_cht + γ ln(ppop_ct) + λ3,cht, with variables defined as:
    - s_cht: market share of variety (HS10 product from country c at time t) in the industry.
    - s_0t: market share of the domestic variety.
    - λ1,ch: variety-specific fixed effect.
    - λ2,t: time fixed effect.
    - p_cht: price (unit values inclusive of cost of insurance and freight and tariff duties).
    - lls_cht: market share of the variety within the HS10 product.
    - ppop_ct: population of the country producing the variety.
    - λ3,cht: unobserved variety-time error.
  - Quality defined as the expectation of the relative market share conditional on price, nested share and population:
    - λ_ch_t ≡ λ̂1,ch + λ̂2,t + λ̂3,cht.
- Data and instruments:
  - U.S. 10-digit trade data from USA Trade Online for 2002-2018.
  - Population data from the World Bank World Development Indicators.
  - Industry-level domestic shares from NBER domestic shipments data.
  - Price is endogenous and instrumented using variety-specific unit transportation costs (from the trade dataset), exchange rates (World Economic Outlook database), and oil prices (Haver).
  - lls_cht instrumented with the number of varieties within the product and the number of varieties exported by the partner.
- Percentile benchmarks and loss calculation:
  - When estimating quality for the OECD as a whole, use for each product the 90th percentile of quality across all OECD economies.
  - To calculate potential quality losses if China and the OECD restrict access to each other’s products, assume substitution toward inputs from the rest of the world with quality at the 75th percentile across all non-China, non-OECD economies.
  - Source provides the formal percent-loss formula using normalized qualities and the 75th percentile benchmark.

*Source: REGIONAL ECONOMIC OUTLOOK— Asia and Pacific, INTERNATIONAL MONETARY FUND | October 2023*

---


_Source: https://www.imf.org/-/media/files/publications/reo/apd/2023/october/english/ch3annex.pdf_
