## ch1annex ## Source details **Canonical URL:** [ch1annex](https://www.imf.org/-/media/files/publications/weo/2022/october/english/ch1annex.pdf) ## Other formats - [Markdown version](/-/media/files/publications/weo/2022/october/english/ch1annex.pdf.md) - [Structured JSON version](/-/media/files/publications/weo/2022/october/english/ch1annex.pdf.json) --- ### Methods and identification - Primary econometric approaches: - Local Projection Instrumental Variables (LP-IV) for dynamic causal impulse response functions (IRFs). - Structural VAR (SVAR) with zero contemporaneous restrictions to recover structural shocks and cumulative orthogonalized IRFs. - Key instruments and identifying assumptions: - Instruments for three channels in LP-IV: changes in real gas prices; US monetary policy shocks from Jarocinski and Karadi (2020); cereal harvest shocks from De Winne and Peersman (2016). - Harvest shocks constructed as prediction errors from regressions of calorie-weighted rest-of-world harvest quantities (maize, rice, soybeans, wheat) after allocating annual production to quarters using crop and country-specific harvesting calendars. - Shipping-cost instrument: closure events of the Suez Canal used to instrument the Baltic Dry Index (BDI). - SVAR variable ordering and short-term restrictions: top block (delta log global oil supply, global real economic activity indicator, log oil prices) treated as exogenous within the month to bottom-block variables (log gas prices, log fertilizer prices, log cereal prices). Short-term restrictions imply fertilizer prices don’t affect gas prices within the month and cereal prices don’t affect fertilizer prices within the month. - Sample and data frequency: - LP-IV estimation sample for oil/fertilizer/monetary/harvest analysis: 1991Q1 to 2021Q1. All variables deflated by the US GDP deflator. - Monthly triple-difference/speculative analysis: 1980M1-2022M3. - Cross-country food pass-through sample: more than 100 countries, period 1991-2020. - All regressors in P, X and W incorporated as monthly log-differences; lag polynomials up to L=18 for country-level LP-IV pass-through exercise. ### Dynamic causal effects: energy, harvest, and monetary policy shocks on cereal prices - LP-IV specification normalizes unit effect so that a 1% increase in a shock (e.g., monetary policy shock) corresponds to a 1% increase in the 3-month treasury bill rate. - Key IRF findings from SVAR and LP-IV: - A 10% negative oil supply shock leads to an increase in oil prices of 0.4% after 9 months. - Effects of a 10% negative oil supply shock: - On fertilizer prices: mostly statistically insignificant. - On gas prices: peak at 10% after 8 months and stabilize afterwards. - Estimation detail: dynamic effect 훳훳_h of each regressor estimated via classical IV in a 2SLS approach and represented as IRFs (h, 훳훳_h). ### Speculative demand, financialization, and cereal prices - Definition of speculative periods: - Periods where cereal prices and net long positions of non-commercial traders change in the same direction. - Binary dummy S_t equals 1 for speculative periods, 0 otherwise. - Triple-difference specification (monthly 1980M1-2022M3) includes: - Dummy D_t equal to 1 after 2003 (period of intensified financialization of commodities markets). - I_t equal to delta log of the S&P500 index. - Empirical result: - Financialization led the pass-through from the S&P500 to cereal price to be 56 percent higher in speculative periods than in non-speculative periods. - Table A1 (select coefficients and details): - I_t: 0.114 (0.0732) - S_t: 0.00304 (0.00460) - S_t * I_t: -0.188 (0.133) - D_t: 0.0110** (0.00510) - D_t * S_t * I_t: 0.557** (0.239) - _cons: -0.00219 (0.00273) - N = 507 - Note: * 0.10 ** 0.05 *** 0.01 ### Dynamic causal effects of food commodity price shocks on domestic food inflation - Estimation approach: - Country-level lag-augmented LP-IV following Jordà (2005), Montiel Olea and Plagborg‐Møller (2021), De Winne and Peersman (2021). - Endogenous variables X_t: IMF’s food price index and the Baltic Dry Index (BDI) as measures of food commodity prices and shipping costs respectively. - Controls W_i,t (predetermined): exchange rate (LCU/USD), headline inflation, real oil prices. - Country characteristics c_k,i: real income per capita (2010 USD) and trade openness (% of GDP), expressed in standard deviations from the global mean; interactions with food price index included to capture heterogeneity in pass-through. - Identification: - Harvest shocks used as instrument for food commodity prices (per De Winne and Peersman 2021). - Suez Canal closure events used as instrument for BDI (per Carriere-Swallow et al. 2022). - Main pass-through magnitudes (Table A2; month-on-month log-difference coefficients with robust clustered SEs in parentheses): - food price index (mom log-diff): - h=1: 0.009 (0.005) - h=3: 0.104*** (0.015) - h=6: 0.207*** (0.032) - h=9: 0.288*** (0.042) - h=12: 0.302*** (0.046) - h=18: 0.326*** (0.057) - exchange rate (LCU/USD) (mom log-diff): - h=1: -0.040*** (0.006) - h=3: 0.080 (0.049) - h=6: 0.197** (0.071) - h=9: 0.268*** (0.077) - h=12: 0.341*** (0.086) - h=18: 0.354*** (0.091) - Baltic Dry Index (BDI) (mom log-diff): - h=1: -0.003 (0.003) - h=3: -0.019+ (0.011) - h=6: 0.002 (0.020) - h=9: 0.040 (0.027) - h=12: 0.088** (0.030) - h=18: 0.075* (0.037) - headline CPI (mom log-diff): - h=1: 1.234*** (0.056) - h=3: 1.624*** (0.419) - h=6: 1.722** (0.615) - h=9: 1.739* (0.700) - h=12: 1.712* (0.754) - h=18: 1.922* (0.895) - real oil price (USD) (mom log-diff): - h=1: -0.006** (0.002) - h=3: -0.000 (0.006) - h=6: -0.013 (0.010) - h=9: -0.024+ (0.013) - h=12: -0.032* (0.015) - h=18: -0.037+ (0.019) - income per capita × food price index: - h=1: -0.010* (0.004) - h=3: -0.025* (0.011) - h=6: -0.045** (0.016) - h=9: -0.060** (0.020) - h=12: -0.047* (0.023) - h=18: 0.016 (0.028) - trade openness × food price index: - h=1: -0.002 (0.005) - h=3: 0.004 (0.016) - h=6: 0.049+ (0.029) - h=9: 0.069+ (0.037) - h=12: 0.049 (0.044) - h=18: -0.003 (0.059) - Key aggregated pass-through statements and interpretation: - The pass-through from a 1pp food commodity price shock to domestic food prices is about 0.30pp after 12 months. - A 1pp increase in shipping costs (BDI) translates into a 0.09pp increase in domestic food prices after 12 months. - Heterogeneity by income and openness: - A country with income per capita 1 standard deviation below average features a food commodity price pass-through of about 35 percent rather than 29 percent 9 months after the shock (implying larger pass-through for lower-income countries). - For a country with trade openness 1 standard deviation above average, pass-through increases from 29 to 37 percent (9 months after the shock). - Cross-country model fit and sample: - Observations: 13,788 - # of countries: 97 - R-sq by horizon: h=1: 0.793; h=3: 0.496; h=6: 0.394; h=9: 0.348; h=12: 0.304; h=18: 0.262 - Robust standard errors clustered at the country level. Country fixed effects included but not reported. iv = instrumental variables; fe = fixed effects; mom = month-on-month; LCU = local currency units. ### Notable numerical and timing details - LP-IV horizons for cereal/energy channels: h ∈ {0,1,2,3,4} in the channel-specific specifications; country-level pass-through horizons h = 0,1,2,...,18 months. - Unit normalization: 1% shock in an instrumented variable corresponds to a 1% movement in that instrumented variable (e.g., 3-month T-bill rate for monetary shock). - Specific IRF results excerpted: - 10% negative oil supply shock → oil prices +0.4% after 9 months. - Gas prices response to 10% negative oil supply shock: peak at 10% after 8 months. - Financialization timing marker: D_t dummy equals 1 after 2003 to capture the period when asset managers substantially increased involvement in commodities futures and options. *Source: ch1annex (CHAPTER 1 COMMODITY SPECIAL FEATURE), International Monetary Fund | October 2022.* --- _Source: https://www.imf.org/-/media/files/publications/weo/2022/october/english/ch1annex.pdf_