## _wp08270

## Source details

**Canonical URL:** [_wp08270](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08270.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08270.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08270.pdf.json)

---

### II. INTRODUCTION — context and analytical approach
- Paraguay avoided post-war hyperinflation episodes common in the region but experienced persistent moderate and volatile inflation.
- Two post-war high-inflation episodes:
  - early 1950s: inflation briefly crossed over the one hundred percent threshold;
  - early 1990s: inflation rose above forty percent during economic liberalization.
- Drivers of inflation volatility:
  - External: exposure as an exporter of primary commodities, location between large neighbors, spillovers from regional financial crises, world food and energy price increases.
  - Domestic: sharp gyration of a few agricultural product prices in 2006-07 led to year-long decoupling between headline and core inflation.
- Analytical approaches:
  - Mark-up theory of inflation (prices as function of domestic costs and imported prices).
  - Monetary approach (inflation from deviation of real money demand from long-term determinants).
- Data and estimation:
  - Quarterly data for 1991-2007.
  - Monetary aggregates evaluated range from currency in circulation to broad money.

### III. MEASURING INFLATION — headline, core, and rebasing
- Headline CPI presents noise (volatile components) and bias (substitution effects).
- Food items experienced significant supply shocks; food accounts for about 35-40 percent of the CPI basket.
- CPI rebasing:
  - Analysis sample used CPI based on the 1992 household budget survey (considered obsolete).
  - Since January 2008 BCP publishes new CPI based on 2005 household budget survey with less weight on food.
  - Had the weights of the new index been used in 2007, headline inflation could have been 1½ percent lower on average.
  - Fruits and vegetables weight: 6 ¼ percent (reduced to 5 percent in new index).
  - Food as share of CPI: between 35-40 percent of the CPI.
- Volatility examples:
  - Tomato prices (≈1 percent of CPI) movements could shift year-on-year headline CPI by 2-3 percent; August–December 2006 +350 percent, then halved until July 2007; November 2007 −72 percent after prior +68 percent; December 2007 +69 percent rebound.

### Core inflation indicators and empirical properties
- BCP exclusion-based indicators:
  - “CPIX”: headline excluding fruits and vegetables.
  - “CPIX1”: excludes fruits and vegetables, fuel items, and administrated-price products.
- Other indicators:
  - Non-food inflation (excludes all food products).
  - Trimmed indices (symmetric and asymmetric cuts); four trimmed indicators constructed (symmetric 10%/20%, asymmetric variants including a 10% asymmetric cutoff).
- Empirical findings summary:
  - Exclusion-based core indexes: slightly higher volatility than headline and biased downward; CPIX1 bias larger.
  - Trimmed mean indexes: much lower volatility; symmetric trimmed indexes show significant downward bias.
  - Asymmetric trimmed index excluding 10 percent: lower volatility and a small (positive) bias.
  - Forecasting: asymmetric indexes have coefficients with absolute value closer to unity (better predictors); symmetric trimmed indexes tend to underestimate future headline inflation; exclusion-based indexes tend to overestimate future headline inflation.
  - Granger causality: most variables Granger-cause headline inflation at longer horizons (3–6 months), except CPIX1 and non-food inflation.
- Recommendation:
  - An asymmetric trimmed index, preferably with a lower cutoff of 10 percent of the basket, could usefully complement BCP core indicators.
  - Of exclusion-based indexes, CPIX presents better statistical properties than CPIX1.
  - Non-food inflation is a relatively poorer core measure.

- Table 1 selected properties (1995-2007, year-on-year change):
  - Headline: Mean 8.8; Median 8.2; Std. dev. 3.6; Bias 0.0; RMSE with avg. (36 months): 3.6; RMSE with avg. (18 months): 2.4.
  - CPIX: Mean 8.5; Median 8.0; Std. dev. 3.7; Bias -0.3; RMSE (36): 3.5; RMSE (18): 2.3; attractor 0.17; attractor horizon 0.58; Granger p-values 0.53 (36), 0.07 (18), 0.00* (lags 3), 0.00* (lags 6).
  - CPIX1: Mean 7.8; Median 7.4; Std. dev. 3.7; Bias -1.0; RMSE (36): 3.7; RMSE (18): 2.5; attractor 0.06; attractor horizon 0.34; Granger p-values 0.38, 0.71, 0.03*, 0.05.
  - Non-food: Mean 8.3; Median 8.6; Std. dev. 3.9; Bias -0.5; RMSE (36): 3.2; RMSE (18): 2.5; attractor 0.12; attractor horizon 0.39; Granger p-values 0.62, 0.02*, 0.04*, 0.05.
  - Symmetric trimmed At 10 %: Mean 6.1; Median 5.5; Std. dev. 3.1; Bias -2.7; RMSE (36): 3.8; RMSE (18): 3.0; attractor 0.22; attractor horizon 0.87; Granger p-values 1.38, 0.05, 0.00*, 0.00*.
  - Symmetric trimmed At 20%: Mean 4.9; Median 4.4; Std. dev. 2.8; Bias -3.9; RMSE (36): 4.6; RMSE (18): 4.0; attractor 0.23; attractor horizon 0.90; Granger p-values 1.44, 0.03*, 0.00*, 0.00*.
  - Asymmetric trimmed At 10 %: Mean 9.0; Median 8.2; Std. dev. 3.4; Bias 0.2; RMSE (36): 3.2; RMSE (18): 2.0; attractor 0.19; attractor horizon 0.74; Granger p-values 0.80, 0.16, 0.00*, 0.00*.
  - Asymmetric trimmed At 20%: Mean 7.4; Median 6.9; Std. dev. 3.3; Bias -1.4; RMSE (36): 3.3; RMSE (18): 2.2; attractor 0.20; attractor horizon 0.78; Granger p-values 1.13, 0.04*, 0.00*, 0.00*.

### Analytical frameworks and econometric methodology
- Mark-up theory:
  - Formal: P = e^{μ} · ULC^{α} · P_{m}^{β} · P_{a}^{γ}; log-linear p = μ + α ulc + β p_m + γ p_a.
  - If 1 = α + β + γ, prices are weighted average of unit labor cost, other domestic inputs, and foreign prices.
  - Fuel and administrated prices included separately to capture direct/indirect effects.
- Monetary theory:
  - Money-market equilibrium and log-linear money demand yield long-run relation: m_t - p_t = β_1 y_t + β_2 i_t + ν_t and presented in source notation as p_t = β_1 m_t + β_2 y_t + β_3 i_t + ν_t.
  - Restrictions tested: money illusion (β_1 = 1) and constant velocity (β_1 = 1 and β_2 = -1 with ν_t constant).
- Econometric methodology:
  - Non-stationary variables; cointegration techniques applied.
  - Johansen (1991) VECM used to identify cointegrating relations and short-run dynamics.
  - DOLS (Stock and Watson, 1991) used for robustness of cointegrating coefficients.
  - Short-run specification in error-correction form Δx_t = α β' x_{t-1} + Σ Γ Δx_{t-i} + ν + ε_t with k = 2 (one difference term).

### Mark-up model — long-run (cointegrating) results
- Preferred dependent variable: core inflation CPIX1.
- Cointegrating results (price coefficient normalized to -1):
  - ULC coefficients:
    - Unrestricted: -0.79*; Restricted: -0.78*; DOLS: 0.77*.
  - CPI Brazil coefficients:
    - Unrestricted: -0.24*; Restricted: -0.22*; DOLS: 0.14*.
  - Loading coefficient:
    - Unrestricted: 0.05; Restricted: 0.05.
  - Probability of restriction: 0.63.
  - "*" denotes significance at the 5 percent level.
- Interpretation:
  - Unit labor cost (ULC) is strongly significant and main determinant of domestic prices; estimated ULC coefficient generally between ¾ to 0.9 across specifications.
  - Imported inflation from Brazil influences Paraguayan inflation; Brazilian CPI coefficient stable around ¼–⅓ in preferred VECM specs; DOLS gives 0.14.
  - U.S. dollar exchange rate and nominal effective exchange rate not significant in mark-up regressions.
  - Baseline interpretation: about 80 percent of core inflation (CPIX1) explained by growth in labor costs and remaining 20 percent by imported inflation from Brazil (CPI Brazil).
- Sensitivity:
  - Using core tradable prices makes Brazilian CPI coefficient similar weight as ULC (~0.8); homogeneity constraint rejected.
  - Oil and administrated prices not significant in mark-up regressions.

### Mark-up model — short-term dynamics and error-correction performance
- Cointegrating vector used: CPIX1 - 0.88 ULC - 0.22 CPI Brazil.
- Short-run entries (selected; Table 3 summary):
  - Loading coeff. (CPIX1): 0.06.
  - ULC short-term coefficient: 0.29* (Std.dev. -0.06) with R2 0.67.
  - CPI Brazil short-term coefficient: 0.17 (Std.dev. -0.10) with R2 0.38.
  - "*" denotes significance at the 5 percent level.
- Performance:
  - Model explains less than half of short-term inflation dynamics; ECM for prices performs poorly.
  - Loading coefficient often not significant or has wrong sign for prices, but is significant with correct sign for ULC dynamics.
  - Implication: cointegrating vector may be more relevant for wage equation than for CPI equation; mechanism could be stronger link from prices to wages than wages to prices.
- Institutional mechanism:
  - Minimum wage adjusts automatically when cumulative inflation reaches 10 percent after last adjustment; can amplify wage sensitivity to past price movements.

### Monetary approach — cointegration and money demand findings
- VAR setup and lag choices:
  - Information criteria suggest one lag in difference (k = 2) though residuals often better with longer VAR.
  - For VAR with two lags: eigenvalue and trace tests indicate one cointegration vector; dynamics driven by two stochastic trends (prices and real income) and one cointegrating vector interpretable as money demand.
- Currency-in-circulation money demand (key estimates):
  - Income elasticity of currency demand higher than Baumol-Tobin prediction of 0.5 and higher than quantity theory prediction of 1; examples from other countries cited (Albania 2.3; Armenia 2.7; Russia 1.8).
  - Best-performing opportunity-cost proxy: backward year-on-year depreciation of the guaraní with respect to the U.S. dollar.
- Table 4 (currency-in-circulation model) reported coefficients:
  - Currency in circ. coefficient: -1.08* (Unrestricted), -1(c) (Constrained 1/), -1(c) (Constrained 2/).
  - Output coefficient: 1.97*, 1.50*, 1.65*.
  - Opportunity cost coefficient: -0.65*, -0.75*, -0.69*.
  - Loading coeff.: -0.11*, -0.10*, -0.10*.
  - Probability of restriction: 0.48, 0.21.
  - Notes: 1/ Restriction on coefficient of currency. 2/ Also weak exogeneity restriction. "*" denotes significance at the 5 percent level.
- Narrow vs broad money:
  - M1 and M2 estimates weaker and often suggest a second cointegration relation; instability partly explained by banking crises (1995, 1997, 1998, 2002, 2003).
  - M3 (broad money) gives coefficient of money close to unity; income coefficient close to unity when opportunity cost measured as spread within M3 and LRM; unit elasticity of income accepted → velocity stationary.
  - LRM (policy sterilization paper) significance suggests policy rate impacts portfolio allocation.
- Loading coefficients and half-lives (Table 6 comparisons):
  - Currency in circulation: Loading coefficient 0.07*; Half life 4-7 (quarters); Price homogeneity Yes; Constant velocity No; Income coefficient 1.5*.
  - M1: Loading coefficient -0.13*; Half life 13-25; Price homogeneity Yes; Constant velocity No; Income coefficient -1.6*.
  - M2: Loading coefficient 0.02*; Half life 25; Price homogeneity Yes; Constant velocity No; Income coefficient 1.6*.
  - M3: Loading coefficient -0.04*; Half life 25; Price homogeneity Yes; Constant velocity Yes; Income coefficient 0.6*.
  - Opportunity cost proxies: Currency in circulation — Depreciation with U.S. dollar; M3 — Sterilization paper (LRM).
- Why inflation did not immediately rise despite high currency growth:
  - Since 2005 excess currency growth over nominal GDP raised concerns.
  - Appreciation of the guaraní by about 20 percent in 2008 partially offset currency growth effects by increasing demand for real balances.
  - Table 7 (determinants of inflation in 2007 and 2008, Average Q1-Q3 for 2008) (preserved entries):
    - Currency growth: 2007: 124.2 (Growth) → Contrib. 24.2; 2008: 32.4 (Growth) → Contrib. 32.4.
    - Real income: 2007: -1.65 → Contrib. 6.4; 2008: -10.64 → Contrib. 4.4 → Contrib. -7.3.
    - Opportunity cost: 2007: 0.69 → Contrib. -9.8; 2008: -6.8 → Contrib. -12.7 → Contrib. -8.8.
    - Expected inflation: 2007: 6.8 → Contrib. 16.4.
    - Actual inflation: 2007: 5.3; 2008: 10.5.
  - Cointegrating vector used: m - p - 1.6 y + 0.7 r.

### Short-term determinants of core inflation — VECM final specification and dynamics
- Modeling approach:
  - Combine long-run money-demand relation (with price homogeneity and weak exogeneity) and mark-up long-run relation into a VECM for CPIX.
  - Start with two-lag VECM, reduce to significant variables, add exogenous variables for the “final” equation.
- Table 8 — Initial vs Final specification (selected significant coefficients):
  - Loading coefficients:
    - Money demand equation: -0.05*** (Initial), -0.04*** (Final).
    - Mark-up equation: 0.04.
  - Short run coefficients (Final):
    - d Price (-1): 0.42***.
    - d Price (-2): -0.22**.
    - d ULC: 0.17**.
    - d CPI Brazil (-1): 0.03*.
    - d Adm prices: 0.10**.
    - d Fruits and veg.: 0.02***.
    - Constant: 0.50***.
  - Fit and diagnostics:
    - R2: 0.65 (Initial), 0.74 (Final).
    - Log likelihood: 208.8 (Initial), 218.4 (Final).
    - Durbin-Watson: 2.02 (Initial), 1.86 (Final).
    - Akaike: -6.25 (Initial), -6.51 (Final).
    - Schwartz: -5.99 (Initial), -6.17 (Final).
  - Notes and interpretation:
    - First autoregressive coefficient ~0.49; second ~-0.30.
    - Contemporaneous change in ULC significant; ULC may anchor inflation inertia and raise input costs.
    - Exchange rate with Brazil matters in short run; Brazilian quarterly price changes enter with one quarter lag.
    - Administrated prices significant with estimated coefficient 0.10 vs share in CPIX of 8.3 percent.
    - Fruits and vegetables have modest effect on core inflation dynamics despite exclusion from core CPI.

### Model performance, forecasts, and dynamic contributions
- Model explains about 3/4 of inflation variance over the sample.
- Loading coefficient to cointegrating relation reduced in extended model: -0.04 (extended) vs -0.07 (restricted VECM); up to -0.13 in some unrestricted specs.
- Out-of-sample 2008 forecast: model captures uptick in core inflation but underestimates actual by about one percentage point.
- Dynamic decomposition findings:
  - Bulk of inflation dynamics driven by evolution of currency in circulation given model structure.
  - Currency in circulation contributes less to short-term spikes (e.g., 1999, 2003, 2005) which are not directly related to currency in circulation.
  - Since 2004:
    - Currency in circulation contributed by 10 to 12 percent to inflation.
    - Buoyant money demand related to transaction motive (real income) added negative impact increasing from 2 to 5 percent since 2004.
    - Appreciation of the guaraní reduced inflation by 2.5 percent on average.
  - Exchange rate contributions in 2002-03:
    - Depreciation against U.S. dollar added about 3.2 percent to inflation (opportunity cost channel).
    - Depreciation against Brazilian real added about 0.7 percent (imported goods channel).

### Key econometric and empirical findings — concise points
- Money talks:
  - Currency in circulation is the most relevant monetary aggregate for inflation dynamics.
  - An increase of currency in circulation by 1 percent leads to an increase of inflation by 0.05 points after one quarter, and by 1 percent in the long-run.
  - Income coefficient > 1 in money demand suggests trend increased monetization.
  - Quantitative theory of money holds for broad money (long-run) but has little short-term relevance.
- Pass-through matters:
  - Short run: 1 percent increase of imported inflation from Brazil adds 0.03 percent to Paraguay inflation.
  - Long run: exchange rate with U.S. dollar is key for portfolio allocation and money demand; depreciation with respect to U.S. dollar may capture some pass-through to prices.
- Food feeds inflation:
  - Food ≈ 35 percent of CPI; food prices have spillover effects on non-food prices; Granger-causality tests indicate final impact larger than weight.
- Inertia strong:
  - A 1 percent shock to inflation in a quarter gives additional 0.4 percent after one quarter and 0.2 percent after two quarters.
  - Wage indexation mechanism may lock in inflation increases; inertia linked to expectations and policy credibility.

### Risks, policy implications, and instruments
- Risks for 2008-09 inflation outlook:
  - Reversal of appreciation of the guaraní vs U.S. dollar.
  - Further depreciation vs Brazilian real.
  - End of moderation of administrated prices.
- Monetary policy effectiveness and instruments:
  - Currency in circulation is the aggregate most easily controlled by the Central Bank and the best predictor of inflation.
  - Controlling currency growth via emission of sterilization paper (LRM) moderates inflationary impact.
  - Policy interest rate significant for broad money portfolio allocation.
  - LRM rose from < 1/10 of currency in circulation in 1998 to close to ¼ of currency in circulation by end-2007.

### CPI rebasing, micro findings, and data notes
- New CPI (base month Dec. 2007) vs old (Dec. 1992) — selected changes:
  - Food: 35.3 → 32.0.
  - Fruits and vegetables: 6.2 → 5.1.
  - Meat: 10.7 → 10.4.
  - Clothing: 8.6 → 4.9.
  - Transport: 7.4 → 14.8.
  - Communication: 0.7 → 3.4.
  - Fuel: 3.6 → 5.8.
  - Administrated prices: 7.7 → [7.0].
- Implication: new index from January 2008 should lead to lower measured inflation; had new weights been used in 2007 headline inflation could have been 1.5 percent lower on average.
- Micro-level statistics (1994-2007, using 1992 basket) selected:
  - Food: Number items 91; Weight in CPI 39.0; Share of tradables 89; Average monthly change 0.80; Standard deviation 2.02.
  - Fruits and vegetables: Items 17; Weight 6.2; Average monthly change 1.35; Standard deviation 9.20.
  - Administrated prices: Items 8; Weight 7.7; Average monthly change 0.90; Standard deviation 1.96.
  - Core inflation (CPIX): Items 276; Weight 93.8; Average monthly change 0.70; Standard deviation 0.67.
  - Core inflation (CPIX1): Items 262; Weight 82.5; Average monthly change 0.65; Standard deviation 0.67.
  - All items: Items 293; Weight 100.0; Average monthly change 0.72; Standard deviation 1.0.
- Granger-causality: food inflation is a predictor of non-food inflation in Paraguay (robust to lag selection).
- Data and methodological notes:
  - ULC = average wage index minus labor productivity; wages available semi-annually, minimum wages monthly, employment annual with lag; quarterly series extrapolated using linear relationships and Chow-Lin (1971).
  - Price of foreign inputs proxied by NEER, exchange rate with US$, and CPI of Argentina and Brazil (converted to guaranies).
  - Monetary aggregates defined: Currency in circulation (CC); M1 = CC + cash in vault + demand deposits; M2 = M1 + time and saving deposits + certificates of deposits; M3 = M2 + dollar deposits.
  - Opportunity cost measures include depreciation-based proxies and deposit returns; 100 bp is 0.01 as expressed in source.

*Source: _wp08270 (IMF working paper content provided in the supplied PDF excerpt).*

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

### _wp08270 - References..............................................................................................................

### II. INTRODUCTION
- Paraguay evaded the bouts of hyperinflation that affected most countries in Latin America over the post-war period, but, as inflation started to decline in neighboring countries, it remained afflicted with moderate inflation.
- In the region, it shares these characteristics of never experiencing three digit inflation over the past fifty years only with Colombia.
- Paraguay only experienced two episodes of high inflation in the post-war period:
  - in the early 1950s inflation briefly crossed over the one hundred percent threshold during the civil war period;
  - in the early 1990s economic liberalization and the opening up of the economy sent inflation shooting up above forty percent.
- Over the past ten years, inflation in Paraguay has been on average higher than in most other countries in the region.
- Inflation has been highly volatile; drivers include:
  - external factors related to Paraguay’s position as an exporter of primary commodities (subject to related supply shocks);
  - location between two large neighbors and exposure to spillovers from regional financial crises;
  - world increase of food and energy prices.
- Domestic components to inflation volatility are also important:
  - in 2006-07, the sharp gyration of a few agricultural product prices led to year-long decoupling between headline and core inflation.
- Analytical approaches used in the paper:
  - mark-up theory of inflation: prices modeled as a function of domestic costs and imported prices;
  - monetary approach: inflation results from the deviation of real money demand from its long-term determinants.
- Monetary aggregates evaluated range from currency in circulation to broad money to assess robustness.
- All econometric models are estimated using quarterly data for 1991-2007.

### III. MEASURING INFLATION
- Headline inflation (CPI) presents both noise (due to volatile components) and bias (due to substitution effects).
- Food items have experienced significant supply shocks, contributing to overall inflation beyond their share in the basket.
- Paraguay is not a full-fledged inflation targeting country as inflation is only one among many objectives; since 2003, the Central Bank has announced an end-year target for inflation with a broad band of +/- (text truncated in source).

### KEY RESULTS AND INTERPRETATIONS
- Long-run determinants:
  - Monetary factors, in particular currency in circulation, have played a major role in determining long-run inflation.
  - The exchange rate with respect to the U.S. matters for currency demand in the long-run.
- Short-run determinants:
  - Foreign prices, in particular from Brazil, and some food products have had a large impact on the short-term dynamics of inflation.
  - The exchange rate with respect to Brazil matters for cost-push inflation in the short-run.
  - The evolution of some food prices matters beyond their large share in the CPI basket, possibly because wage indexation—or expectations of continued moderate inflation—tends to lock up price increases.
- Methodological scope:
  - Paper focuses on determinants of core inflation indicators given high headline volatility.
  - Two different analytical perspectives (mark-up and monetary approaches) are used to assess long-term determinants and then encompass short-term dynamics.

### STRUCTURE OF THE PAPER (as stated in source)
- Section II discusses headline and core inflation indexes.
- Sections III and IV investigate long-run determinants of inflation using, respectively:
  - the analytical framework of the mark-up theory; and
  - the monetary approach.
- Section V presents the short-term inflation dynamics.
- Section VI includes concluding remarks.
- Annexes listed in the source include analyses spanning inflation history, micro-level price analysis, the new CPI, food inflation in Latin America, data issues and database used in econometric analysis, and econometric results.

*Source: _wp08270 - References..............................................................................................................*

### 2.5 percentage points.

### _wp08270 - 2.5 percentage points.

### Inflation measurement and CPI rebasing
- Sample period for analysis: 1991-2007, using a CPI based on the 1992 household budget survey considered obsolete.
- Since January 2008, the BCP started publishing a new CPI based on the 2005 household budget survey.
- The new basket gives less weight to food items. Had the weights of the new index been used in 2007, headline inflation could have been 1½ percent lower on average.
- Fruits and vegetables weight: 6 ¼ percent of the CPI basket (reduced to 5 percent in the new index).
- Food as share of CPI: between 35-40 percent of the CPI.

### Food components and their inflation dynamics
- Fruits and vegetables
  - Represent 6 ¼ percent of the CPI basket (5 percent in the new index).
  - Coefficient of variation about 5 times larger than that of the index in general.
  - Political interference since 2006 affected volatility (periodic imposition/removal of sanitary controls).
  - Tomato example: despite representing about 1 percent of the CPI, wide tomato price variations could move year-on-year headline CPI by 2-3 percent.
  - Tomato price volatility examples: between August and December 2006 price increased by 350 percent, then reduced by half until July 2007; November 2007 saw a collapse by 72 percent after a 68 percent gain in previous three months; December 2007 rebounded by 69 percent.
- Meat products
  - Represent 10¾ percent of the CPI basket (relatively unchanged in the new index).
  - Beef constitutes about 4/5 of meat products.
  - Domestic meat prices influenced by international beef prices and foot-and-mouth disease (FMD) episodes regionally.
  - Beef prices contributed on average by 2 percent to annual inflation between 2003 and 2006.
  - Chronology of regional FMD impacts:
    - Starting in 2000 regional outbreak hit Paraguayan beef producers; in 2002 several outbreaks led to closure of Paraguayan export markets.
    - In 2005 Paraguay recovered “FMD-free with vaccination” status; combined with export restrictions from Argentina, beef components of the CPI index increased by 140 percent between July 2002 and July 2006.
    - In 2007 reopening of Brazilian markets shifted demand away from Paraguay and contributed to a decline in early 2007; 2008 saw increases in line with overall inflation.
- Other food items
  - Represent 18.4 percent of the basket (16.5 percent in the new basket).
  - Historically evolved in line with non-food prices.
  - Starting mid-2007, milk and bread products increased worldwide; these represent about 8-10 percent of the basket and increased by around 30 percent between June 2007 and June 2008.
  - Milk price increases partly due to a supply effect from a severe drought in the Chaco region of Argentina and Paraguay.

### Energy, administrated prices, and their role
- Energy products represent about 3½ percent of the basket (increased to 5¾ percent in the new basket).
- Domestic gasoline and natural gas prices follow international trends with a lag; diesel price is subject to occasional subsidies as the government tries to stabilize diesel, creating sharp-step movements.
- Administrated prices reflect the role of oil as an input for public transport; until mid-2007 energy and administrated prices contributed to lower headline inflation, but since mid-2007 administrated prices have increased in line with overall inflation while fuel products increased by twice as much.

### Tradable vs non-tradable inflation dynamics
- Non-tradable goods
  - During the first half of the sample, non-tradable inflation tended to be consistently higher than average core inflation across categories (administrated prices, education, health, other services).
  - Since 2002, non-tradable inflation has been consistently lower, even excluding administrated prices—possibly reflecting lower real wages in the public sector since 2001 and a trend decline in unit labor costs.
- Tradable goods
  - Spikes in inflation in 1998, 2003, and 2006 driven primarily by tradable goods.
  - Paraguay experienced significant depreciation with respect to the U.S. dollar in 1998 and 2003.
  - In 2006 depreciation with respect to the U.S. dollar was relatively mild but the guarani depreciated with respect to the real; higher fuel prices also contributed to the 2006 episode.

### Core inflation measures described
- BCP exclusion-based core indicators:
  - “CPIX”: headline inflation excluding fruits and vegetables.
  - “CPIX1”: excludes fruits and vegetables, fuel items, and products subject to administrated prices.
- Other suggested core indicators:
  - Simple non-food inflation index (excludes all food products).
  - Trimmed indices: exclude items according to monthly volatility rather than by category; four trimmed indicators constructed:
    - Symmetric trimmed excluding 10 percent most volatile and 10 percent least volatile.
    - Symmetric trimmed excluding 20 percent most and 20 percent least.
    - Asymmetric trimmed excluding ⅓ most volatile and ⅔ least volatile (two variants at 10 percent and 20 percent asymmetric cutoffs implied).
- Remarks on excluding all food: radical due to food representing between 35-40 percent of CPI.

### Properties of core inflation indicators (summary of empirical findings)
- Over the sample period all exclusion-based core indexes show slightly higher volatility than headline inflation and are biased downward relative to headline; bias greater for CPIX1 due to larger increases in administrated prices and fuel (they increased by 11.5 and 16.3 percent respectively, against 7.8 percent for other prices).
- Trimmed mean indexes show much lower volatility but symmetric trimmed indexes exhibit significant downward bias.
- The asymmetric trimmed index excluding 10 percent of items has both lower volatility and a small (positive) bias.
- Medium-term proxy performance (RMSE with moving average):
  - Non-food inflation and the asymmetric index at 10 percent have lower RMSE when medium-term inflation is measured as a moving average of 36 months.
  - For a moving average of 18 months, only asymmetric indexes have significantly lower RMSE.
- Forecasting properties:
  - Asymmetric indexes present coefficients with absolute value closer to unity (better predictor of future headline inflation).
  - Symmetric trimmed indexes tend to underestimate future headline inflation.
  - Exclusion-based indexes tend to overestimate future headline inflation, and by more than the asymmetric index.
- Granger causality:
  - Most variables tend to Granger-cause headline inflation at longer horizons (3 to 6 months), except CPIX1 and non-food inflation which do not show this predictive power.
- Table 1 (properties over 1995-2007, year-on-year change) key figures (preserved exactly as presented):
  - Headline: Mean 8.8; Median 8.2; Std. dev. 3.6; Bias 0.0; RMSE with avg. (36 months): 3.6; RMSE with avg. (18 months): 2.4.
  - CPIX: Mean 8.5; Median 8.0; Std. dev. 3.7; Bias -0.3; RMSE (36): 3.5; RMSE (18): 2.3; others: attractor 0.17; attractor horizon 0.58; Granger causality p-values 0.53 (36 months), 0.07 (18 months), 0.00* (lags 3), 0.00* (lags 6).
  - CPIX1: Mean 7.8; Median 7.4; Std. dev. 3.7; Bias -1.0; RMSE (36): 3.7; RMSE (18): 2.5; attractor 0.06; attractor horizon 0.34; Granger causality p-values 0.38, 0.71, 0.03* ,0.05.
  - Non-food: Mean 8.3; Median 8.6; Std. dev. 3.9; Bias -0.5; RMSE (36): 3.2; RMSE (18): 2.5; attractor 0.12; attractor horizon 0.39; Granger p-values 0.62, 0.02* ,0.04* ,0.05.
  - Symmetric trimmed At 10 %: Mean 6.1; Median 5.5; Std. dev. 3.1; Bias -2.7; RMSE (36): 3.8; RMSE (18): 3.0; attractor 0.22; attractor horizon 0.87; Granger p-values 1.38, 0.05, 0.00*, 0.00*.
  - Symmetric trimmed At 20%: Mean 4.9; Median 4.4; Std. dev. 2.8; Bias -3.9; RMSE (36): 4.6; RMSE (18): 4.0; attractor 0.23; attractor horizon 0.90; Granger p-values 1.44, 0.03*, 0.00*, 0.00*.
  - Asymmetric trimmed At 10 %: Mean 9.0; Median 8.2; Std. dev. 3.4; Bias 0.2; RMSE (36): 3.2; RMSE (18): 2.0; attractor 0.19; attractor horizon 0.74; Granger p-values 0.80, 0.16, 0.00*, 0.00*.
  - Asymmetric trimmed At 20%: Mean 7.4; Median 6.9; Std. dev. 3.3; Bias -1.4; RMSE (36): 3.3; RMSE (18): 2.2; attractor 0.20; attractor horizon 0.78; Granger p-values 1.13, 0.04*, 0.00*, 0.00*.
- Summary recommendation from the section:
  - An asymmetric trimmed index, preferably with a lower cutoff (of 10 percent of the basket), could usefully complement the core inflation indicators currently used by the BCP.
  - Of the exclusion-based core indexes used by the BCP, CPIX presents better statistical properties than the narrower CPIX1.
  - Non-food inflation is a relatively poorer core inflation index because it tends to present higher volatility, a larger bias than CPIX, or less predictive power for future inflation at large horizons.

### Analytical and empirical framework to assess determinants of inflation
- Two theoretical frameworks used:
  - (i) Markup theory of inflation: consumer prices as a mark-up over producer’s costs.
    - Formal expression: P = e^{μ} · ULC^{α} · P_{m}^{β} · P_{a}^{γ}
    - Log-linear form: p = μ + α ulc + β p_m + γ p_a
    - If 1 = α + β + γ (linear homogeneity), prices are a weighted average of unit labor cost, other domestic inputs, and foreign prices.
    - Including fuel and administrated prices separately captures direct and indirect effects (spillovers).
  - (ii) Monetary theory of inflation: inflation driven by excess money supply over money demand.
    - Money market equilibrium: (M/P)_s = (M/P)_d
    - Log-linear money demand yielding testable long-run relation: m_t - p_t = β_1 y_t + β_2 i_t + ν_t (presented as equation (5): p_t = β_1 m_t + β_2 y_t + β_3 i_t + ν_t in the source notation).
    - Restrictions tested include money illusion (β_1 = 1) and constant velocity (β_1 = 1 and β_2 = -1 with ν_t constant).
- Econometric methodology:
  - Variables are non-stationary; cointegration techniques used.
  - Cointegrated VAR (VECM) of Johansen (1991) used to determine cointegration relations, distinguish stochastic trends and cointegration, and encompass both long-run and short-run dynamics.

*Source: _wp08270 - 2.5 percentage points.*

### Annex Table V.1 describes the variables used in the regression and Annex Table V.2 presents the results of

### _wp08270 - Annex Table V.1 describes the variables used in the regression and Annex Table V.2 presents the results of

### Stationarity, cointegration, and estimation methodologies
- Annex Table V.2 reports ADF and Schmidt-Phillips stationarity tests: the variables of interest are generally non stationary in levels but stationary in first differences.
- Cointegration technique references: Juselius (2005) discussed step-by-step application to a money demand model for Denmark.
- Estimation approaches used:
  - Johansen procedure to identify cointegrating relations and the short term dynamics.
  - Dynamic OLS (DOLS) methodology of Stock and Watson (1991) to check robustness; DOLS gives consistent estimates of the coefficient of a cointegration relation but does not test the existence of cointegration.
- Short-run specification (error-correction form):
  - General short-run equation shown as equation (6): Δx_t = α β' x_{t-1} + Σ Γ Δx_{t-i} + ν + ε_t (lag structure k = 2 implies a single term in difference for short term dynamics).
  - x_t is the vector of endogenous variables; ν captures short-run or long-run constants/trends.
  - Long-run error-correction term: β' x_{t-1}; α is the loading coefficient (speed of adjustment).
- Interpretation of α (with price coefficient normalized to 1 in the cointegration vector):
  - Negative and significant α implies error-correction: deviations of price from long-run relation lead to contraction of prices next period and return to long-term value.
  - Positive α implies explosive dynamics: deviations are self-reinforcing.
- Sample and estimation strategy:
  - Model estimated on quarterly data over 1991Q1 to 2007Q3 using a two-stage estimation strategy.
  - Two long-run relationships (2) and (5) estimated independently in a cointegrating setting and then combined to study short-term dynamics.
  - Two-stage approach noted as less efficient but can yield estimates closer to theoretical assumptions; similar approaches cited in Sekine (2001), Oomes and Ohnsorge (2005), Sacerdoti and Xiao (2001).

### Mark-up model — long-run (cointegrating) results (Table. 2 summary)
- Model specification and estimation:
  - Results reported with a lag structure of 3 quarters for the VAR; robustness tested with different lag structures.
  - Preferred dependent variable: core inflation index CPIX1.
- Unrestricted / Restricted / DOLS cointegrating vector (coefficient of price CPIX1 normalized to -1):
  - ULC coefficients reported as:
    - Unrestricted: -0.79*
    - Restricted: -0.78*
    - DOLS: 0.77*
  - CPI Brazil coefficients reported as:
    - Unrestricted: -0.24*
    - Restricted: -0.22*
    - DOLS: 0.14*
  - Loading coefficient reported as:
    - Unrestricted: 0.05
    - Restricted: 0.05
  - Probability of restriction: 0.63
  - Note: "*" denotes significance at the 5 percent level.
- Interpretation of long-run coefficients:
  - Unit labor cost (ULC) is strongly significant and the main determinant of domestic prices.
  - Estimated ULC coefficient generally between ¾ to 0.9 across specifications; considered on the high side relative to many other country studies.
  - Imported inflation proxying:
    - Consumer prices in Brazil (CPI Brazil) influence inflation in Paraguay more than other imported-inflation indicators.
    - U.S. dollar exchange rate and nominal effective exchange rate are not significant.
    - Brazilian CPI coefficient stable around ¼--⅓ in preferred VECM specifications.
  - Possible explanation: larger share of imports from Brazil and higher economic integration with Brazil (especially in the east of the country).
  - Robustness: DOLS tends to give a lower coefficient for Brazilian inflation (0.14).

### Mark-up model — further findings and sensitivity
- Sensitivity to inflation measure:
  - Using core tradable prices instead of core prices: coefficient of Brazilian CPI has same weight as ULC, around 0.8; homogeneity constraint is rejected in that specification.
  - Using CPIX1 as dependent variable: homogeneity constraint is accepted at high significance level; using CPIX as dependent variable: constraint is not accepted.
- Constraints and model performance:
  - The linear homogeneity restriction is accepted for baseline model.
  - Baseline interpretation: about 80 percent of core inflation (CPIX1) explained by growth in labor costs (ULC) and remaining 20 percent by imported inflation from Brazil (CPI Brazil).
  - Estimating with DOLS reduces coefficient on Brazilian inflation to 0.14.
- Non-significant variables:
  - Oil and administered prices are not significant in the mark-up regressions; this lack of significance is robust to alternative modeling and price indicators.

### Short-term dynamics and error-correction performance (Table 3 summary)
- Short-term dynamics reported with cointegrating vector: CPIX1 - 0.88 ULC - 0.22 CPI Brazil.
- Table 3 entries (with cointegrating vector above):
  - Loading coeff. (CPIX1): 0.06
  - R2 (CPIX1): -0.05
  - Coeff./Std.dev. (CPIX1): 0.41
  - ULC coefficient in short term: 0.29* with Std.dev. -0.06 and R2 0.67
  - CPI Brazil short-term coefficient: 0.17 with Std.dev. -0.10 and R2 0.38
  - Note: "1/ with cointegrating vector: CPIX1 - 0.88 ULC - 0.22 CPI Brazil"
  - Note: "*" denotes significance at the 5 percent level.
- Overall short-run performance:
  - The model explains less than half of the short-term dynamics of inflation (performs poorly as an ECM for prices).
  - Loading coefficient of the cointegrating vector in short-term price dynamics is usually not significant (or significant with wrong sign), implying deviations from long-term determinants have little short-term influence on prices.
  - By contrast, the loading coefficient is highly significant with correct sign for ULC dynamics; model explains short-term dynamics of ULC better.
  - Implication: identified cointegrating vector may be more relevant for a wage equation than for a CPI equation; possible stronger link from prices to wages rather than wages to prices.
- Institutional mechanism note:
  - Minimum wage adjustment mechanism: automatic when cumulative inflation reaches 10 percent after last adjustment; this can amplify wage sensitivity to past price movements.
  - Minimum wage: direct influence mainly in formal sector but may serve as a reference wage affecting other labor costs.

### Monetary theory of inflation (introductory remarks)
- Inflation measure in monetary section: core inflation CPIX.
- VAR specification: based on judgment and statistical tests; lag structure chosen by judgment rather than only information criteria.
- Preliminary results:
  - In an unconstrained model, the coefficient of currency is close to unity while coefficient of income is significantly above unity.
  - Constraint of long run price homogeneity (unit coefficient of money) cannot be rejected at reasonable significance levels (no money illusion).
  - Constant velocity assumption is narrowly rejected; the vector resulting from constraints is not interpreted as a cointegrating vector by the trace test (further details follow in the full monetary section).

*Source: _wp08270 - Annex Table V.1 describes the variables used in the regression and Annex Table V.2 presents the results of*

### Annex tables VI.3 and VI. 4 report the full detail of the regressions using different definitions of monetary

### _wp08270 - Annex tables VI.3 and VI. 4 report the full detail of the regressions using different definitions of monetary

### VAR specification, lag selection, and cointegration
- Information criteria (AIC and BIC) suggest the use of only one lag in difference (k = 2).
- Residuals of the VAR often have better statistical properties with a longer VAR; some coefficients of the second difference are significant.
- Number of cointegrating vectors usually not affected by lag structure, but the income coefficient is consistently higher and beyond the reasonable with a shorter lag structure.
- For a VAR with two lags:
  - Both eigenvalue and trace tests indicate the existence of one cointegration vector.
  - Dynamics are driven by two stochastic trends (one for prices and one for real income) and one cointegration vector interpretable as a money demand relation.

### Currency-in-circulation money demand (key estimates and robustness)
- Estimated income elasticity of currency demand:
  - Higher than Baumol-Tobin prediction of 0.5.
  - Higher than quantity theory prediction of 1.
  - In line with other empirical studies (examples: Albania 2.3; Armenia 2.7; Russia 1.8).
- Opportunity cost proxies tested:
  - Best-performing proxy: backward year-on-year depreciation of the guaraní with respect to the U.S. dollar.
  - Quarterly backward depreciation or forward depreciation (perfect foresight) models are less robust.
  - When both annual backward depreciation and return on guaraní time deposits are introduced, cointegration tests suggest one or no cointegrating relation; some coefficients become insignificant or have unexpected signs.
- Recursive/rolling estimation:
  - Currency demand model broadly stable for the past three years; unstable when 2004 is excluded.
  - Excluding 2005-07 has little impact on coefficients.
  - Instability when 2004 excluded may be due to high variability of the exchange rate around that period.
- Table 4 (Model with currency in circulation) — reported coefficients:
  - Cointegrating vector (CPIX111)
  - Currency in circ. coefficient: -1.08* (Unrestricted), -1(c) (Constrained 1/), -1(c) (Constrained 2/)
  - Output (income) coefficient: 1.97*, 1.50*, 1.65*
  - Opportunity cost coefficient: -0.65*, -0.75*, -0.69*
  - Loading coeff.: -0.11*, -0.10*, -0.10*
  - Proba. of restriction: 0.48, 0.21
  - Notes: 1/ Restriction on coefficient of currency. 2/ Also weak exogoneity restriction.

### Narrow money (M1, M2) vs. broad money (M3) results
- General pattern:
  - Estimates for M1 and M2 are generally weaker than for currency in circulation due to significant and sharp variations of these aggregates over the sample period.
  - Test statistics often suggest the existence of a second cointegration relation for M1/M2, difficult to interpret.
  - Banking crises during the sample period (1995, 1997, 1998, 2002, 2003) may explain instability in deposit-based aggregates (M1, M2) because deposits can be lost in crises while cash remains safe.
- M3 (broad money) results:
  - Estimation of M3 gives a coefficient of money close to unity.
  - Income coefficient for M3 is also close to unity (when opportunity cost is measured as spread within M3 and LRM).
  - Unit elasticity of income accepted in this specification → velocity of money is stationary.
  - Significance of LRM suggests monetary policy rate has impact on public’s decision to hold broad money, but:
    - Loading factor significant but with the wrong sign → equilibrium interpretation difficult.
    - Eigenvalue test suggests existence of another cointegration relation.
    - LRM not widely available during first half of sample period.
- Loading coefficients and half-lives across aggregates (Table 6 comparison):
  - Currency in circulation: Loading coefficient 0.07*; Half life of deviation (in quarters) 4-7; Price homogeneity Yes; Constant velocity No; Income coefficient 1.5*.
  - M1: Loading coefficient -0.13*; Half life 13-25; Price homogeneity Yes; Constant velocity No; Income coefficient -1.6*.
  - M2: Loading coefficient 0.02*; Half life 25; Price homogeneity Yes; Constant velocity No; Income coefficient 1.6*.
  - M3: Loading coefficient -0.04*; Half life 25; Price homogeneity Yes; Constant velocity Yes; Income coefficient 0.6*.
  - Notation: 1/ When with correct sign. 2/ In model with price homogeneity.
  - Relevant opportunity cost entries (as reported): Currency in circulation — Depreciation with U.S. dollar; M3 — Sterilization paper (LRM).

### Why inflation did not immediately pick up despite high currency growth
- Since 2005 excess growth of currency in circulation over nominal GDP raised concerns about inflationary pressures.
- Long-run currency demand relation captures:
  - Monetization effect via high income coefficient.
  - Appreciation of the exchange rate (with respect to the dollar) making local currency more attractive.
- Table 7 — Determinants of inflation in 2007 and 2008 (Average Q1-Q3 for 2008):
  - Currency growth:
    - In 2007: 124.2 (Growth) → Contrib. 24.2
    - In 2008: 32.4 (Growth) → Contrib. 32.4
  - Real income:
    - In 2007: -1.65 → Contrib. 6.4
    - In 2008: -10.64 → Contrib. 4.4 → Contrib. -7.3 (table shows both growth and contrib; preserved as in source)
  - Opportunity cost:
    - In 2007: 0.69 → Contrib. -9.8
    - In 2008: -6.8 → Contrib. -12.7 → Contrib. -8.8
  - Expected inflation:
    - In 2007: 6.8 → Contrib. 16.4
  - Actual inflation:
    - In 2007: 5.3
    - In 2008: 10.5
  - Note: appreciation of the guaraní by about 20 percent in 2008 partially offset currency growth effects.
- Model implication:
  - Appreciation increased demand for real balances in local currency, limiting inflationary impact of higher currency growth.
  - Model predicts currency growth higher than warranted to avoid feeding inflationary pressures.
- Cointegrating vector used (Figure 4b): m - p - 1.6 y + 0.7 r

### Short-term determinants of core inflation (VECM dynamics and final specification)
- Modeling approach:
  - Use long-run coefficients from model with price homogeneity and weak exogeneity.
  - Add a second long-term relation from mark-up model.
  - Focus on core inflation measure CPIX.
  - Start with VECMs with two lags (“initial” equation), reduce to significant variables and add exogenous variables (“final” equation).
- Table 8 — Dynamics of core inflation (Initial vs Final; only significant coefficients shown except mark-up long-run):
  - Loading coefficients to long run cointegration relations:
    - Money demand equation: -0.05*** (Initial), -0.04*** (Final)
    - Mark up equation: 0.04
  - Short run:
    - d Price (-1): 0.49*** (Initial), 0.42*** (Final)
    - d Price (-2): -0.30*** (Initial), -0.22** (Final)
    - d ULC: 0.17**
    - d CPI Brazil (-1): 0.06*** (Initial), 0.03* (Final)
    - d Adm prices: 0.10**
    - d Fruits and veg.: 0.02***
    - Constant: 0.55** (Initial), 0.50*** (Final)
  - Statistics:
    - R2: 0.65 (Initial), 0.74 (Final)
    - Log likelihood: 208.8 (Initial), 218.4 (Final)
    - Durbin-Watson: 2.02 (Initial), 1.86 (Final)
    - Akaike criteria: -6.25 (Initial), -6.51 (Final)
    - Schwartz criteria: -5.99 (Initial), -6.17 (Final)
  - Notes:
    - First autoregressive coefficient around 0.49; second autoregressive coefficient -0.30.
    - Contemporaneous change in ULC significant; ULC may anchor inflation inertia and raise input costs.
    - Exchange rate with Brazil matters more for imported-inflation dynamics than guaraní/U.S. dollar rate in short run.
    - Quarterly changes in prices in Brazil enter with a lag of one quarter.
    - Administered prices significant with estimated coefficient 0.10 vs share in CPIX of 8.3 percent.
    - Fruits and vegetables have modest effect on core inflation dynamics despite exclusion from core CPI.

### Model fit, forecasts, and dynamic contributions
- Model performance:
  - Extended model explains about 3/4 of the variance of inflation over the sample period.
  - Loading coefficient to cointegrating relation reduced by half in extended model: -0.04 (extended) vs -0.07 (restricted VECM); up to -0.13 in some unrestricted specifications.
  - Out-of-sample forecast for 2008:
    - Model captures uptick in core inflation in 2008 but underestimates actual by about one percentage point.
- Dynamic contributions (univariate decomposition):
  - Bulk of inflation dynamics driven by evolution of currency in circulation given model structure.
  - Currency in circulation has less influence on short-term spikes (e.g., rebounds in 1999, 2003, 2005 not directly related to currency in circulation).
  - Since 2004:
    - Currency in circulation contributed by 10 to 12 percent to inflation.
    - Buoyant money demand related to transaction motive (real income) added a negative impact on inflation increasing from 2 to 5 percent since 2004.
    - Appreciation of the guaraní contributed to reduce inflation by 2.5 percent on average.
  - Exchange rate contributions:
    - Depreciation of guaraní against U.S. dollar added about 3.2 percent to inflation in 2002-03 (via opportunity cost channel).
    - Depreciation against Brazilian real added about 0.7 percent of inflation in the same period (via imported goods channel).

*Italic: Source — staff estimates and analyses reported in the document _wp08270 (Annex tables VI.3 and VI.4 and accompanying sections).*

### 2002. The spike in inflation in 2005 can also be explained by exchange rates: the slowdown

### _wp08270 - 2002. The spike in inflation in 2005 can also be explained by exchange rates: the slowdown

### Dynamic contributions to core inflation (selected observations)
- The slowdown in the appreciation of the guaraní, despite continued currency growth, contributed to about 2 percent to inflation in 2005.
- The depreciation against the Brazilian currency added another 1 percent to inflation in 2005.
- Over the sample period, the trend decline of unit labor cost has contributed to a decline of inflation, although occasionally it locked-in some inflation increases (for example in 2003 or 2006).
- The moderation of administrated prices since 2003 has contributed to reduce core inflation by about 1 percentage point over the most recent period, but this is described as a temporary measure.
- In 2008, the uptick in core inflation is explained by:
  - continued increase of monetary supply in excess over money demand;
  - money supply continued to increase while the contribution of the other long term components of money demand, output and opportunity cost, is now broadly constant;
  - the contribution of cost-push factors, notably administrated prices, switched from negative to slightly positive.

### Key econometric and empirical findings (summary points from paper)
- Money talks.
  - Currency in circulation appears as the most relevant monetary aggregate for inflation dynamics.
  - An increase of currency in circulation by 1 percent leads to an increase of inflation by 0.05 points after one quarter, but by 1 percent in the long-run.
  - In the money demand function, the coefficient of income larger than unity suggests a trend increased monetization over the sample period.
  - In the long-run, the quantitative theory of money holds for broad money, but this result has little relevance for short-term inflation.
- Pass-throughs matters.
  - In the short run, a 1 percent increase of imported inflation from Brazil adds 0.03 percent to inflation in Paraguay.
  - In the long-run, the exchange rate with respect to the US dollar remains a key factor for portfolio allocation decisions and for money demand.
  - The significance of the depreciation with respect to the U.S. dollar could possibly also capture some pass-through effect from exchange rates to prices.
- Food feeds inflation.
  - Food accounts for about 35 percent of the CPI basket.
  - Food prices tend to have a spillover effect on non-food prices; Granger-causality tests and coefficients in a price equation indicate the final impact is larger than its weight.
- Inertia remains strong.
  - A shock of 1 percent to inflation in a given quarter gives rise to an additional increase by 0.4 percent after one quarter and by 0.2 percent after two quarters.
  - The wage indexation mechanism may be partly responsible for locking in any inflation increase; high inertia could be related to inflation expectations and the credibility of anti-inflationary efforts.

### Risks for the inflation outlook (2008-09)
- Supply shocks, in particular from food, and exchange rate shocks have affected inflation.
- Specific risks that could lead to higher inflation:
  - a reversal of the appreciation of the guaraní against the U.S. dollar;
  - further depreciation against the Brazilian real;
  - an end of the moderation of administrated prices.

### Monetary policy effectiveness and instruments
- Evidence of monetary policy effectiveness is presented through both volume and policy rates.
- Currency in circulation is described as:
  - the monetary aggregate most easily controlled by the Central bank (as opposed to broader monetary aggregates);
  - the best predictor of inflation.
- Policy instruments and mechanisms discussed:
  - Controlling the rate of currency growth through emission of sterilization paper allows moderation of the inflationary impact of money.
  - For broad money, the policy interest rate is significant in influencing the portfolio allocation of real money balances.
  - Central Bank sterilization paper (Letras de Regulacion Monetaria or LRM) rose from representing less than a tenth of currency in circulation in 1998 to close to ¼ of currency in circulation by end-2007.

### CPI measurement, new index and micro-level findings
- New CPI (base month Dec. 2007) vs older CPI (base month Dec. 1992):
  - Food: old index 35.3 → new index 32.0 (in IPC comparison table using new breakdown).
  - Fruits and vegetables: 6.2 → 5.1.
  - Meat: 10.7 → 10.4.
  - Clothing: 8.6 → 4.9.
  - Transport: 7.4 → 14.8.
  - Communication: 0.7 → 3.4.
  - Memorandum: Fuel 3.6 → 5.8; Administrated prices 7.7 → [7.0].
- All things being equal, the new index used from January 2008 should lead to a lower inflation measurement.
  - Had the weights of the new index been used in 2007, headline inflation could have been 1.5 percent lower on average.
  - The new basket integrates new products (cell phones, internet services, personal computers, cable television, etc.) which have experienced low or negative price increases; it has a larger weight for oil related products.
- Micro-level statistics (selected values from 1994-2007 analysis using 1992 basket):
  - Food: Number items 91; Weight in CPI 39.0; Share of tradables 89; Average monthly change 0.80; Standard deviation 2.02; Coeff. of seasonal variation Q1 1.02, Q2 1.02, Q3 1.00, Q4 1.00.
  - Fruits and vegetables: Number items 17; Weight 6.2; Average monthly change 1.35; Standard deviation 9.20; Coeff. of seasonal variation Q1 1.14, Q2 1.10, Q3 0.99, Q4 1.00.
  - Administrated prices: Number items 8; Weight 7.7; Average monthly change 0.90; Standard deviation 1.96; Coeff. of seasonal variation Q1 1.02, Q2 1.01, Q3 1.00, Q4 1.00.
  - Core inflation (CPIX): Items 276; Weight 93.8; Average monthly change 0.70; Standard deviation 0.67.
  - Core inflation (CPIX1): Items 262; Weight 82.5; Average monthly change 0.65; Standard deviation 0.67.
  - Tradable: Items 101; Share in tradables 66.7; Average monthly change 0.76; Standard deviation 1.40.
  - Imported: Items 65; Weight 21.9; Average monthly change 0.79; Standard deviation 3.15.
  - All items: Items 293; Weight 100.0; Average monthly change 0.72; Standard deviation 1.0.
- Granger-causality tests (selected result):
  - Unlike most Latin American countries, food inflation in Paraguay is a predictor of non-food inflation; the result is robust to the number of lags considered.

### Data and methodological notes (selected)
- Unit labor cost (ULC) definition: average wage index minus labor productivity for the economy as a whole.
  - Average wages available semi-annually; minimum wages monthly; employment annual with lag.
  - Extrapolations use linear relationships and Chow-Lin (1971) methodology to construct quarterly series.
- Price of foreign inputs proxies used: nominal effective exchange rate; exchange rate with the US$ (with or without adjustment for U.S. inflation); prices of imported consumer goods from Argentina and Brazil converted to guaranies.
- Inflation measures used in estimations: CPIX1 preferred; also headline inflation and core inflation CPIX.
- Monetary aggregates defined:
  - Currency in circulation: CC.
  - Narrow money M1 = CC + cash in vault and demand deposits.
  - Guarani Broad Money M2 = M1 + time and saving deposits + certificates of deposits.
  - Broad Money M3 = M2 + dollar deposits.
- Opportunity cost measurement examples:
  - Compare zero nominal return of holding domestic currency to return of holding dollars in cash (depreciation of exchange rate) and return on local currency time and saving deposits.
  - For Guarani broad money, alternative assets could be dollar deposits or local bonds; Treasury bills not widely traded post-2003 default.

*Source: IMF working paper content provided in the supplied PDF excerpt.*

### Annex Table V.1 : Database Used

### Annex Table V.1 : Database Used

### Activity and price
- GDP — lysa: Quarterly GDP from National Accounts. 1/
- CPI — lcpi: Consumer price index in Greater Asuncion.
- Core prices — lcpix: CPI exlcuding fruits and vegetables
- Core tradables — lcpixt: Tradable prices, excluding fruits and vegetables
- Core non tradables — lcpixnt: Non tradable prices

### Wages
- lwag: Average wage in Greater Asuncion
- lwmin: Minimum wages
- lulc: Unit labor cost — Average wage cost per unit of output, using labor force data from national accounts.

### Monetary aggregates 2/
- lcc: Currency in circulation
- lm1: Narrow money — Currency in circulation + cash in vault and demand deposits
- lm1a: M1A — Narrow money + time and savings deposits
- lm2: M2 — Narrow money + time and savings deposits, certificates of deposits
- lm3: M3 — M2 and foreign currency deposits

### Opportunity cost of money on: 3/
- r_ddg: Demand deposit in local currency
- r_tsg: Time and saving dep. in local currency
- r_d$: Dollar deposit — Weighted average of time and saving dollar deposits
- r_d$g: Dollar deposit in Gs. — Above plus y-o-y nominal Gs. depreciation
- r_lrm: Policy rate on LRM — Avg. interest rate on sterilzation paper (LRM)
- r_m1: M1 — Weighted average
- r_m1a: M1A — Weighted average
- r_m2: M2 — Weighted average
- r_m3: M3 — Weighted average, converted in Gs.
- r_$y: Dollar cash — Y-o-y depreciation of the Gs
- r_$q: Dollar cash — Q-o-q depreciation of the Gs
- r_$fq: Dollar cash — Forward looking q-o-q depreciation of the Gs

### Exchange rate and foreign prices
- lneer: Nominal effective exchange rate — INS based NEER with correction from hyperinflation in Brazil before 1994.
- ler: Nominal exch. rate — Exchange rate with U.S. dollar
- lcpi_usa: U.S. CPI — CPI price in USA, converted in Gs
- lcpi_bra: Brazil CPI — CPI price in Brazil, converted in Gs
- lcpi_arg: Argentina CPI — CPI price in Argentina converted in Gs

### Sources and notes
- Sources: Banco Central de Paraguay (BCP); International Finance Statistics (IFS, IMF); and authors'
- 1/ Extrapolated using monthly indicator of activity for 1991-94 and 2007.
- 2/ BCP after January 1995, backdated using IFS.
- 3/ Expressed in percentage: 100 bp is 0.01.

*Source: _wp08270 - Annex Table V.1 : Database Used*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2008/_wp08270.pdf_