## wpiea2021012-print-pdf — Sections 3.1, 3.2 and 4.2

## Source details

**Canonical URL:** [wpiea2021012-print-pdf — Sections 3.1, 3.2 and 4.2](https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021012-print-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2021/english/wpiea2021012-print-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2021/english/wpiea2021012-print-pdf.pdf.json)

---

### Data and sample
- Sample and outcome variable
  - Sample: 70 low and middle-income countries.
  - Period: 1980 -2012.
  - Outcome variable: household consumption expenditure per capita (World Bank’s World Development Indicators dataset).
  - Rationale: food consumption expenditure data unavailable or inseparable from total household consumption; household consumption per capita is less prone to measurability issues and households in poorer countries tend to spend a large share of their budget on food.
  - Illustrative food budget shares (World Economic Forum 2016): Algeria 40.1%, Azerbaijan 42.5%, Cameroon 45.6%, Nigeria 56.4%, Philippines 40.9%; Australia 9.8%, Canada 9.1%, Singapore 6.7%, US 6.4%.
- Imported food price index: construction and components
  - Commodities included: wheat, sugar, soybeans, soybeans oil, maize, and rice.
  - Weights: contemporaneous prices weighted by period-average imported quantities to avoid endogeneity (equation (1) and equation (2)).
  - Base year: 1995.
  - Country-year price index P_{i,t} constructed as product over j of P_{j,t}^{w_{i,j•}}.
  - Note: detailed data on share of volume of each commodity in each country’s basket were not available.
- Price shocks: indicators and methodology
  - Country-by-country regression (equation (3)): ln(P_{i,t}) = α_{i,0} + α_{i,1} t + α_{i,2} * sum_{p=1 to 3} θ_{i,p} ln(P_{i,t−p}) + ε_{i,t}; residual ε_{i,t} extracted.
  - Autocorrelation: controlled via first and second order lags; robustness with one or three lags.
  - Three indicators:
    - Shock: residual ε_{i,t}.
    - Frequency indicator: number of positive residuals during consecutive four non-overlapping years.
    - Filter_Shock: cyclical component from Hodrick-Prescott filter on ln(import price index) with λ = 6.25; linear filter focused on in paper.
  - Note: prices P_{j,t} are world market prices (same for all countries); transmission differs across countries.
- Discretionary government consumption: construction and controls
  - Country-by-country regression (equation (4)): ∆GOV_t = β_0 + β_1 * time + β_2 * ∆GOV_{t−1} + β_3 * Z_t + β_4 * GOV_GAP_t + ε_t, where Z includes Inflation and Inflation Square; cyclical GDP per capita used as proxy for output gap.
  - Bootstrap correction: output gap coefficient corrected with 250 bootstrap replications to address generated-variable bias.
  - Discretionary Gov exp (equation (5)) standardized: Discretionary Gov exp_{i,t} = (ResidGov_{i,t} − ResidGov¯) / σ(ResidGov_{i,t}).
- Other fiscal variable
  - Government subsidies and other transfers included as additional fiscal policy measure; may include transfers to workers, direct subsidies on basic foodstuff, pensions, social safety nets.
  - Caveat: subsidies may also apply to non-food items; disaggregated data not available.

### Stylized facts on import food price shocks and household consumption
- Shock occurrence and timing
  - Import price shocks were frequent over 1980-2010.
  - Upward trend from 2002 to 2011; highest values in 2006-2007 and 2011 (2006-2008 and 2011 surges); falls in 2009 and earlier 2012.
- Correlations and simple regression
  - Shock and Filter_Shock are strongly correlated.
  - Hodrick-Prescott filter parameter used: λ = 6.25 (results unchanged when λ = 150).
  - Pooled regression reported: Consumption = 6.96 - 3.5 Shock (axis: Log(Household Consumption Percapita); Shock).
- Descriptive statistics by household consumption quartiles (summary Table 1)
  - Mean import food price shocks:
    - Lowest household consumption quartile: 0.08
    - Highest household consumption quartile: 0.05
  - Positive shocks (average weight of positive shocks):
    - Lowest quartile: 4.62
    - Highest quartile: 3.55
  - Discretionary government consumption expenditure:
    - Lowest quartile: average negative value (more episodes of unpredicted reduction).
    - Highest quartile: average positive value.
  - Government transfers (share of expense):
    - Lowest quartile: 23%
    - Intermediate quartile: 29%
    - Highest quartile: 34%
  - Remittances per capita:
    - Lowest quartile: 3
    - Other quartiles: 7 and 4 (respectively)
  - Official development assistance (share of GDP):
    - Lowest quartile: 12% of GDP
    - Intermediate quartile: 5
    - Highest quartile: 1
  - Private credit: lowest in lowest household consumption countries on average.
  - Exchange rate regime flexibility: no systematic difference across lowest/highest consumption groups.
- Limits: descriptive patterns are not causal; econometric approach used to identify causal relationships and control variables.

### Econometric framework and identification
- Dynamic panel model (equation (6)) specification
  - Dependent variable: HCONS_{i,t} (household consumption per capita).
  - Key regressors: HCONS_{i,t-1}; Shock_{i,t}; disGovexp_{i,t}; Shock_{i,t} * disGovexp_{i,t}; Z_{i,t} (controls); country and year fixed-effects.
  - Expected signs: α2 (Shock) negative; α4 (Shock * disGovexp) positive.
  - Threshold where shock effect is zero: disGovexp* = -α2 / α4.
- Control variables (from Appendix Table 7): Revenue; Net remittances received (per capita); Net official aid assistance (per capita); Credit to private sector (% of GDP); GDP per capita (log, constant 2011); Trade openness; Exchange rate regime (rank 1–15); data sources WDI, FAO STATISTIQUES, IMF WEO.
- Estimation strategy
  - Nickel bias from lagged dependent variable addressed using system-GMM (Blundell and Bond, 1998) with instruments in levels and differences.
  - Import food price shocks and discretionary government consumption treated as exogenous.
  - Model validation: over-identification tests, first-order serial independence, absence of second-order serial correlation.
  - Practical choices: period averaged over non-overlapping 4-year blocks to improve consistency; limit instrument count per Roodman (2009).

### Estimation results — baseline (two-step system-GMM, Windmeijer-corrected)
- Data averaged over eight non-overlapping 4-year periods (1980–2011).
- System-GMM diagnostics (selected, Table 2):
  - Number of Observation: 400, 368, 294, 294 (columns 1–4).
  - Number of Country: 70, 68, 66, 66.
  - Ar1(p-value): 0.0758, 0.0515, 0.0541, 0.0645.
  - Ar2(p-value): 0.370, 0.108, 0.128, 0.162.
  - Hansen(p-value): 0.356, 0.373, 0.265, 0.167.
- Key estimated coefficients (Table 2, selected)
  - L.Log(Household Consumption Per capita): 0.899***; 0.830***; 0.843***; 0.879***.
  - Shock (import food price shock): -1.848**; -2.685***; -2.010*** (columns 1–3).
  - Discretionary Gov Expenditure: -0.161; -2.747; -2.442; -2.739*.
  - Shock * Discretionary Gov Expenditure: 29.06**; 37.86**; 22.04*.
  - Private Credit: 0.205*; 0.336**; 0.233; 0.221.
  - Positive Shock (frequency): -0.0737* (column 4).
  - Positive Shock * Discretionary Gov Expenditure: 1.180* (column 4).
- Interpretation and thresholds
  - Import food price shocks significantly and negatively affect household consumption.
  - Interaction Shock * Discretionary Gov Expenditure positive and significant: discretionary spending during positive import food price shocks mitigates the negative effect on household consumption.
  - Estimated discretionary policy threshold referenced: discretionary policy amounts to 0.06- 0.09 points (deviation of government consumption expenditure from long-term trend) are effective; table gives threshold around 0.07 points for African subsamples.
  - Financial deepening: private credit increases household consumption in most regressions.
  - Remittances noted to positively affect household consumption.

### Geographic heterogeneity (Africa, SSA, non-Africa)
- Subsample diagnostics (Table 3, selected)
  - Number of Observation (columns 1–7): 112; 139; 155; 118; 192; 118; 147.
  - Number of Country: 26; 33; 33; 27; 43; 27; 35.
  - Ar2(p-value): 0.0623; 0.0408; 0.0190; 0.0738; 0.00662; 0.0520; 0.0749.
  - Hansen(p-value): 0.366; 0.160; 0.918; 0.923; 0.276; 0.121; 0.682.
- Selected coefficients and findings
  - L.Log(Household Consumption Per capita): 0.952***; 0.907***; 0.757***; 0.900***; 0.707***; 0.895***; 0.906***.
  - Shock: -1.844*; -2.031***; -1.807* (negative effects in SSA and Africa).
  - Shock * Discretionary Gov Expenditure: 24.56**; 23.25*; -10.66 (positive and significant in SSA and Africa; negative and not significant in others).
  - Filter Shock: -6.097**; Filter Shock * Discretionary Gov Expenditure: 91.99* (in HP-filtered specification).
  - Positive Shock: -0.144***; -0.160*.
  - Discretionary Gov Expenditure_Positive Shock: 2.096**; 2.096**.
- Interpretation
  - Import food price shocks significantly decrease household consumption in SSA and African samples; negative effect dampened by discretionary government expenditure around 0.07 deviation points.
  - In non-African developing countries, shocks negatively affect consumption but interactive fiscal effect not statistically significant.
  - Consistent results across shock measures (frequency and HP-filtered).

### Robustness: government transfers (subsidies and other transfers)
- Motivation: fiscal mitigation may operate through transfers/subsidies; government transfers (% of expense) used as alternative fiscal variable.
- System-GMM diagnostics (Table 4):
  - Number of Observation: 172; 172; 202 (columns 1–3).
  - Number of Country: 49; 49; 60.
  - Ar1(p-value): 0.0916; 0.0544; 0.0950.
  - Ar2(p-value): 0.177; 0.0651; 0.116.
  - Hansen(p-value): 0.264; 0.852; 0.650.
- Key coefficients (Table 4)
  - L.Log(Household Consumption Per capita): 0.789***; 0.330**; 0.573***.
  - Shock: -1.512* (column 1).
  - Government Transfers: 0.00377; -0.00335; 0.00581.
  - Shock * Government Transfers: 0.0172** (column 1).
  - Positive Shock: -0.0235** (column 2).
  - Positive Shock * Government Transfers: 0.00106*** (column 2).
  - Filter Shock: -2.084** (column 3).
  - Filter Shock * Government Transfers: 0.0445** (column 3).
- Interpretation
  - Government transfers and subsidies mitigate the negative effect of food price shocks on household consumption: positive and significant interaction terms across specifications.
  - Data limitation: transfers could not be disaggregated to food-specific subsidies; adding food production does not change results but food production estimates inconsistent due to missing data.

### Heterogeneity by exchange rate regime
- Exchange regime split: less flexible regimes (rank 1 to 6) vs more flexible regimes (rank 7 to 15); data recorded 1984–2010.
- Diagnostics and samples (Table 5):
  - Number of Observation: 153; 101; 133; 82 (columns shown).
  - Number of Country: 48; 39; 42; 30.
  - Ar1(p-value): 0.0662; 0.0799; 0.0878; 0.0458.
  - Ar2(p-value): 0.380; 0.188; 0.122; 0.896.
  - Hansen(p-value): 0.673; 0.770; 0.611; 0.668.
- Selected coefficients and findings (Table 5)
  - L.Log(Household Consumption Per capita): 0.853***; 0.855***; 0.599**; 0.952***.
  - Shock: -1.156** (fixed regime sample); -1.762 (another fixed-flexible split); 0.539 (flexible regime sample, not significant).
  - Shock * Discretionary Gov Expenditure: 9.762** (fixed regime); 4.559 (not significant in flexible regime).
  - Positive Shock: -0.150* (in fixed-regime specification).
  - Positive Shock * Government Transfers: 0.00451** (mitigating effect significant in less flexible regimes).
  - Trade Openness: 0.467*** (in one flexible-regime specification).
- Interpretation
  - Import food price shocks affect household consumption primarily in less flexible (more fixed) exchange rate regimes.
  - Dampening effects of discretionary government consumption and transfers are significant in less flexible regimes, consistent with larger fiscal multipliers under fixed exchange rate regimes.

### Conclusion and policy implications
- Main findings
  - Import food price shocks negatively and significantly impact household consumption in the pooled sample and especially in African countries and countries with less flexible exchange rate regimes.
  - Discretionary government consumption expenditure and government transfers/subsidies mitigate the adverse effects of positive import food price shocks on household consumption.
  - Financial deepening (private credit) is associated with higher household consumption in many specifications.
- Policy lessons
  - Discretionary government expenditure can be an effective tool to stabilize household consumption during food price shocks.
  - Fiscal consolidation should consider composition: subsidies and transfers can be important for short-run shock absorption and should be targeted to those in need.
  - Measures that reduce household consumption vulnerability to import food price shocks (including financial deepening) could substitute for fiscal interventions and help lower fiscal deficits.
- Caveats and limitations
  - Results subject to data limitations: sample sizes, missing data for food production, inability to separate subsidies from other transfers.
  - Some standard error concerns due to extracted explanatory variables; bootstrap not feasible with available observations.
  - Use of averaged shocks (sub-periods) driven by GMM concerns with yearly data.

*Source: wpiea2021012-print-pdf — Sections 3.1, 3.2 and 4.2.*

### 3.1 Data and Sample _____________________________________________________ 7

### 3.1 Data and Sample

### Sample and outcome variable
- Sample: 70 low and middle-income countries.
- Period: 1980 -2012.
- Outcome variable: household consumption expenditure per capita (World Bank’s World Development Indicators dataset).
- Rationale: food consumption expenditure data were unavailable or not separable from total household consumption; household consumption per capita is less prone to measurability issues and households in poorer countries tend to spend a large share of their budget on food.
- Illustrative food budget shares (World Economic Forum 2016): Algeria 40.1%, Azerbaijan 42.5%, Cameroon 45.6%, Nigeria 56.4%, Philippines 40.9%; Australia 9.8%, Canada 9.1%, Singapore 6.7%, US 6.4%.

### Imported food price index: construction and components
- Price index constructed per country using the most common imported food commodities (joint report by FAO, IMF, and UNCTAD (2011)); most commodities are cereals.
- Commodities included: wheat, sugar, soybeans, soybeans oil, maize, and rice.
- Weights: contemporaneous prices are weighted by period-average imported quantities to avoid endogeneity (see equation (1) and equation (2)).
- Base year: 1995.
- Equation summary:
  - Equation (1): definition of relative weight w_{i,j•} using P_{j•} and Q_{i,j•} with summation j = 1 to 6.
  - Equation (2): country-year price index P_{i,t} = product over j of P_{j,t}^{w_{i,j•}}.
- Note: detailed data on share of volume of each commodity in each country’s basket were not available.

### Price shocks: three indicators and methodology
- Regression to extract unpredictable residuals (equation (3)):
  - ln(P_{i,t}) = α_{i,0} + α_{i,1} t + α_{i,2} * sum_{p=1 to 3} θ_{i,p} ln(P_{i,t−p}) + ε_{i,t}.
  - Regressions run separately for each country to account for differential transmission.
  - Autocorrelation controlled by including first and second order lags; robustness noted when using one or three lags.
- Three indicators constructed:
  - Shock: residual ε_{i,t} from equation (3).
  - Frequency indicator: number of positive residuals of equation (3) during consecutive four non-overlapping years (proxy for frequency of positive food price shocks).
  - Filter_Shock: cyclical component from Hodrick-Prescott filter applied to ln(import price index); smoothing parameter λ = 6.25 (Ravn and Uhlig (2002)). The linear filter shock displayed the same variation as Shock; paper focuses on the linear filter.
- Note: prices used are world market prices (same P_{j,t} for all countries) and transmission differs across countries.

### Discretionary government consumption: construction and controls
- Methodology: follows Fatas and Mihov (2003) and Agnello and Sousa (2009); regressions run country-by-country to compute the discretionary (exogenous) component of government consumption.
- Regression (equation (4)):
  - ∆GOV_t = β_0 + β_1 * time + β_2 * ∆GOV_{t−1} + β_3 * Z_t + β_4 * GOV_GAP_t + ε_t.
  - ∆GOV is government expenditure as a share of GDP in t minus its value in t−1.
  - Z includes Inflation and Inflation Square; cyclical component of GDP per capita is used as a proxy for the output gap.
- Bootstrap correction: output gap coefficient corrected using a bootstrap process with 250 replications in country-by-country regressions to address generated-variable bias.
- Discretionary Gov exp (equation (5)): standardized Resid Gov
  - Discretionary Gov exp_{i,t} = (ResidGov_{i,t} − ResidGov¯) / σ(ResidGov_{i,t}), where Resid Gov¯ is the yearly average of Resid Gov for all countries in the sample and σ is the standard error of Resid Gov for the entire sample.

### Other fiscal variable
- Government subsidies and other transfers are included as an additional fiscal policy measure (may include transfers to workers, direct subsidies on basic foodstuff, pensions, social safety nets).
- Caveat: subsidies may also apply to non-food items (fuel, energy); disaggregated data were not available. The paper assumes food-related subsidies are an important share of total subsidies and transfers.

*Source: wpiea2021012-print-pdf — 3.1 Data and Sample.*

### 3.2 Stylized Facts

### 3.2 Stylized Facts

### Patterns of import food price shocks
- The two import price shock variables (Shock and Filter_Shock) are strongly correlated (Figure 1).
- Cyclical component of GDP per capita is obtained using the Hodrick-Prescott filter with a filter parameter λ of 6.25. Results did not change when λ=150; the study uses λ=6.25.
- Frequency and timing (observed period references):
  - Import price shocks were frequent over the period 1980-2010.
  - An upward trend is observed from 2002 to 2011.
  - Highest values in 2006-2007 and then in 2011 (materializing the 2006-2008, and 2011 surges).
  - Falls observed in 2009 and in earlier 2012.
- Figure 2 (pooled time series cross section for developing countries, period 1980-2012) shows household expenditure per capita is negatively correlated with import food price shocks.
  - Regression reported on the figure: Consumption = 6.96 - 3.5 Shock
  - Axis/labeling: Log(Household Consumption Percapita); Shock; Average Values

### Descriptive statistics by household consumption quartiles (summary from Table 1, Appendix)
- Mean import food price shocks:
  - Lowest household consumption quartile: 0.08
  - Highest household consumption quartile: 0.05
- Positive shocks (average weight of positive shocks):
  - Lowest quartile: 4.62
  - Highest quartile: 3.55
- Discretionary government consumption expenditure:
  - Average negative value for the lowest household consumption quartile (indicating more episodes of unpredicted reduction in government consumption expenditures).
  - Average positive value for the highest household consumption quartiles.
- Government transfers and subsidies (share of expense):
  - Lowest quartile: 23%
  - Intermediate quartile: 29%
  - Highest quartile: 34%
- Remittances per capita:
  - Lowest quartile: 3
  - Other quartiles: 7 and 4 (respectively)
- Official development assistance (share of GDP):
  - Lowest quartile: 12% of GDP
  - Intermediate quartile: 5
  - Highest quartile: 1
- Financial private credit:
  - On average, lowest household consumption countries have the lowest level of financial private credit.
- Exchange rate regime flexibility:
  - No evidence that the lowest or highest household consumption groups systematically have more or less flexible exchange rate regimes.

### Interpretation and limits of descriptive patterns
- The descriptive patterns are not causal; the paper proceeds to an econometric approach to investigate causal relationships and control for other variables.

### Econometric framework: model specification
- Objective: identify the dampening effect (if any) of fiscal policy on household consumption in times of food price shocks.
- Fiscal policy variable: discretionary component of government consumption expenditure; robustness checks use government transfers (as a percentage of expense).
- Dynamic panel model (equation (6)):
  - Dependent variable: HCONS (household consumption per capita for country i at period t).
  - Key regressors:
    - HCONS_{i,t-1} (first lag of consumption expenditures)
    - Shock_{i,t} (import food price shocks)
    - disGovexp_{i,t} (discretionary government consumption)
    - Interaction: Shock_{i,t} * disGovexp_{i,t}
    - Z_{i,t} (vector of control variables)
    - d_i (country fixed-effects)
    - w_t (year fixed-effects)
    - ε_{i,t} (error term)
  - Expectations:
    - α2 (coefficient on Shock) expected to be negative.
    - α4 (coefficient on Shock * disGovexp) expected to be positive.
- Threshold at which the negative effect of shocks on household consumption is completely cleared (from equation (7)):
  - ∂(HCONS)_{i,t} / ∂(SHOCK)_{i,t} = α2 + α4 * disGovexp_{i,t} = 0
  - => disGovexp* = -α2 / α4

### Control variables (definitions referenced from Appendix table 7)
- Revenue (total fiscal revenue)
- Net remittances received (per capita)
- Net official aid assistance (per capita)
- Credit to the private sector (ratio of private credit over GDP)
- GDP per capita (constant 2011, measured in logarithm)
- Trade openness (ratio of imports and exports over GDP)
- Exchange rate regime (Reinhart and Rogoff (2004) classification ranked 1 to 15, from most fixed to most flexible)
- Data sources: WDI, FAO STATISTIQUES, IMF World Economic Outlook databases

### Identification and estimation strategy
- OLS with country fixed-effects may be biased due to correlation between the lagged dependent variable and the error term (Nickel bias; Nickell 1981), especially in small samples.
- Estimation approach:
  - Use GMM approach developed by Blundell and Bond (1998).
  - System-GMM is preferred: instrument variables in first difference with those in level and vice versa.
  - Import food price shocks and discretionary components of government consumption are considered exogenous in the model.
- Model validation criteria:
  - Rejection of the over-identification hypothesis
  - Presence of first-order serial independence
  - Absence of second-order serial correlation (Arellano, 2003; Arellano and Bond, 1991)
- Practical estimation choices and robustness:
  - The 33 years period is split into sub-periods of 4 years to obtain more consistent and asymptotically efficient estimators and to neutralize high-frequency noise.
  - The issue of too many instruments in system-GMM is addressed by limiting the fixed number of lags (Roodman, 2009).

*Source: wpiea2021012-print-pdf - 3.2 Stylized Facts*

### 4.2 Estimation Results

### 4.2 Estimation Results

### Baseline results: Household consumption, import food price shocks and fiscal policy
- Estimation method: two-step system-GMM with the Windmeijer correction (Windmeijer, 2005). Data averaged over eight non-overlapping 4-years periods during 1980–2011.
- System-GMM diagnostics (selected regressions, Table 2):
  - Number of Observation: 400, 368, 294, 294 (columns 1–4).
  - Number of Country: 70, 68, 66, 66.
  - Ar1(p-value): 0.0758, 0.0515, 0.0541, 0.0645.
  - Ar2(p-value): 0.370, 0.108, 0.128, 0.162.
  - Hansen(p-value): 0.356, 0.373, 0.265, 0.167.
- Key estimated coefficients (Table 2):
  - L.Log(Household Consumption Per capita): 0.899***; 0.830***; 0.843***; 0.879*** (t-statistics reported).
  - Shock (import food price shock): -1.848**; -2.685***; -2.010*** (columns 1–3).
  - Discretionary Gov Expenditure: -0.161; -2.747; -2.442; -2.739*.
  - Shock * Discretionary Gov Expenditure: 29.06**; 37.86**; 22.04*.
  - Private Credit: 0.205*; 0.336**; 0.233; 0.221.
  - Positive Shock (frequency): -0.0737* (column 4).
  - Positive Shock * Discretionary Gov Expenditure: 1.180* (column 4).
- Interpretation:
  - Import food price shocks significantly and negatively affect household expenditure.
  - Discretionary government consumption expenditure interacts positively with shocks: discretionary spending involved during positive import food price shocks mitigates the negative effect on household consumption.
  - Estimated discretionary policy threshold noted in text: discretionary policy amounts to 0.06- 0.09 points (deviation of government consumption expenditure from its long-term trend) are referenced as effective; table gives a threshold around 0.07 points for African subsamples.
  - Financial deepening: increase of financial credit to the private sector strongly and positively increases household consumption in most regressions.
  - Remittances appear to positively affect household consumption (noted in text).

### Geographic heterogeneity: Africa versus non-Africa and SSA
- Subsample estimations (Table 3) — selected diagnostics and coefficients:
  - Number of Observation (columns 1–7): 112; 139; 155; 118; 192; 118; 147.
  - Number of Country: 26; 33; 33; 27; 43; 27; 35.
  - Ar2(p-value) (columns 1–7): 0.0623; 0.0408; 0.0190; 0.0738; 0.00662; 0.0520; 0.0749.
  - Hansen(p-value) (columns 1–7): 0.366; 0.160; 0.918; 0.923; 0.276; 0.121; 0.682.
- Key estimated coefficients (selected):
  - L.Log(Household Consumption Per capita): 0.952***; 0.907***; 0.757***; 0.900***; 0.707***; 0.895***; 0.906***.
  - Shock: -1.844*; -2.031***; -1.807* (columns 1–3; SSA and Africa showing negative effects).
  - Shock * Discretionary Gov Expenditure: 24.56**; 23.25*; -10.66 (positive and significant in SSA and Africa; negative and not significant in others).
  - Filter Shock: -6.097** (in a column using HP filter).
  - Filter Shock * Discretionary Gov Expenditure: 91.99* (in the same filtered specification).
  - Positive Shock (where used): -0.144***; -0.160* (negative effects on household consumption).
  - Discretionary Gov Expenditure_Positive Shock: 2.096**; 2.096** (mitigating interaction in positive-shock specifications).
- Interpretation:
  - Import food price shocks significantly decrease household consumption in SSA and African samples. The negative effect is dampened by discretionary government expenditure around a deviation of 0.07 points from long-term trend.
  - In non-African (other) developing countries, shocks negatively affect consumption but the interactive effect with fiscal policy is not statistically significant.
  - Results are consistent across alternative shock measures (frequency of positive shocks and Hodrick-Prescott filtered shocks).

### Robustness check: Government transfers (subsidies and other transfers)
- Motivation: fiscal policy may operate directly through subsidies and transfers; government transfers (share of GDP) used as alternative fiscal policy variable.
- System-GMM diagnostics (Table 4):
  - Number of Observation: 172; 172; 202 (columns 1–3).
  - Number of Country: 49; 49; 60.
  - Ar1(p-value): 0.0916; 0.0544; 0.0950.
  - Ar2(p-value): 0.177; 0.0651; 0.116.
  - Hansen(p-value): 0.264; 0.852; 0.650.
- Key estimated coefficients (Table 4):
  - L.Log(Household Consumption Per capita): 0.789***; 0.330**; 0.573***.
  - Shock: -1.512* (column 1).
  - Government Transfers: 0.00377; -0.00335; 0.00581.
  - Shock * Government Transfers: 0.0172** (column 1).
  - Positive Shock: -0.0235** (column 2).
  - Positive Shock * Government Transfers: 0.00106*** (column 2).
  - Filter Shock: -2.084** (column 3).
  - Filter Shock * Government Transfers: 0.0445** (column 3).
- Interpretation:
  - Government subsidies and transfers mitigate the negative effect of food price shocks on household consumption: interaction terms are positive and significant across specifications.
  - Positive import price shocks reduce household consumption; government transfers play a mitigating role.
  - Data limitation: government subsidies and other transfers could not be separated; results unchanged when adding food production but food production estimates not consistent due to missing data.

### Heterogeneity by exchange rate regime
- Exchange regime split: less flexible exchange rate regimes (ranked 1 to 6) vs more flexible regimes (ranked 7 to 15). Exchange regime data recorded from 1984 to 2010.
- Diagnostics and samples (Table 5):
  - Number of Observation (columns shown): 153; 101; 133; 82.
  - Number of Country: 48; 39; 42; 30.
  - Ar1(p-value): 0.0662; 0.0799; 0.0878; 0.0458.
  - Ar2(p-value): 0.380; 0.188; 0.122; 0.896.
  - Hansen(p-value): 0.673; 0.770; 0.611; 0.668.
- Key estimated coefficients (Table 5):
  - L.Log(Household Consumption Per capita): 0.853***; 0.855***; 0.599**; 0.952***.
  - Shock: -1.156** (fixed regime sample); -1.762 (another fixed-flexible split); 0.539 (flexible regime sample, not significant).
  - Shock * Discretionary Gov Expenditure: 9.762** (fixed regime); 4.559 (not significant in flexible regime).
  - Positive Shock: -0.150* (in some fixed-regime specification).
  - Positive Shock * Government Transfers: 0.00451** (mitigating effect significant in less flexible regimes).
  - Trade Openness: 0.467*** (in one flexible-regime specification).
- Interpretation:
  - Import food price shocks affect household consumption primarily in countries with less flexible (more fixed) exchange rate regimes.
  - Dampening effects of discretionary government consumption expenditure and government transfers on household consumption during import price shocks are significant in countries with less flexible exchange rate regimes, consistent with literature that fiscal multipliers are larger under fixed exchange rate regimes.

### Conclusion and policy implications
- Main findings:
  - Import food price shocks negatively and significantly impact household consumption in the pooled sample and especially in African countries and countries with less flexible exchange rate regimes.
  - Discretionary government consumption expenditure and government transfers/subsidies mitigate the adverse effects of positive import food price shocks on household consumption.
  - Financial deepening (private credit) is associated with higher household consumption in many specifications.
- Policy lessons highlighted in text:
  - Policymakers should consider discretionary government expenditure as a tool to stabilize household consumption in bad times.
  - Fiscal consolidation should account for composition: government subsidies and transfers can be important for short-run shock absorption and should be targeted to those in need.
  - Measures that reduce household consumption vulnerability to import food price shocks (including financial deepening) could substitute for fiscal interventions and help lower fiscal deficits.
- Caveats and limitations noted in the source:
  - Results should be taken with caution due to data limitations (sample sizes, missing data for food production, inability to separate subsidies from other transfers).
  - Some standard error concerns due to extracted explanatory variables; bootstrap not feasible with available observations.
  - The paper works on averaged shocks using sub-periods because GMM inconsistency with yearly data motivates working on average shocks.

*Source: 4.2 Estimation Results (from wpiea2021012-print-pdf)*

### REFERENCES

### wpiea2021012-print-pdf - REFERENCES

### Bibliographic scope
- Bibliography includes empirical and theoretical literature on public deficits, consumption, fiscal policy, food prices, social protection, panel data econometrics, fiscal multipliers, and related topics.
- Key cited authors and works include: Agnello & Sousa (2009); Alem & Soderbom (2012); Schclarek (2004); Arellano (2003); Arellano & Bond (1991); Arezki & Bruckner (2011); Bacchetta & Gerlach (1997); Besley & Kanbur (1988); Bhalla et al. (2018); Blank, Devereux & Handa (2010); Blundell & Bond (1998); Carmignani (2010); Chambers (1989); Clark (1992); Collier & Dehn (2001); Combes et al. (2014); Deaton (1989); Deaton, Miller, & al. (1995); del Granado, Coady, & Gillingham (2012); Essers (2013); FAO et al. (2011); Fatás & Mihov (2003, 2013); Feldstein (1982); Frankel, Vegh, & Vuletin (2013); Furceri & Zdzienicka (2012); Green et al. (2013); Ilzetzki & Vegh (2008); Ilzetzki, Mendoza, & Végh (2013); Lawlor et al. (2019); Minot et al. (2009); Mitchell (2008); Nasreen & Anwar (2014); Nickell (1981); Reinhart & Rogoff (2004); Roodman (2009); Seale et al. (2003); Watson (2013); Windmeijer (2005); Wooldridge (2002); Zhu et al. (2014).

### Appendix — Figure 4 (Fiscal Policy by Geographic Area)
- Note: On the y-axis of the figure at the left, gov-expenditure is the discretionary component of government consumption expenditure.

### Appendix — Table 1. Summary Statistics (by household consumption quartile)
- First quartile of Household Consumption (Obs where provided)
  - Log(Household Consumption): Obs 196, Mean 5,69, Std. Dev. 0,31, Min 4,70, Max 6,12
  - Shock: Obs 171, Mean 0,08, Std. Dev. 0,10, Min -0,15, Max 0,32
  - Positive Shock: Obs 171, Mean 2,10, Std. Dev. 1,33, Min 0,00, Max 4,00
  - Filter Shock: Obs 196, Mean 0,00, Std. Dev. 0,04, Min -0,09, Max 0,08
  - Discretionary Gov Expenditure: Obs 189, Mean -0,06, Std. Dev. 1,51, Min -20,72, Max 0,16
  - Government Transfers: Obs 61, Mean 23,72, Std. Dev. 15,80, Min 1,48, Max 65,04
  - Private Credit: Obs 182, Mean 15,99, Std. Dev. 14,15, Min 1,77, Max 100,81
  - GDP Per capita (Log): Obs 196, Mean 11,03, Std. Dev. 2,64, Min 2,40, Max 16,56
  - Remittances: Obs 161, Mean 3,28, Std. Dev. 5,34, Min 0,01, Max 43,93
  - Official Dev Assistance: Obs 190, Mean 12,35, Std. Dev. 11,62, Min 0,15, Max 90,16
  - Fiscal Revenue: Obs 159, Mean 17,99, Std. Dev. 6,82, Min 1,19, Max 44,16
  - Trade Openness: Obs 182, Mean 16,58, Std. Dev. 25,20, Min -41,49, Max 171,64
  - Exchange Rate Regime: Obs 175, Mean 7,17, Std. Dev. 4,09, Min 2, Max 15

- Second quartile of Household Consumption
  - Log(Household Consumption): Obs 196, Mean 6,68, Std. Dev. 0,34, Min 6,12, Max 7,32
  - Shock: Obs 164, Mean 0,07, Std. Dev. 0,09, Min -0,13, Max 0,32
  - Positive Shock: Obs 164, Mean 1,95, Std. Dev. 1,23, Min 0,00, Max 4,00
  - Filter Shock: Obs 196, Mean 0,00, Std. Dev. 0,04, Min -0,08, Max 0,07
  - Discretionary Gov Expenditure: Obs 193, Mean 0,05, Std. Dev. 0,15, Min -1,32, Max 0,76
  - Government Transfers: Obs 100, Mean 29,53, Std. Dev. 16,58, Min 2,28, Max 66,75
  - Private Credit: Obs 189, Mean 30,09, Std. Dev. 24,29, Min 2,43, Max 148,48
  - GDP Per capita (Log): Obs 189, Mean 11,09, Std. Dev. 2,68, Min 5,18, Max 17,18
  - Remittances: Obs 176, Mean 7,26, Std. Dev. 13,71, Min 0,01, Max 95,09
  - Official Dev Assistance: Obs 184, Mean 5,60, Std. Dev. 6,23, Min 0,01, Max 43,87
  - Fiscal Revenue: Obs 140, Mean 25,42, Std. Dev. 9,35, Min 10,82, Max 59,77
  - Trade Openness: Obs 184, Mean 13,42, Std. Dev. 14,85, Min -39,92, Max 69,34
  - Exchange Rate Regime: Obs 190, Mean 6,91, Std. Dev. 3,69, Min 1, Max 15

- Third quartile of Household Consumption
  - Log(Household Consumption): Obs 196, Mean 7,78, Std. Dev. 0,33, Min 7,33, Max 8,64
  - Shock: Obs 180, Mean 0,05, Std. Dev. 0,09, Min -0,18, Max 0,27
  - Positive Shock: Obs 180, Mean 1,76, Std. Dev. 1,23, Min 0,00, Max 4,00
  - Filter Shock: Obs 196, Mean 0,00, Std. Dev. 0,03, Min -0,06, Max 0,06
  - Discretionary Gov Expenditure: Obs 194, Mean 0,03, Std. Dev. 0,26, Min -3,36, Max 0,26
  - Government Transfers: Obs 105, Mean 34,00, Std. Dev. 15,18, Min 6,27, Max 69,51
  - Private Credit: Obs 182, Mean 42,20, Std. Dev. 30,34, Min 3,68, Max 151,92
  - GDP Per capita (Log): Obs 184, Mean 10,54, Std. Dev. 2,58, Min 6,70, Max 17,13
  - Remittances: Obs 171, Mean 4,08, Std. Dev. 5,79, Min 0,00, Max 29,23
  - Official Dev Assistance: Obs 182, Mean 1,80, Std. Dev. 2,95, Min -0,07, Max 26,01
  - Fiscal Revenue: Obs 137, Mean 24,79, Std. Dev. 8,48, Min 5,49, Max 59,92
  - Trade Openness: Obs 182, Mean 11,02, Std. Dev. 11,16, Min -21,11, Max 50,67
  - Exchange Rate Regime: Obs 179, Mean 7,32, Std. Dev. 4,46, Min 1, Max 15

### Appendix — Table 6. Countries and Number Of Observations
- Africa
  - Algeria (6), Benin (6), Botswana (3), Burkina Faso (7), Burundi (2), Cameroon (3), Chad (1), Comoros (2), Congo (5), Gabon (5), Gambia (3), Guinea-Bissau (1), Kenya (7), Lesotho (7), Liberia (2), Madagascar (7), Malawi (2), Mali (3), Mauritania (1), Mauritius (3), Morocco (6), Mozambique (6), Namibia (6), Nigeria (3), Rwanda (4), Senegal (5), Sierra Leone (3), South Africa (3), Swaziland (6), Tanzania (5), Togo (6), Tunisia (6), Uganda (4)
- Sub-Saharan Africa (repeat/overlap listing as given)
  - Benin (6), Botswana (3), Burkina Faso (7), Burundi (2), Cameroon (3), Congo (5), Gambia (3), Guinea-Bissau (1), Kenya (7), Lesotho (7), Liberia (2), Madagascar (7), (2), Mali (3), Mauritius (3), Mozambique (6), Namibia (6), Nigeria (3), Rwanda (4), Senegal (5), Sierra Leone (3), South Africa (3), Sudan (6), Swaziland (6), Tanzania (5), Togo (6), Uganda (4)
- Others
  - Albania (3), Armenia (4), Azerbaijan (4), Bangladesh (7), Belarus (2), Belize (4), Bhutan(2), Brazil (4), Cambodia (4), China (4), Colombia (7), Costa Rica (6), Dominican Republic (4), Ecuador (5), Egypt (4), El Salvador (6), Georgia (4), Guatemala (5), Honduras (6), India (6), Indonesia (7), Iran (5), Jordan (7), Kazakhstan (5), Kyrgyz Republic (5), Laos (2), Lebanon (3), Macedonia (4), Malaysia (6), Mexico (6), Moldova (4), Mongolia (1), Nicaragua (3), Pakistan (6), Panama (7), Paraguay (5), Peru (3), Philippines (6), Tajikistan (3), Thailand (5), Turkey (3), Ukraine (2), Vietnam (3)
- Note: Number of observations in the parentheses.

### Appendix — Table 7. Variables, Definitions And Sources
- Price shocks
  - Definition: see section (3)
  - Sources: WEO (2015), FAOSTATISTIQUES (2016)
- Household Consumption (Per capita)
  - Definition: Total household consumption expenditure as % of GDP
  - Source: WDI (2015)
- Government Transfers
  - Definition: Government subsidies, grants, transfers, social security and assistances, and employer social benefits (as% of expense)
  - Source: WDI (2015)
- GOV
  - Definition: Government consumption expenditure, include all government current expenditure to purchase goods and services (including compensation of employees but excluding military expenditure) as % of GDP
  - Source: WDI (2015)
- Discretionary Gov Expenditure
  - Definition: Current government expenditure (see section 2)
  - Sources: WDI (2015) and author’s calculations
- Exchange Rate Regime
  - Definition: Categorical variable ranking from 1(lowest flexibility) to 15 (highest flexibility)
  - Sources: (Reinhart and Rogoff 2004; Ilzetzki,Reinhart,and Rogoff 2017)
- GDP Per capita
  - Definition: logarithm of GDP per Capita in constant term 2011
  - Source: WDI (2015)
- Trade Openness
  - Definition: trade openness, measured as the sum of goods and services import and export over the total GDP
  - Source: WDI (2015)
- Private Credit
  - Definition: ratio of the credit addressed by the financial and bank sectors to the private sector in % of GDP
  - Source: WDI (2015)
- Remittances
  - Definition: Net remittances received par individual from abroad in % of GDP
  - Source: WDI (2015)
- Official Development Assistance
  - Definition: Net Official assistance received Per capita
  - Source: WDI (2015)
- Inflation
  - Definition: Inflation Deflator
  - Source: WDI (2015)
- Source: Authors

*wpiea2021012-print-pdf - REFERENCES*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2021/english/wpiea2021012-print-pdf.pdf_
