## _wp10135

## Source details

**Canonical URL:** [_wp10135](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10135.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10135.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10135.pdf.json)

---

### Introduction: recent trends and broad drivers
- GDP-weighted regional inflation fell from 11.5 percent in 1996 to 2.5 percent in 2000, then rose every year until 2008 when it reached double digits.
- In MENACA, inflation increases were broad and steady since 2000; by 2007 almost every country had higher inflation than in 2000.
- Key policy-related observations:
  - Pegs to the US dollar constrained many countries (particularly in the Maghreb) from raising nominal interest rates.
  - Some GCC countries raised rates but “not enough to keep up with inflation.”
  - Maintaining an official or unofficial peg is identified as a key policy decision affecting inflation outcomes.
- Global factors emphasized:
  - Resurgence in inflation coincided with sharp increases in food and energy prices.
  - Decline in the value of the US dollar has been tied to commodity prices and world inflation; a subsequent strengthening of the US dollar could reduce world inflation.
  - Literature cited (summarized): Borio and Filardo; Rogoff; Ball; D’Agostino.

### Background and overview of inflation in MENACA (GDP-weighted series, 1996–2009)
- Year : percent
  - 1996: 11.5
  - 1997: 6.8
  - 1998: 5.4
  - 1999: 4.5
  - 2000: 2.4
  - 2001: 2.7
  - 2002: 3.3
  - 2003: 4.0
  - 2004: 5.0
  - 2005: 5.0
  - 2006: 6.0
  - 2007: 8.6
  - 2008: 13.0
  - 2009: 7.6
- Cross-country examples and patterns:
  - Sudan: 132.8 (1996) → 46.7 (1997) → single digits by 2000.
  - Tajikistan: 418.5 (1996), 88.0 (1997), 43.2 (1998), 27.5 (1999), 32.9 (2000).
  - Kuwait: 3.0 (1996), 0.8 (1997), 0.6 (1998), 3.1 (1999), 1.6 (2000).
  - Saudi Arabia: 0.9 (1996), -0.4 (1997), -0.2 (1998), -1.3 (1999), -1.1 (2000).
- Grouped patterns:
  - GCC, Maghreb, Former Soviet Union, and Other groups followed a fall-then-rise inflation pattern.
  - Classification by oil-export intensity (Top 5 oil, Some oil, Little oil) showed similar patterns; exception: little-oil exporters did not experience a post-crisis drop in inflation.

### Analysis of factors determining inflation — commodity prices, import pass-through, and persistence
- Commodity price dynamics:
  - World commodity prices (fuel and nonfuel) are central; division fuel vs nonfuel is important due to oil dependence and fuel price controls/subsidies.
  - Nonfuel commodity prices declined modestly during the 1990s and began rising starting in 2002; steep rise in energy prices began later than the rise in inflation.
  - Qualitative evidence: nonfuel primary commodities price index, energy price index, and total primary commodities index (in SDR terms, 2000=100) increased in the 2000s, temporally associated with rising inflation.
- Import prices and pass-through:
  - Most MENACA countries import the majority of commodities; import prices may be as good predictors of inflation as world commodity prices.
  - Literature findings cited (examples):
    - Duma (2008): low pass-through in Sri Lanka; most pass-through after 4 months.
    - Leigh and Rossi (2002): Turkey full pass-through complete after one year.
    - Gottschalk et al (2008): higher oil prices → higher inflation in Sierra Leone.
    - Habermeier et al (2009): commodity prices responsible for inflation; core inflation lower than headline in Middle East.
    - Loungani and Swagel (2001): fuel and nonfuel commodity prices significant in fixed exchange rate regimes among developing countries.
- Inflation persistence and domestic demand:
  - Lagged inflation (inflationary momentum) is a strong determinant; adaptive expectations and overlapping contracts contribute to persistence.
  - Output gap is important but hard to measure; economic growth often used as proxy; relationship varies across studies.
  - MENACA: GDP-weighted economic growth was significantly higher after 2000 than before 2000, but no U-shaped growth pattern corresponding to the inflation pattern.

### Interest rates, money growth, and measurement issues
- Interest rates and inflation:
  - Restrictions on free adjustment of interest rates can “strengthen the negative correlation between the inflation data and the real interest rate data, while at the same time making the relationship meaningless.”
  - Al-Mashat and Billmeier (2007): interest rates in Egypt had greater impact on inflation after adoption of an interest rate corridor in 2005.
  - Nominal interest rates likely positively related to inflation, but endogeneity concerns exist.
- Money growth and monetary factors:
  - Long-run expectation: money growth should be directly positively related to inflation when economic growth is included and velocity is stable.
  - Mechanism: money growth affects aggregate demand, mediated by interest rates or credit availability.
  - Evidence cited:
    - Nelson (2002): money growth matters for the US and UK even with interest rates included.
    - Bonato (2007): in Iran, money growth is the main determinant of inflation.
    - Khan and Schimmelpfennig (2006): M2 and private sector credit growth dominant in Pakistan with a 12-month lag.
    - Fanizza and Söderling (2007): monetary policy has little effect in MENA; money velocity shifted in early 2000s.
    - Saizar and Chalk (2008): monetary policy more effective in floating regimes in emerging markets.
  - Robust regression result: "The percent change in broad money had a robust coefficient of seven or eight percent with a high level of confidence."

### Exchange rates, US dollar, and pass-through
- Exchange rate channels:
  - Direct: prices of importables.
  - Indirect: current account and aggregate demand.
- Empirical findings:
  - Billmeier and Bonato (2004): low pass-through in Croatia due to administered prices; similar in many MENACA countries.
  - A stronger US dollar associated with lower inflation; level of the dollar appeared more important than change.
  - Cross-country regressions: "The percent change in the exchange rate had a negative coefficient of six or seven percent ... significant with over 99.99 percent confidence."
  - US inflation: in full-sample regressions coefficients "about a half"; “between 70 and 90 percent when fuel prices (not significant) were included.”
- Historical pattern:
  - Fall and rise of inflation in MENACA roughly corresponded to increased strength and subsequent weakness of the US dollar, though bivariate correlation and timing imperfect.

### Main empirical results on commodity prices and inflation
- Region-wide correlations:
  - Positive correlation between inflation and changes in nonfuel commodity prices.
  - No relation or sometimes negative relationship between inflation and changes in fuel prices.
- Estimated magnitudes:
  - Coefficients on nonfuel commodity prices were in the range of four to six percent in most regressions.
  - Coefficients on fuel prices were tiny (pass-through of only one or two percent), and when significant often negative.
  - Comparison: coefficient of 7 percent estimated for Brazil (IMF REO April 2008); roughly 3.3 percent estimated for the region as a whole for 2006-07 in that report.
- Robustness:
  - Results held when commodity prices measured in SDRs and in dollars, with lags of one year, and when substituting narrow indices.
  - Nonfuel commodity coefficients not very robust across subregions.
- Explanations for weak/negative fuel-price coefficients:
  - Subsidies and price controls mute pass-through.
  - Fiscal channel: higher fuel prices raise subsidy payments and could offset inflationary effects.
  - Remittances and real exchange rate support can offset inflation.
  - FSU vs non-FSU: nonFSU countries had small negative coefficients on oil prices; FSU countries had positive coefficients.

### The tested model and key estimation results
- Model notation:
  - Inflation_t = F( α0 + α1 inflation_t-1 + α2 US NEER_t + α3 Δ US NEER_t + α4 Δ exchange rate_t + α5 local interest rate (or monetary aggregate)_t + α6 Δ commodity prices_t, t-1 + α7 Δ import prices_t, t-1 + α8 GDP growth_t + α9 international interest rates, exchange rates, and inflation )
- Data and lags:
  - Annual data used primarily; monthly tests attempted but data limitations restricted scope.
  - Monthly tests: inflation lagged 1 month; other lagged variables usually lagged 12 months.
  - Commodity prices measured in SDRs and dollars; SDR series considered more stable.
  - Inflation series often capped at 10 percent for robustness.
- Selected reported regression coefficients (1997–2008; examples):
  - Inflation lag coefficient: around 0.45 to 0.49 (high significance).
  - US NEER level: negative coefficients in the range -0.12 to -0.07 (significant).
  - US inflation: positive coefficients around 0.47 to 0.86 where significant.
  - Exchange rate percent change: negative coefficients around -0.06 to -0.07 (significant).
  - M2 growth (lagged): coefficients around 0.07 to 0.08 (significant).
  - Nonfuel commodity prices (lagged, SDRs): reported as 0.04, 0.06, 0.05, 0.04, 0.04, 0.06, 0.08, 0.09 across regressions.
  - Fuel prices: around -0.01 to -0.02 (insignificant or marginal).
  - Adjusted R-Square examples: 0.73, 0.74, 0.74, 0.74, 0.74, 0.71, 0.71, 0.68, 0.69.
  - Observations reported in some regressions: 284, 300, 289.
- Robustness:
  - Coefficients and standard errors robust to adding/deleting monetary and exchange rate variables.
  - Removing US NEER increased significance of nonfuel commodity changes; removing variables reduced adjusted R-squared.
- Additional findings:
  - "Inflation in the previous period was the most significant explanatory variable, with a coefficient of about 50 percent in most regressions (or 20-30 percent if inflation was not capped). The significance was always well above the 99.99 percent level."
  - Credit growth substituted for broad money growth yielded similar results but smaller coefficients and lower adjusted R-squared.

### Lags, international interest rates, and regional differences
- Lags:
  - Moriyama (2008) in Sudan: exchange rate and money supply affected inflation with an 18-24 month lag.
  - In MENACA sample: exchange rate lagged one year — coefficient shrank slightly; M2 coefficient maximized with one-year lag; two-year lags often insignificant.
- International interest rates:
  - LIBOR not significant; US Prime rate and US T-bill rate significant.
  - US NEER stable and significant across many regressions.
  - US inflation significant in some regressions (more so in GCC).
- Regional and oil-intensity differences:
  - FSU: lagged inflation coefficient smaller; local exchange rate coefficient larger and highly significant; money supply coefficient smaller.
  - Excluding FSU: change in exchange rate not significant; money growth coefficient larger in non-GCC countries.
  - GCC: coefficients on US inflation larger and more significant; exchange rate and money supply coefficients not robust.
  - Oil intensity:
    - Fuel prices insignificant in all regressions.
    - Nonfuel commodity significance often breaks down when splitting by oil intensity.
    - Inflationary inertia larger for countries with more oil.
    - US NEER coefficients larger for countries with little oil.
    - M2 coefficients larger for countries with more oil (except GCC due to fixed exchange regimes).
- Exchange rate regimes:
  - Aggregate results replicated across fixed vs floating regimes: inflationary inertia significant; fuel prices not significant; nonfuel commodity prices often significant.
  - Exchange rate coefficient oddly larger for fixed-rate regimes when measured against the SDR.
  - Inflationary inertia coefficient greater for non-fixed rate countries (about a third → about a half).
  - US NEER larger for fixed regimes; US inflation significant in some fixed-regime regressions, not in floating.
  - Regime indicator (fixed vs floating) not significant in pooled regressions.

### Monthly data, political breaks, crisis period, and time trend
- Monthly vs annual:
  - Monthly results similar to annual but with restricted tests.
  - Month-significant coefficients retained: exchange rate change (smaller), M2 growth (smaller), nonfuel commodity changes (significant).
  - Fuel price changes remained insignificant.
- Political developments:
  - Break-dummies tested for 1998–2005: break in 2001 (9/11) found significant; other years insignificant.
- Global Financial Crisis:
  - Regressions run through 2007 and through most recent data (including 2008) showed similar results; inflation peaked in 2008.
  - Crisis strengthened some coefficients; pre- and post-crisis results largely consistent.
- Time trend (Year):
  - Year variable significant in almost every regression; coefficients around 0.15–0.17 (described as "around 15 percent, meaning that inflation would increase by about 0.15 points every year").
  - Including Year reduced coefficients and significance of US inflation and nonfuel commodity prices (nonfuel significance restored if US NEER excluded).
  - Interpretation: Year captures additional factors (e.g., financial deepening) not in other variables.

### Summary conclusions and policy implications
- Energy (fuel) price changes:
  - Generally not significant in regression results; sometimes coefficients negative; result robust across specifications and subsets.
  - Prevalence of fuel price controls/subsidies likely mutes pass-through.
- Nonfuel commodity prices:
  - Changes significant in most regressions, but not robust across subregions or when Year is included.
- US dollar/US NEER:
  - Strength of the US dollar has a significant impact on inflation; US NEER robustly significant.
  - Coefficient greater for fixed rate countries and for countries with less oil.
- Inflationary inertia:
  - Lagged inflation large and significant in all regressions (around 0.3 to 0.5 across subsamples; full-sample about 0.5).
- Exchange rate and money supply:
  - Both significant but importance varies by cross-section (FSU, fixed-exchange-rate, oil intensity).
- Real growth and credit:
  - Real growth not significant in regressions.
  - Credit growth behaves like money supply growth but with lower coefficients and adjusted R-squared.
- Time trend and structural break:
  - Time trend significant and positive in most regressions.
  - Dummy for 2001 break significant.
- Policy implications highlighted:
  - Countries with exchange rate flexibility have instruments to combat inflation; any resurgence should be met with a strong policy response.
  - Caution against relying on falling fuel prices to address high inflation; falling nonfuel commodity prices may more reliably lower inflation.
  - Countries formally or informally linked to the US dollar have limited policy options and might consider greater exchange rate flexibility as an alternative to remaining aligned with the US dollar.

### Data, sample, measurement, and unit-root tests
- Inflation measured with CPI from IMF IFS Statistics; period average CPI used (end-period CPI tested with similar results).
- Commodity prices, import prices, deposit/lending rates, exchange rates, monetary aggregates: IMF IFS Statistics.
- GDP data: IMF WEO Statistics.
- Sample period: 1996–2009.
  - Rationale: captures steep drop then resurgence in inflation; pre-1996 data patchier; financial changes make earlier periods less comparable.
- Treatment of extremes:
  - Inflation series capped at 10 percent in most regressions (also tested cap at 15 percent); capping did not qualitatively change results.
- Lag choices:
  - Lagged commodity prices used as they resulted in higher adjusted R-squared; nonfuel coefficients more significant when lagged.
- Unit-root tests:
  - Fisher type unit root tests (ADF-based: Inverse chi-squared (50), Inverse normal Z, Inverse logit t(129), Modified inv. chi-squared Pm).
  - For inflation, unit roots were rejected with over 99.99 percent confidence.
  - For inflation capped at 10 percent, confidence values ranged from 99.53 percent to 99.97 percent.
- Countries analyzed (included):
  - Algeria, Armenia, Azerbaijan, Bahrain, Djibouti, Egypt, Georgia, Iran, Jordan, Kazakhstan, Kuwait, Kyrgyz Republic, Lebanon, Libya, Mauritania, Morocco, Oman, Pakistan, Qatar, Saudi Arabia, Sudan, Syria, Tajikistan, Tunisia, United Arab Emirates.
- Countries excluded due to lack of data:
  - Afghanistan, Iraq, Somalia, Turkmenistan, Uzbekistan, West Bank and Gaza, Yemen.

*Source: IMF IFS Statistics; chapter text from _wp10135 - 1. Inflation, 1996-2007 and associated sections cited in the provided PDF excerpt.*

### 1. Inflation, 1996-2007 ................................................................................................

### _wp10135 - 1. Inflation, 1996-2007 ................................................................................................

### Introduction: recent trends and broad drivers
- From 1996 until 2000 regional inflation, weighted by GDP, declined from 11.5 percent to 2.5 percent; since 2000 this trend reversed with increases in every year until 2008 when inflation jumped into double digits.
- After 2000 inflation made a comeback globally; in the MENACA region the increase has been broad and steady since 2000 and in 2007 almost every single country in this region had higher inflation than in 2000.
- Central bank policies, exchange rate regimes (official or unofficial pegs to the US dollar), and money supply growth are highlighted as ultimately important drivers of inflation in MENACA.
- Policy-related observations:
  - Many countries, particularly in the Maghreb, failed to increase nominal interest rates in the face of increasing inflation because of constraints of pegs to the US dollar.
  - Some GCC countries increased rates despite the peg but “not enough to keep up with inflation.”
  - Maintaining adherence to an official or unofficial peg is often identified as the key policy decision affecting inflation outcomes.
- Global factors emphasized:
  - The resurgence in inflation coincided with sharp increases in food and energy prices.
  - The decline in the value of the US dollar has been tied to commodity prices and to world inflation; a subsequent strengthening of the US dollar could reduce world inflation.
  - Literature and views cited (summarized in the chapter): Borio and Filardo (global factors increasingly important), Rogoff (globalization changed Phillips curve), Ball (disputes some global-inflation claims), D’Agostino (US inflation and global liquidity).

### Background and overview of inflation in MENACA
- GDP-weighted inflation, 1996–2009 (year : percent):
  - 1996: 11.5
  - 1997: 6.8
  - 1998: 5.4
  - 1999: 4.5
  - 2000: 2.4
  - 2001: 2.7
  - 2002: 3.3
  - 2003: 4.0
  - 2004: 5.0
  - 2005: 5.0
  - 2006: 6.0
  - 2007: 8.6
  - 2008: 13.0
  - 2009: 7.6
- Cross-country patterns:
  - Despite wide variation in levels, almost all countries in the sample experienced a decline in inflation between 1996 and 2000 and an increase from 2001 to 2008.
  - Examples from the data:
    - Sudan reduced inflation from triple digits to single digits over 1996–2000 (Table shows 132.8 in 1996, 46.7 in 1997, then falling to single digits by 2000).
    - Tajikistan lowered inflation from over 400 percent to just over 30 percent (Table shows 418.5 in 1996, 88.0 in 1997, 43.2 in 1998, 27.5 in 1999, 32.9 in 2000).
    - Kuwait lowered its rate from 3 percent to 1.6 percent over 1996–2000 (Table shows 3.0 in 1996, 0.8 in 1997, 0.6 in 1998, 3.1 in 1999, 1.6 in 2000).
    - Saudi Arabia lowered its rate from 0.9 percent to minus 1.1 percent over 1996–2000 (Table shows 0.9 in 1996, -0.4 in 1997, -0.2 in 1998, -1.3 in 1999, -1.1 in 2000).
  - Grouped patterns:
    - GCC, Maghreb, Former Soviet Union, and Other groups all followed a similar fall-then-rise pattern in inflation.
    - Classification by oil-export intensity (Top 5 oil, Some oil, Little oil) also shows similar patterns; one noted exception: countries that export little oil did not experience a drop off in inflation after the global financial crisis.

### Analysis of factors determining inflation
- Factor input costs and commodity prices:
  - World commodity prices (fuel and nonfuel) are central variables considered for their impact on consumer prices; division into fuel versus nonfuel is particularly relevant in MENACA because of oil dependence and prevalence of fuel price controls/subsidies.
  - Nonfuel commodity prices declined modestly during the 1990s and began rising starting in 2002; the steep rise in energy prices began later than the rise in inflation.
  - Figure summary (qualitative): Nonfuel primary commodities price index, Energy price index, and Total primary commodities price index (in SDR terms, 2000=100) exhibit increases in the 2000s that are temporally associated with rising inflation.
- Import prices and pass-through:
  - Most MENACA countries import the majority of commodities; import prices may be at least as good predictors of inflation as world commodity prices, with differences attributable to manufactured goods and services imports and local commodity production shares.
  - Empirical findings from other studies referenced:
    - Duma (2008): low pass-through of external shocks to inflation in Sri Lanka; oil pass-through low and negative; most pass-through after 4 months.
    - Leigh and Rossi (2002): in Turkey full pass-through of external shocks complete after one year.
    - Gottschalk et al (2008): higher oil prices result in higher inflation in Sierra Leone.
    - Habermeier et al (2009): all commodity prices, including fuel, responsible for inflation; core inflation lower than headline inflation in Middle East.
    - Loungani and Swagel (2001): both fuel and nonfuel commodity prices significant in fixed exchange rate regimes among developing countries.
- Inflation persistence and domestic demand:
  - Inflationary momentum (lagged inflation) is a strong determinant; adaptive expectations and overlapping contracts contribute to persistence.
  - Output gap and growth:
    - Output gap is important but difficult to measure; economic growth is often used as a proxy.
    - Relationship between growth and inflation varies; generally positive in many studies, though some MENACA evidence suggests a backward-bending Phillips curve in high-inflation countries (Khan 2004).
    - In MENACA as a whole, GDP-weighted economic growth was significantly higher after 2000 than before 2000, but there was no U-shaped growth pattern corresponding to the inflation pattern.
- Real interest rates and measurement issues:
  - Higher real interest rates should reduce aggregate demand and inflation, but measurement problems exist because real rates are calculated as nominal rates minus inflation.
  - The chapter notes a statistical artifact: if nominal interest rates and reported inflation are noisy, a spurious negative relationship between inflation and real interest rates can arise (since real rate = nominal minus inflation).
- Role of exchange rates and the US dollar:
  - The decline in the US dollar has been linked to higher commodity prices and higher world inflation; a stronger dollar could have offsetting effects on inflation.
  - Exchange rate regime constraints (pegs) are repeatedly noted as limiting domestic policy responses to rising inflation.

*Source: IMF IFS Statistics; chapter text from _wp10135 - 1. Inflation, 1996-2007.*

### 20.      Furthermore, interest rates do not always move freely, which can complicate the

### _wp10135 - 20.      Furthermore, interest rates do not always move freely, which can complicate the

### Interest rates and inflation
- Restrictions on the free adjustment of interest rates would "strengthen the negative correlation between the inflation data and the real interest rate data, while at the same time making the relationship meaningless."
- Empirical finding: Al-Mashat and Billmeier (2007) found that interest rates in Egypt began to have a greater impact on inflation following the adoption of an interest rate corridor in 2005.
- Nominal interest rates are likely positively related to inflation, but regressions that include nominal interest rates suffer from endogeneity because "higher inflation is usually associated with higher nominal interest rates" and causation may run from inflation to nominal rates rather than the reverse.

### Money growth and monetary factors
- Long-run expectation: "Money growth in the long run should be directly positively related to inflation, as long as economic growth is included in the regression and there is no reason to expect significant shifts in money velocity."
- Mechanism: "Money growth would affect inflation through its impact on aggregate demand, which in turn would be impacted by interest rates (to the extent that these are flexible) or other measures of the availability of credit (Nelson 2002)."
- Evidence cited:
  - Nelson (2002): money growth matters for the US and UK even when interest rates are included.
  - Bonato (2007): in Iran money growth is the main determinant of inflation, though real interest rates, real output, and the exchange rate are also important.
  - Khan and Schimmelpfennig (2006): monetary factors, including M2 and private sector credit growth, were dominant in Pakistan with a lag of 12 months.
  - Fanizza and Söderling (2007): monetary policy has little effect on inflation in the Middle East and North Africa; money velocity shifted significantly in the early 2000s.
  - Saizar and Chalk (2008): monetary policy is more effective in floating rate regimes in emerging market countries.
- Robust regression result: "The percent change in broad money had a robust coefficient of seven or eight percent with a high level of confidence."

### Exchange rates, effective exchange rates, and pass-through
- Nominal effective exchange rate ("US NEER") and simple nominal exchange rates were used due to data availability constraints.
- Channels:
  - Direct: affecting prices of importables.
  - Indirect: influencing the current account and aggregate demand.
- Empirical findings:
  - Billmeier and Bonato (2004): low exchange rate pass-through in Croatia due to administered prices; similar situations prevail in many MENACA countries.
  - A stronger US dollar was associated with lower inflation; level of the value of the dollar appeared more important than the change.
  - "The percent change in the exchange rate had a negative coefficient of six or seven percent in cross-country regressions with capped inflation and was significant with over 99.99 percent confidence."
  - Regional variation: coefficients on variables measuring the strength of the dollar were significant with high confidence for MENA countries, but insignificant or barely significant in regressions for FSU countries.

### International factors (world interest rates, major currencies, and commodities)
- World interest rates and exchange rates can influence aggregate demand in countries with fixed or floating regimes, especially where capital movements are unrestricted.
- Major world interest rates (US Treasury rates, LIBOR, etc ...) are "more reliably measured and more reliably market determined than many local interest rates," so their impact on inflation should be easier to detect and stronger.
- Al-Hassan (2008) incorporated international and regional interest rates in a study of inflation in the GCC countries.
- Caveats:
  - Capital controls could reduce the impact of world interest rates.
  - In flexible exchange rate regimes or questionable fixed arrangements, expectations about currency movements would adjust the impact of world interest rates on local aggregate demand, making practical use difficult.
- Currency level effects:
  - Movements and levels of major currencies affect aggregate demand and can serve as proxies for real or nominal effective exchange rate data where local data are lacking.
  - In MENACA, many countries link currencies to the US dollar; "a decline in the value of the US dollar could have the same effect as a decline in the value of the local currency."
  - A strong US dollar could reflect policies such as high interest rates that "could spill over into countries that align their currencies with the dollar."
  - US inflation was found to be significant in full-sample regressions with coefficients "about a half," i.e., "every percent increase in US inflation could result in a half percent increase in the MENACA region." Coefficients increased to "between 70 and 90 percent when fuel prices (not significant) were included in the regression."
- Historical patterns:
  - The fall and rise of inflation in the MENACA region "has roughly corresponded to the increased strength and subsequent weakness of the US dollar," though the bivariate correlation was limited and timing did not perfectly align.
  - SDR exchange rates and the US NEER show more moderate patterns than dollar/euro swings.

### Commodity prices and inflation (Main Results)
- For the region as a whole:
  - Positive correlation between inflation and changes in nonfuel commodity prices.
  - No relation or sometimes negative relationship between inflation and changes in fuel prices.
- Estimated magnitudes:
  - "The coefficients on nonfuel commodity prices were in the range of four to six percent in most regressions," implying "even a doubling of commodity prices would only suggest an increase of four to six points to inflation."
  - This is compared to a "coefficient of 7 percent ... estimated for Brazil" in the April 2008 IMF REO for Latin America and "roughly 3.3 percent ... estimated for the region as a whole for 2006-07" in that report.
  - "The coefficients on fuel prices were tiny, indicating a pass-through of only one or two percent, and in the rare cases when they were significant they were negative."
- Robustness and variants:
  - Results held when testing commodity price changes measured in SDRs and in dollars, with lags of one year, and when substituting narrow indices (average crude, US Gulf wheat) for aggregate indices.
  - Significance of nonfuel commodity coefficients was "not very robust when the sample of countries was divided into subregions."
- Explanations for a negative or insignificant fuel-price coefficient:
  - Subsidies and price controls can mute pass-through (Habermeier et al., described).
  - Fiscal channel: higher fuel prices increase subsidy payments and could "leave less money left over for other expenditures," offsetting inflationary effects.
  - Remittances: Lueth and Ruiz-Arranz (2006) found remittances positively correlated with oil prices in Sri Lanka; similar dynamics could reduce recipient-country inflation via real exchange rate support.
  - FSU vs non-FSU differences: nonFSU countries had small negative coefficients on oil prices; FSU countries had positive coefficients on fuel prices.
- Import prices:
  - Changes in commodity prices are "not simply a proxy for changes in overall import prices." Including an import price index often produced an insignificant coefficient and had little impact on the nonfuel commodity price coefficient. In regressions using uncapped inflation, import prices were significant and reduced the coefficients on nonfuel commodity prices.

### The tested model and estimation details
- Model tested (notation preserved):
  - Inflation_t = F( α0 + α1 inflation_t-1 + α2 US NEER_t + α3 Δ US NEER_t + α4 Δ exchange rate_t + α5 local interest rate (or monetary aggregate)_t + α6 Δ commodity prices_t, t-1 + α7 Δ import prices_t, t-1 + α8 GDP growth_t + α9 international interest rates, exchange rates, and inflation )
- Data and lags:
  - Tests used annual data; monthly tests were attempted but many series were not available monthly.
  - In monthly tests inflation was lagged by 1 month and other lagged variables were usually lagged by 12 months, with different lags also tested.
  - Simple nominal exchange rates were used for MENACA because nominal effective exchange rates were unavailable for most countries; nominal effective exchange rates were tested where available.
  - Commodity prices were measured in SDRs and in dollars; SDR series considered more stable and representative of MENACA trading partners.
- Estimation summary (selected reported regression results, 1997-2008):
  - Inflation lag coefficient: around 0.45 to 0.49 (t-statistics reported, significance levels *** for 99 percent).
  - USNEER level: negative coefficients in the range of -0.12 to -0.07 (t-statistics significant).
  - US inflation: positive coefficients around 0.47 to 0.86 where significant.
  - Exchange rate percent change: negative coefficients around -0.06 to -0.07 (significant).
  - M2 growth (lagged): coefficients around 0.07 to 0.08 (significant).
  - Nonfuel commodity prices (lagged, measured in SDRs): coefficients reported at 0.04, 0.06, 0.05, 0.04, 0.04, 0.06, 0.08, 0.09 across regressions (t-statistics ranging from marginal to highly significant).
  - Fuel prices: coefficients around -0.01 to -0.02 (insignificant or marginally significant; negative when significant).
  - Adjusted R-Square values reported in regressions: e.g., 0.73, 0.74, 0.74, 0.74, 0.74, 0.71, 0.71, 0.68, 0.69.
  - Observations in reported regressions: values reported include 284, 300, 289.
  - Notes: Fixed effects used; t-statistics shown in parentheses; inflation capped at 10 percent in some regressions.
- Robustness:
  - Coefficients and standard errors were robust when monetary and exchange rate variables were added or deleted.
  - Dropping exchange rates and money growth (or vice-versa) did not substantially alter results, although removing the US NEER increased the significance of nonfuel commodity price changes.
  - Removing any of these variables reduced adjusted R-squared for the regressions.
- Additional specific findings:
  - "Inflation in the previous period was the most significant explanatory variable, with a coefficient of about 50 percent in most regressions (or 20-30 percent if inflation was not capped). The significance was always well above the 99.99 percent level."
  - "Berger and Harjes (2009) found that US liquidity developments led Euro area liquidity. If the same were the case regarding the MENACA region higher liquidity in the US could lead MENACA liquidity, causing inflation in the MENACA region and a weaker US dollar to occur contemporaneously. This would result in a correlation between the level of the dollar and inflation in the MENACA region."
  - Growth of credit to the private sector substituted for broad money growth yielded similar results but with a smaller coefficient, lower significance, and smaller adjusted R-squared.

*IMF working paper excerpt*

### 49.      Moriyama (2008) found that in Sudan changes in the exchange rate and money

### _wp10135 - 49.      Moriyama (2008) found that in Sudan changes in the exchange rate and money

### Exchange rate and money supply lags
- Moriyama (2008) found that in Sudan changes in the exchange rate and money supply affected inflation with an 18-24 month lag.
- For the MENACA sample:
  - Lagging the exchange rate by one year: the coefficient on the exchange rate shrank very slightly and was significant with slightly lower probability.
  - The coefficient and t-statistic on money supply growth (M2) were maximized with a lag of one year.
  - When either variable was lagged by two years it became insignificant.
- The coefficient on the exchange rate is described as consistent with other studies (Eichenbaum, Burstein, and Rebelo 2002; Laflèche 1997).

### International interest rates and global influences
- Some evidence that international interest rates mattered:
  - LIBOR was not significant in regressions.
  - The US Prime rate and the US T-bill rate were significant.
  - These results hold for countries that link their currencies to the US dollar and also for the FSU region, where fewer such linkages exist.
- US NEER:
  - Appears as a stable and significant coefficient across many regressions (see tables).
- US inflation:
  - Significant in some regressions; not robust across regions.
  - More likely significant and larger in the GCC.

### Real growth, import prices, and exports
- Real growth:
  - Significant in some regressions, but not when any variable measuring the strength of the US dollar or money supply was included.
  - When significant it had a positive coefficient.
  - Real nonoil growth (estimated by subtracting oil imports from GDP) was not significant.
- Import prices:
  - Not significant in any regression.
  - Including import prices did not change the significance of other variables, including commodity prices.
- Exports:
  - Change in exports divided by GDP was not significant.
  - Oil exports were not significant, even in the subset of GCC countries.

### Local and real interest rates
- Nominal deposit and lending rates:
  - Not significant in regressions.
  - Including them increased the coefficient on the US NEER and reduced confidence in coefficients on US inflation and the exchange rate.
- Real deposit and lending rates:
  - Negatively related to inflation with over 99.99 percent confidence.
  - Coefficient on the real deposit rate larger and significant at a greater confidence level than the lending rate.
  - Real interest rates computed as nominal rate minus inflation.
  - If adaptive expectations used (subtracting previous period’s inflation), coefficients on real deposit and lending rates became insignificant.
- Caution:
  - Endogeneity and prevalence of controls on interest rates in MENACA countries make inferences risky.
  - Behavior of coefficients on most other variables does not change substantially when interest rates are added.

### Key regression coefficients and significance (selected from Table 3)
- Inflation (lagged one period): 0.42, 0.56, 0.37, 0.41, 0.41, 0.40, 0.39, 0.38, 0.40 (T-statistics repeatedly *** indicating strong significance in these columns).
- US NEER: -0.15, -0.14, -0.13, -0.10, -0.06, -0.11, -0.09, -0.10 (T-statistics show multiple *** and significant values).
- US inflation: 0.48, 0.49, 0.54, 0.96, 0.55, 0.00, 0.49 (mixed significance, some * and **).
- Exchange rate (percent change): -0.04, -0.06, -0.07, -0.06, -0.06, -0.04, -0.07 (several */**/*** significance levels).
- M2 (lagged one period, percent change): 0.13, 0.09, 0.09, 0.08, 0.07, 0.07, 0.06, 0.08 (several ***).
- Nonfuel commodity prices (lagged one period, measured in SDRs): 0.05, 0.08, 0.06, 0.02, 0.05 (a number of */** significance).
- Fuel prices: -0.02 (insignificant in many specifications).
- Nominal deposit rate and nominal lending rate: coefficients around 0.03 to -0.05 (generally insignificant).
- Real deposit rate: -0.21 (T-statistic -7.13) ***.
- Real deposit rate with adaptive expectations: -0.02 (T-statistic -0.64) (insignificant).
- Adjusted R-Squared values reported: 0.67, 0.63, 0.67, 0.68, 0.69, 0.69, 0.68, 0.76, 0.69.
- Observations across columns: 211, 202, 202, 202, 202, 183, 202, 202.

### Regional differences (FSU, NoFSU, GCC)
- FSU characteristics:
  - Lagged inflation coefficient somewhat smaller.
  - Local exchange rate coefficient larger and significant at a much higher confidence level.
  - Money supply coefficient smaller and significant at lower confidence.
  - US NEER coefficient somewhat smaller and significant with lower confidence.
  - Adjusted R-Squareds lowest in FSU regressions.
- Excluding FSU:
  - Change in exchange rate not significant; coefficient on money growth larger (especially for non-GCC countries).
  - Possible explanation: greater prevalence of controls in MENA countries insulating citizens from exchange rate movements.
- GCC:
  - Coefficients on US inflation larger and significant with greater confidence.
  - Coefficients on changes in exchange rate and money supply not robust.
  - Adjusted R-Squareds highest in GCC regressions.

### Oil export intensity differences
- Fuel prices:
  - Changes in fuel prices insignificant in all regressions.
- Nonfuel commodity prices:
  - Significance breaks down in most regressions when splitting by oil intensity.
- Inflationary inertia:
  - Significant in all regressions; coefficients substantially larger for countries with more oil.
- US NEER:
  - Countries with little oil had substantially larger coefficients on US NEER with greater likelihood of significance.
  - Medium oil producers: coefficients slightly larger and marginally more significant than large oil producers.
- M2 (broad money):
  - Countries with more oil had larger coefficients with greater likelihood of significance on change in broad money.
  - Suggests large oil producers might have more independent monetary policies; GCC exception possibly due to fixed exchange rate regimes.
- Exchange rate:
  - Significant only in low oil producing countries; significance disappears when FSU countries are excluded.

### Exchange rate regimes: fixed versus floating
- General replication:
  - Aggregate results replicated across regimes: inflationary inertia significant; fuel prices generally not significant; nonfuel commodity prices often significant (often at 90 percent confidence).
- Exchange rate coefficient:
  - Oddly larger for fixed-rate regimes (many fixed rates are against the US dollar while exchange rates here are measured against the SDR).
  - In fixed rate countries the exchange rate may capture part of the effect of the US NEER.
- Inflationary inertia:
  - Coefficient greater for non-fixed rate countries than fixed rate countries (increasing from about a third to about half).
- US NEER:
  - Significantly larger for fixed rate regimes than for floating regimes, though still sometimes significant for non-fixed regimes.
- US inflation:
  - Significant in some regressions with fixed rate regimes, but never with floating regimes.
- Regime indicator:
  - A coefficient representing fixed vs floating regime had no significance in any regression; no variable behaved significantly differently when filtered only for fixed or floating regimes.
- Observations and adjusted R-squared (selected from Tables 6a and 6b):
  - Fixed regime observations: 112 (various columns).
  - Floating regime observations: 172.
  - Adjusted R-Squared values reported for sub-samples around 0.69–0.72 (fixed) and 0.72 (float), and up to 0.80 in NoFSU float sub-sample.

*Italic: Content drawn from the provided IMF PDF chapter/section text.*

### 68.      Results using monthly data were similar to the results using annual data, though the

### Results using monthly data and robustness checks

### Monthly versus annual data
- Results using monthly data were similar to the results using annual data, though the range of tests was restricted because many series were not available monthly.
- Coefficients that remained significant using monthly data:
  - Change in the exchange rate: significant but smaller and significant at a lower level of confidence.
  - Growth of the money supply (M2): significant but smaller and significant at a lower level of confidence.
- Coefficients that remained insignificant:
  - Changes in fuel prices: remained insignificant.
- Coefficients that remained significant:
  - Changes in nonfuel commodity prices: significant.

### Political developments
- Political factors were not directly available as variables.
- Dummies testing for breaks were implemented for:
  - 2001 (9/11) → a break in 2001 was found to be significant.
  - All other years during 1998-2005 → breaks in all other years were insignificant.
- Literature cited in discussion:
  - Aisen and Veiga (2006): find higher inflation associated with greater ideological polarization.
  - Alesina and Drazen (1991): find higher inflation related to greater reliance on seignorage from political standoffs.

### Global Financial Crisis (1997–2007 and through 2008)
- Regressions run using data from 1997 through the most recently available data and separately through 2007 to check for crisis-driven quirks.
- Inflation peaked in 2008, with the rise having been stopped by the end of 2008.
- The crisis strengthened some results by increasing the size and significance of some coefficients, but almost all coefficients that were significant using data through the crisis were also consistent using data that stop before the crisis.
- Note: most post-crisis regressions only have complete data through 2008, so the difference between pre- and post-crisis regressions is just that one year.

Key coefficients reported in Table 2b (1997-2007 regressions, fixed effects; t-statistics in parentheses):
- Constant examples: 10.65 (3.85), 8.05 (2.79), 9.65 (2.48), 9.20 (2.34), 8.91 (2.74), 10.88 (3.65), 10.33 (2.59), 0.33 (0.97), 0.31 (0.89), 0.34 (1.01).
- Inflation (lagged one period) coefficients: 0.42, 0.43, 0.42, 0.41, 0.42, 0.42, 0.42, 0.48, 0.48, 0.47 (t-statistics up to (9.69)***).
- USNEER examples: -0.10, -0.08, -0.09, -0.08, -0.07, -0.09, -0.08 (t-statistics around (-4.44)*** to (-2.65)***).
- Exchange rate (percent change) examples: -0.05, -0.06, -0.06, -0.06, -0.06, -0.05, -0.05, -0.05 (t-statistics around (-3.07)*** to (-3.09)***).
- M2 (lagged one period, percent change) examples: 0.08, 0.07, 0.07, 0.07, 0.07, 0.08, 0.08, 0.08 (t-statistics around (5.28)*** to (5.12)***).
- Nonfuel commodity prices (lagged one period, measured in SDRs): coefficients include 0.05, 0.04, 0.03, 0.04, 0.05, 0.05, 0.07, 0.07, 0.06 (various significance levels up to (4.29)***).
- Fuel prices (lagged one period, measured in SDRs): 0.01, 0.00, 0.00, 0.00 (t-statistics (0.62), (0.48), (-0.21), (0.16)) — generally insignificant.
- Import prices (lagged one period): 0.02, 0.02, 0.02 (t-statistics (0.82), (0.91), (1.27)) — generally not significant.
- Adjusted R-Square examples: 0.73, 0.74, 0.74, 0.74, 0.74, 0.71, 0.71, 0.72, 0.72, 0.72.
- Observations: 264, 264, 264, 264, 264, 275, 275, 264, 264, 264.
- Notes: Inflation is capped at 10 percent. Many variables lagged one period. Nonfuel and fuel prices measured in SDRs in some regressions.

### Time trend (Year) inclusion and effects
- A year (time trend) variable was not included in main regressions because its justification is vague, but was tested separately.
- The year variable was significant in almost every regression with high confidence.
- Coefficients on year were around 0.15–0.17 (described as "around 15 percent, meaning that inflation would increase by about 0.15 points every year independently of the factors included in these regressions").
- Including the year had minimal impact on most coefficients except:
  - US inflation: coefficients shrank greatly and became insignificant when the year was added.
  - Nonfuel commodity prices: coefficients shrank and became insignificant when the year was added (although excluding US NEER restored significance).
- Changes in fuel prices remained insignificant in all regressions including the year.
- Key coefficients from Table 7 (1997–2008 regressions with Year; fixed effects):
  - Year coefficients: 0.17, 0.15, 0.15, 0.16, 0.16, 0.16, 0.16, 0.17, 0.15 (t-statistics (3.75)***, (3.18)***, (3.03)***, (3.16)***, (3.51)***, (3.28)***, (3.42), (3.06)***, (2.04)**).
  - Inflation (lagged): 0.48, 0.49, 0.48, 0.48, 0.47, 0.49, 0.49, 0.51, 0.48 (t-statistics up to (10.32)***).
  - USNEER examples: -0.10, -0.09, -0.09, -0.10, -0.09, -0.09, -0.09 (t-statistics around (-4.39)*** to (-2.81)***).
  - M2 examples: 0.07, 0.07, 0.06, 0.06, 0.07, 0.07, 0.07, 0.06 (t-statistics around (4.50)*** to (4.08)***).
  - Nonfuel commodity prices became smaller and often insignificant when Year included (examples: 0.02, 0.00, 0.06, 0.04, 0.01, 0.00 with t-statistics ranging from (2.91)*** to (0.12)).
  - Fuel prices remained mixed/insignificant (examples: 0.01, -0.01, 0.00, 0.01 with t-statistics (1.30), (-0.91), (-0.82), (0.48)).
  - Adjusted R-Square examples: 0.74, 0.74, 0.74, 0.74, 0.74, 0.70, 0.70, 0.72, 0.74.
  - Observations: 284, 284, 284, 284, 284, 289, 289, 300, 284.
- Interpretation offered: the significant time trend suggests the presence of additional factors (e.g., financial deepening) not captured by included variables.

### Summary and conclusions (selected findings)
- Energy (fuel) price changes:
  - Main controversial result: changes in energy prices were generally not found to be significant in regressions on inflation; sometimes coefficients were even negative.
  - This result was robust across regressions with different variable mixes, geographic subsets, oil intensity, exchange rate regimes, and lags.
  - Prevalence of price controls on petroleum products (and subsidies) may explain weak pass-through from fuel prices to measured inflation.
  - Import prices similarly showed weak significance.
- Nonfuel commodity prices:
  - Changes in nonfuel commodity prices were found significant in most regressions, but significance was not robust across country subsets or when a trend variable was included.
- US dollar strength:
  - The strength of the US dollar has a significant impact on inflation.
  - The coefficient was greater for fixed rate countries and for countries with less oil, but was significant in at least some regressions across cross sections.
  - US NEER is robustly significant; US REER gives similar results with slightly lower confidence; US Treasury bill rate yields similar results.
- Inflationary inertia:
  - Confirmed as important. Lagged inflation coefficients large and significant in all regressions (around 0.3 to 0.5 across subsamples; full-sample coefficient about 0.5).
  - Persistence more significant in regressions limited to non-fixed exchange rate regimes.
- Exchange rate and money supply:
  - Both significant but with variance across cross sections.
  - FSU and fixed-exchange-rate countries: changes in exchange rates had larger coefficients.
  - Non-FSU, floating-rate, more-oil countries: changes in money supply had larger coefficients.
- Real growth:
  - Not significant in any regressions.
- Credit growth:
  - Behaves similarly to money supply growth but yields lower coefficients and lower adjusted R-squareds.
- Time trend and structural break:
  - Time trend significant and positive in most regressions.
  - Dummy for a 2001 break significant, suggesting political pressures or other changes following 9/11 impacted inflation.
- Policy implications highlighted:
  - Countries with exchange rate flexibility have instruments to combat inflation; any resurgence should be met with a strong policy response.
  - Caution against relying on falling fuel prices to address high inflation; falling nonfuel commodity prices may more reliably lower inflation.
  - Countries formally or informally linked to the US dollar have limited options and might consider greater exchange rate flexibility as an alternative to remaining aligned with the US dollar (which provides stability benefits but can impose costs when economic cycles diverge).

### Table 8 — Summary of regression variables (selected entries)
- Inflation in previous period:
  - Expected: should be highly significant with a coefficient between one and zero.
  - Actual: Highly significant and robust with a coefficient of about 0.5, varying from 0.3 to 0.5 in some subsamples.
- Strength of the US dollar (NEER, exchange rate versus the SDR or the euro):
  - Actual: US NEER robustly significant; rate-of-change of US NEER not significant in most regressions.
- US inflation:
  - Actual: significant in many regressions, usually at the 95 percent level; coefficient for large samples usually about 0.5, sometimes close to or higher than one in subsamples.
- Change in exchange rate (versus the SDR):
  - Actual: coefficients positive and significant overall, significance vanishes when FSU countries are excluded; significance high for FSU countries.
- Change in money plus quasi-money:
  - Actual: significant positive coefficient, but smaller and sometimes not significant when US NEER or US/euro exchange rate included.
- Change in nonfuel commodity prices (denominated in dollars or SDRs):
  - Actual: significant in most full-sample regressions, not robust across subregions; impact stronger in FSU region.
- Change in fuel prices (denominated in dollars or SDRs):
  - Actual: insignificant in almost all regressions; marginal significance where present had tiny coefficients.
- Import prices:
  - Actual: not significant in most regressions.
- International interest rates:
  - Actual: behave much like US NEER but with lower adjusted R-squareds and confidence levels.
- Local real and nominal interest rates:
  - Actual: real interest rates significantly negatively correlated; nominal rates insignificant; endogeneity concerns noted.
- Fixed or floating exchange rate:
  - Actual: regime indicator not significant in pooled regressions; separate regressions by regime showed money supply and US NEER more important for floating rate countries, US inflation more important for fixed rate countries.
- Breaks in years 1998-2005:
  - Actual: a break in 2001 resulting in higher inflation was significant; no other year breaks significant.
- Time trend:
  - Actual: highly significant with a positive coefficient in most regressions; causes nonfuel commodity prices to become insignificant when included.

### Unit root tests
- Fisher type unit root tests based on augmented Dickey-Fuller tests were conducted on inflation and inflation capped at 10 percent using four tests:
  - Inverse chi-squared (50)
  - Inverse normal Z
  - Inverse logit t(129)
  - Modified inv. chi-squared Pm
- Results:
  - For inflation, unit roots were rejected with over 99.99 percent confidence for all tests.
  - For inflation capped at 10 percent, confidence values ranged from 99.53 percent to 99.97 percent.

### Data and countries analyzed
- Focus: countries covered by the IMF’s MCD department for which data were available.
- Included countries: Algeria, Armenia, Azerbaijan, Bahrain, Djibouti, Egypt, Georgia, Iran, Jordan, Kazakhstan, Kuwait, Kyrgyz Republic, Lebanon, Libya, Mauritania, Morocco, Oman, Pakistan, Qatar, Saudi Arabia, Sudan, Syria, Tajikistan, Tunisia, and United Arab Emirates.
- Excluded due to lack of data: Afghanistan, Iraq, Somalia, Turkmenistan, Uzbekistan, West Bank and Gaza, and Yemen.

*Source: _wp10135 - 68.*

### 85.      Inflation was measured using CPI data from IMF IFS Statistics. Period average CPI

### _wp10135 - 85.      Inflation was measured using CPI data from IMF IFS Statistics. Period average CPI

### Data sources and measurement
- Inflation measured using CPI data from IMF IFS Statistics.
- Period average CPI data were used because there was greater coverage for this series.
- Tests were run using end-period CPI data and similar results were obtained.
- Commodity prices, import prices, deposit and lending rates, and exchange rates, and monetary aggregates were taken from IMF IFS Statistics.
- GDP data were taken from IMF WEO Statistics.
- The inflation series was tested to ensure that it was stationary and it was.

### Sample period and rationale
- Data cover the period 1996-2009.
- Rationale for period selection:
  - Includes a volatile period with a steep drop in inflation followed by a resurgence.
  - Prior to 1996 the data become patchier, which would reduce sample sets when adding variables.
  - Extending the sample earlier would require controlling results for reductions in the sample set or truncating periods to maintain comparability.
  - Significant developments in financial markets and financial deepening since the period prior to 1996 imply that going back too far might capture effects that are no longer relevant.

### Treatment of extreme inflation observations
- The inflation series used in most regressions incorporated a cap of 10 percent on inflation in any year for any country.
- Purpose: mitigate influence of countries with very high inflation episodes (notably Sudan and Tajikistan).
- Robustness checks:
  - Capping inflation at 10 percent (or at 15 percent, which was also tested) did not qualitatively change the results.
  - Capping resulted in somewhat more stable coefficients and somewhat diminished the coefficient on the exchange rate.

### Regression specification choices and key empirical findings
- Lag structure for commodity prices:
  - Lagged variables for commodity prices were used because they resulted in higher adjusted R-squareds in the regressions.
  - The coefficient on nonfuel commodity price changes was significant with a greater level of confidence when lagged commodity prices were used.
  - When unlagged commodity prices were used the coefficient on fuel prices was more negative.

*Source: IMF working paper content (sections 85–88 and associated references).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2010/_wp10135.pdf_