## _wp09147

## Source details

**Canonical URL:** [_wp09147](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09147.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09147.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09147.pdf.json)

---

### I. Introduction
- Purpose: analyze high-frequency price and volatility dynamics in emerging market external sovereign bonds and their reaction to local, regional and international macroeconomic news.
- Motivation:
  - Emerging sovereign bonds increasingly affected by "cross-over investors" who hold both emerging and mature sovereign bonds.
  - Compare emerging sovereign bonds to U.S. treasury bonds as a benchmark due to similar information sets and market structures (OTC interdealer trading).
- Expectations / hypotheses:
  - Local macroeconomic releases expected to have a direct and consistent effect on prices and volatility.
  - News from systemically and regionally important economies should strongly affect foreign currency–denominated emerging sovereign bonds.
  - Information absorption may take longer in external emerging markets because of lower liquidity and greater information asymmetries.
  - Distinguish repricing (price impact) from repositioning (volatility impact); surprises (Actual − Expectation) drive repricing.
  - Volatility responses anticipated to dominate price responses and remain elevated longer in emerging markets.
- Principal empirical findings (summary):
  - Initial price adjustment upon arrival of new information is weak and dissipates within minutes.
  - Volatility response is much more pronounced than price response: volatility remains at elevated levels, at up to six times the preannouncement level, for up to one and a half hours after the announcement — about three times as long as in mature bond markets.
  - International (U.S.) news is generally at least as important as domestic news for valuations and volatility in emerging markets.
  - Evidence of asymmetric effects: stronger responses to negative news than to positive news.
  - Disproportionately large impact from news releases containing large surprises.
  - During the subprime crisis, average volatility increased and the impact of international (U.S.) macroeconomic news became more muted.

### III. Intraday Price Data and Announcements (sample and liquidity)
- Sample:
  - Countries: Brazil, Mexico, Russia, Turkey (Argentina and Venezuela excluded for data problems).
  - Sample period: October 1, 2006 ñ February 20, 2008.
  - On-the-run benchmark: 10-year U.S. treasury note (tick-by-tick from Tullett Prebon).
- Bond-specific liquidity and outstanding volume:
  - Brazil: 11 percent 2040 bond, outstanding volume $4.2 billion, average bid-ask spread $0.12 per $100 face value.
  - Mexico: 5.625 percent 2017, outstanding about $3.5 billion, average bid-ask spread $0.22.
  - Russia: 2030 bond, outstanding $20 billion, weight close to 8 percent in EMBIG, average bid-ask spread about $0.23.
  - Turkey: 11.875 percent 2030, outstanding $1.5 billion, annual trading volume about $80 billion, average bid-ask spread $0.42.
  - Emerging market sovereign eurobonds accounted for 18 percent of emerging market debt trading (or $2.3 billion daily) in the last quarter of 2008.
  - Emerging market debt trading was about $13 billion a day at end-2008 (down from $19 billion in the same quarter of 2007).
  - This emerging market trading is roughly a quarter of trading in U.S. treasuries (or $10 billion daily average at end-2008).
- Data construction and cleaning:
  - Primary frequency: 10-minute mid-quotes (Bloomberg); secondary dataset: 1-minute returns between 8:00 a.m. and 9:00 a.m. EST for U.S. macro news impact analysis.
  - Observations removed: outside trading hours (3.00 a.m. ñ 5.00 p.m. EST for Mexican and Brazilian bonds; 2.00 a.m. ñ 5.00 p.m. EST for Turkish and Russian bonds), weekends, major U.S. and U.K. public holidays, days with ≥95 percent zero 10-minute returns.
  - Additional exclusions: non-positive bids/asks, absolute bid/ask price change > 10% of previous price, and technical jumps in U.S. treasury data (~3 percent of sample).
  - Overnight returns replaced with unconditional means.
- Summary statistics (preserve values exactly):
  - Number of trading days: Brazil 325 (10m); Mexico 297; Russia 324; Turkey 325; U.S. 340.
  - Proportion of 8:00–9:00 quotes: Brazil 9.92%, Mexico 10.77%, Russia 11.06%, Turkey 10.16%, U.S. 8.27%.
  - Number of observations (examples): Brazil 27,300 (1min) and 19,500 (10min); U.S. 30,600 (1min) and 20,340 (10min).
  - Liquidity (average number of bid/ask quotes per trading day, period November 6, 2006 ñ February 20, 2008):
    - Brazil: 166 (10m) and 18 (1min)
    - Mexico: 210 and 25
    - Russia: 236 and 29
    - Turkey: 230 and 26
    - U.S.: 659 and 55
  - Liquidity before vs during the subprime crisis:
    - Brazil before crisis: 60 and 9; during crisis: 246 and 25.
    - Mexico before crisis: 129 and 21; during crisis: 276 and 29.
    - Russia before crisis: 188 and 27; during crisis: 272 and 31.
    - Turkey before crisis: 170 and 18; during crisis: 276 and 32.
    - U.S. before crisis: 561 and 52; during crisis: 751 and 58.
    - Note: For all assets, differences in liquidity before and during the subprime crisis are significant at the 5% level.
  - Returns (examples):
    - Mean returns: 0.000 for all 1- and 10-minute series.
    - Standard deviation examples: Brazil 1-minute 0.019; Brazil 10-minute 0.011; Mexico 10-minute 0.013; U.S. 1-minute 0.003.
    - Kurtosis examples: Brazil kurtosis 19.662 for 1min, 37.326 for 10min; U.S. kurtosis 14.501 for 1min, 53.033 for 10min.
    - First-order autocorrelation examples: Brazil 1min −0.171; U.S. 10min 0.055.
  - Absolute returns (examples):
    - Mean absolute returns: Brazil 1min 0.009; U.S. 1min 0.004.
    - U.S. treasury note returns are the least volatile with average absolute return equal to 0.2 basis points (noted qualitatively).
- Announcements and surprise measures:
  - Announcement data and expectations from Bloomberg; surprise standardized as S_k,t = (Actual_k,t − Expectation_k,t)/bσ_k.
  - Release-time indicator A_k,t used when forecasts unavailable.
  - Announcement counts (examples):
    - CPI: Brazil 336 Obs., 288 Exp.; Mexico 23 Obs., 23 Exp.; Russia 64 Obs., 55 Exp.; Turkey 15 Obs., 15 Exp.; U.S. 32 Obs., 32 Exp.; Germany 32 Obs., 32 Exp.
    - Interest rate announcements: Brazil 9 Obs., 9 Exp.; Mexico 13 Obs., 13 Exp.; Russia 15 Obs., 15 Exp.; Turkey 15 Obs., 15 Exp.; U.S. 8 Obs., 7 Exp.; Germany 17 Obs., 17 Exp.
    - Total number of releases varies by country: Turkey 152 releases (implied), Brazil 529 releases (range stated).

### IV. Two-stage modeling of returns and volatility (estimation framework)
- Conditional mean (10-minute specification, equation (2)):
  - R_t = ψ_0 + Σ_{i=1}^I ψ_i R_{t−i} + Σ_{k=1}^K Σ_{j=0}^J ψ_kj S_{k,t−j} + ε_t, where ε_t ~ (0, σ_t^2).
  - Lag length I and response length J determined by model selection criteria.
- Volatility (multiplicative specification, equations (3) and (4)):
  - σ_t = h(deterministic volatility) × g(stochastic volatility) × u_t, u_t i.i.d. mean 1, variance 1.
  - ln|ε_t| = β + ζ(t) + σ_d(t) + Σ_{i=1}^{I0} ϕ_i ln|ε_{t−i}| + Σ_{k=1}^K Σ_{j=0}^{J0} ϕ_kj A_{k,t−j} + ln u_t.
  - Deterministic intraday seasonality ζ(t) modeled with cubic splines (hourly knots + extra knot at 8:30 a.m. to capture U.S. opening); separate splines for Mondays and Fridays; allow for structural break at onset of subprime crisis.
  - Long-run daily volatility σ_d(t) estimated via average of 10-minute absolute return innovations over previous day.
  - Short-run persistence captured by lags of ln|ε_t|.
- Inference and robustness:
  - Two-step WLS: use reciprocal of fitted volatility series as WLS weights for conditional mean.
  - Volatility innovations u_t tested to be i.i.d.; HAC (Newey-West) standard errors used with truncation L = trun[ (1/2) (T×N)^{1/3} ].
- Regime detection (Markov-switching VAR):
  - MS(2)-VAR(4,2) used on January 2006–February 2008 (544 observations).
  - Joint transition probability matrix estimated: P = [0.8104 0.7041; 0.1896 0.2959].
  - June 5, 2007 identified as statistically significant date of structural break (start of U.S. subprime crisis period).

### V. Price dynamics and intraday patterns in emerging markets
- Deterministic seasonality and model choice:
  - Deterministic intraday seasonality important; multiplicative volatility model fits emerging markets better than additive models used for mature markets.
  - Cubic splines with hourly knots (extra knot at 8:30 a.m.) capture intraday volatility peaks at U.S. opening.
  - Mondays and Fridays require separate splines (largest volatility on Fridays, least on Mondays).
- Intraday volatility patterns:
  - Inverse U-shaped intraday pattern; spikes at U.S. and U.K. openings, especially for euro-denominated Russian and Turkish bonds.
  - Higher moments (skewness, kurtosis) show time-of-day seasonality; skewness/kurtosis vary by bond and time-of-day (examples documented).
- Crisis effects on intraday patterns:
  - Intraday patterns differ before vs during crisis: pre-crisis U.S. opening associated with increased volatility; during crisis, differences between U.K. and U.S. openings diminish.
  - Average daily volatility increased across all emerging markets during the financial crisis.
  - Crisis interaction terms included to control for changing deterministic volatility patterns.
- Lag-structure selection and final AR-ARCH specifications (summary):
  - Response lengths chosen: J = 0 for returns; J0 = 3 (30 minutes) for volatility.
  - AR-ARCH final specifications (Table 3 summary; structure preserved):
    - Brazil: AR(9); ARCH(13).
    - Mexico: AR(16); ARCH(18).
    - Russia: AR(7); ARCH(8).
    - Turkey: AR(8); ARCH(18).
    - U.S.: AR(1); ARCH(9).
  - Average correlation between observed and fitted volatility series: 0.35.
  - Volatility innovations u_t approximately i.i.d.; HAC standard errors used.

### VI. High-frequency (1-minute) U.S. news impact on returns (model and main findings)
- 1-minute news-response model (between 8:00 and 9:00 a.m. EST; equation (5)):
  - Rt = φ0 + φ1 Rt-1 + Σ_{k=1..K} Σ_{j=0..3} φkj Sk,t-j + εt.
  - Rt: 1-minute log-return; Sk,t: standardized U.S. macro surprise at 8:30 a.m. EST.
  - ContE§ denotes contemporaneous impact (φk0). TotE§ denotes total impact over 3 minutes, Σ_{j=0..3} φkj tested with χ2 Wald statistic.
  - Impulse-response estimates via Monte Carlo; sample period October 1, 2006 ñ February 20, 2008.
- Price-level (returns) findings:
  - Price responses to positive surprises in announcements of real activity indicators are negative (price declines), consistent with a period near the peak of the economic cycle and rising price pressures.
  - "The magnitude of contemporaneous effects on returns varies between 8 and 108 basis points."
  - "The most significant impact of news is observed within three minutes of the announcement"; total effect reported as percentage change in return during four minutes (Σ_{j=0..3} φkj).
- Illustrative contemporaneous (ContE§) and total (TotE§) 1-minute return effects (percentage-point changes) from Table 4 (examples preserved exactly):
  - GDP (Advance): Brazil ContE§ −0.160 TotE§ −1.517; Mexico ContE§ −0.040 TotE§ −3.367; Russia ContE§ −0.051 TotE§ −3.674; Turkey ContE§ 0.153 TotE§ −2.511; U.S. ContE§ −0.086 TotE§ −1.453.
  - Core CPI (U.S. surprise impact): Brazil ContE§ −0.735 TotE§ −3.802; Mexico ContE§ 0.060 TotE§ −4.893; Russia ContE§ 0.012 TotE§ −3.735; Turkey ContE§ 0.116 TotE§ −3.458; U.S. ContE§ −0.185 TotE§ −2.255.
  - Additional items (Personal consumption (Adv), GDP (Pre), Current account, Trade balance, Durable goods orders, Retail sales, Personal spending, Personal income, Housing starts, Building permits, CPI, PPI, Unemployment) reported in Table 4 with contemporaneous and total values by country.

### VI. Volatility impacts, heterogeneity, and nonlinearities
- Volatility-impact general patterns:
  - Macroeconomic news has a much larger effect on volatility than on price levels.
  - Releases of domestic and U.S. macroeconomic data increase volatility by one and a half times on average, with responses lasting for up to one and a half hours (and even longer for some news types).
  - Prolonged volatility reflects a longer portfolio reallocation process owing to lower liquidity.
- Cross-country heterogeneity and news origin:
  - U.S. news reaction: large and mostly homogeneous across emerging markets, with a smaller volatility reaction in Brazilian bonds.
  - Domestic news: more muted and differentiated responses — Brazil reacts instantaneously and significantly to many domestic macro announcements; Russia shows the weakest volatility response to domestic news; Turkey exhibits a protracted pickup in volatility.
  - Mexico and Russia exhibit volatility reversals for some announcements.
- Cross-sectional quantitative examples from Table 5 (ContE§ / TotE§ denote percentage change in volatility; observation window = 40 minutes — Σ_{j=0..3} φkj0):
  - All Domestic News:
    - Brazil: 46.11 / 51.69
    - Mexico: 38.281 / 36.44
    - Russia: 11.89 / −10.29
    - Turkey: −8.20 / 56.70
  - CPI (domestic news):
    - Brazil: 165.19 / 236.41
    - Mexico: 124.19 / 185.53
    - Russia: 163.97 / 614.42
    - Turkey: 61.09 / 113.86
  - All U.S. News:
    - Brazil: 25.54 / 49.88
    - Mexico: 33.70 / 92.89
    - Russia: 27.59 / 62.21
    - Turkey: 15.89 / 77.87
    - U.S.: 33.123 / 68.91
  - Interest Rate (U.S. news):
    - Brazil: 119.71 / 361.92
    - Mexico: 279.94 / 645
    - Russia: −63.69 / −238.14
    - Turkey: −5.77 / −41.10
    - U.S.: 218.51 / 506.45
- Interpretation notes:
  - Coefficients in bold in tables indicate significance at the 5% level (HAC standard errors).
  - Russia: limited meaningful monetary policy interest rate series; irregular timing and preannouncements may mute responses.

### VII. Asymmetries, surprise size, and regime dependence
- Asymmetry by sign:
  - Negative local real activity news produces more volatility over the total observation window than positive news.
  - Negative U.S. real economic shocks elicit stronger responses than positive shocks, particularly for Mexican, Turkish, and U.S. bonds.
  - Inflation news responses more balanced; positive U.S. inflation surprises can cause greater volatility in emerging market bonds.
- Size-of-surprise nonlinearities (70th percentile cutoff for large surprises):
  - Larger surprises in U.S. data trigger more sizable and immediate volatility reactions than large domestic surprises.
  - Examples (Table 7, ContE§ / TotE§ for lower and upper 0.70 quantiles):
    - Brazil, Domestic news: Lower quantile 89.86 / 67.32; Upper quantile 38.93 / 78.57.
    - U.S. news, Upper quantile: Brazil 32.35 / 51.08; Mexico 64.96 / 121.47; Russia 53.45 / 131.86; Turkey 35.49 / 143.52; U.S. 52.23 / 92.21.
- Regime dependence (before vs during subprime crisis, structural break June 5, 2007):
  - With onset of financial turbulence, response to U.S. macro releases became less pronounced; domestic news in Brazil and Turkey gained importance.
  - Example (Table 8, ContE§ / TotE§):
    - All Domestic News (Brazil): Before crisis −7.50 / 27.12; During crisis 45.12 / 56.05.
    - All U.S. News (Brazil): Before crisis 47.18 / 90.68; During crisis 5.90 / 7.09.
  - Interpretation: investors shifted attention away from U.S. (and German) news during early U.S. financial crisis; intraday volatility rose while aggregate effect of U.S. surprises on emerging bond volatility became less consistent and weaker.

### VIII. Mechanisms, market-structure considerations, and interpretation
- Timing and information content:
  - Timely, frequent indicators elicit stronger responses (e.g., weekly trade balance in Brazil vs monthly).
  - GDP, though released late, carries large marginal information content for emerging markets.
- Reliability and preannouncement effects:
  - Perceived data reliability and forecast precision affect market reaction; noisy/high-revision series dampen immediate reaction.
  - Preliminary/advance figures and policy guidance reduce surprise magnitude and market reaction.
- Drivers of prolonged volatility in emerging markets:
  - Greater information asymmetries, lower liquidity, larger share of international investors, and OTC market features lead to longer and more expensive portfolio reallocations and prolonged elevated volatility.

### IX. Key conclusions (Section VII summary)
- Immediate price re-pricing is nearly instantaneous (absorbed within a five-minute period), similar to mature bond markets.
- The repricing process is accompanied by a prolonged period of elevated volatility and trading activity in emerging bond markets — volatility remains elevated for more than one and a half hours (about twice as long as in mature markets).
- International and regional news (notably U.S. inflation and monetary policy announcements) is at least as important as local news for emerging market external bonds.
- Strong asymmetric effects are present: bad news often matters more than good news; large surprises have disproportionately large impacts.
- During the early subprime crisis period, attention shifted away from U.S. (and German) macro news toward more timely/specific domestic indicators in some emerging markets, though no general change in response patterns was identified.

*Source: _wp09147 - References*

### References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  28

### References . . . . . . . . . . . . . . . . . . . . . . . . . . . 28

### I. Introduction
- Purpose: analyze high-frequency price and volatility dynamics in emerging market external sovereign bonds and their reaction to local, regional and international macroeconomic news.
- Motivation:
  - Emerging sovereign bonds increasingly affected by "cross-over investors" who hold both emerging and mature sovereign bonds.
  - Compare emerging sovereign bonds to U.S. treasury bonds as a benchmark due to similar information sets and market structures (OTC interdealer trading).
- Key expectations and hypotheses:
  - Local macroeconomic releases expected to have a direct and consistent effect on prices and volatility.
  - News from systemically and regionally important economies should strongly affect foreign currency–denominated emerging sovereign bonds.
  - Information absorption may take longer in external emerging markets because of lower liquidity and greater information asymmetries.
  - Distinguish repricing (price impact) from repositioning (volatility impact); surprises (Actual − Expectation) drive repricing.
  - Volatility responses anticipated to dominate price responses and remain elevated longer in emerging markets.
- Principal empirical findings summarized:
  - Initial price adjustment upon arrival of new information is weak and dissipates within minutes.
  - Volatility response is much more pronounced than price response: volatility remains at elevated levels, at up to six times the preannouncement level, for up to one and a half hours after the announcement — about three times as long as in mature bond markets.
  - International (U.S.) news is generally at least as important as domestic news for valuations and volatility in emerging markets.
  - Evidence of asymmetric effects: stronger responses to negative news than to positive news.
  - Disproportionately large impact from news releases containing large surprises.
  - During the subprime crisis, average volatility increased and the impact of international (U.S.) macroeconomic news became more muted.

### II. Literature Review
- Mature-market stylized facts:
  - Prices may adjust within one minute of announcements; trading activity (volatility) rises within 10 to 15 minutes and remains elevated for about an hour.
  - Initial stages of adjustment dominated by informed trading; later stages by liquidity trading (Fleming and Remolona, 1999; Balduzzi, Elton, and Green, 2001).
- High-frequency studies cited:
  - Andersen et al. (2003): surprises in macroeconomic releases significantly affect five-minute returns in U.S. dollar spot exchange rates; responses are asymmetric (bad news > good news).
  - Mixed findings on post-announcement volatility persistence across studies (Bauwens, Omrane, and Giot, 2005; Ederington and Lee, 1993; Fleming and Remolona, 1999).
- Insights for indicator selection:
  - Timeliness and incremental information content matter (e.g., GDP less important than CPI, employment, industrial production, factory orders) (Veredas, 2006).

### III. Intraday Price Data and Announcements
- Sample and assets:
  - Countries: Brazil, Mexico, Russia, Turkey (Argentina and Venezuela excluded for data problems).
  - Sample period: October 1, 2006 ñ February 20, 2008.
  - On-the-run benchmark for mature comparison: 10-year U.S. treasury note (tick-by-tick from Tullett Prebon).
- Bond-specific details and liquidity highlights:
  - Brazil: 11 percent 2040 bond, outstanding volume $4.2 billion, average bid-ask spread $0.12 per $100 face value.
  - Mexico: 5.625 percent 2017, outstanding about $3.5 billion, average bid-ask spread $0.22.
  - Russia: 2030 bond, outstanding $20 billion, weight close to 8 percent in EMBIG, average bid-ask spread about $0.23.
  - Turkey: 11.875 percent 2030, outstanding $1.5 billion, annual trading volume about $80 billion, average bid-ask spread $0.42.
  - Emerging market sovereign eurobonds accounted for 18 percent of emerging market debt trading (or $2.3 billion daily) in the last quarter of 2008.
  - Emerging market debt trading was about $13 billion a day at end-2008 (down from $19 billion in the same quarter of 2007).
  - This emerging market trading is roughly a quarter of trading in U.S. treasuries (or $10 billion daily average at end-2008).
- Data construction and cleaning:
  - Primary frequency: 10-minute mid-quotes (Bloomberg); secondary dataset: 1-minute returns between 8:00 a.m. and 9:00 a.m. EST for U.S. macro news impact analysis.
  - Observations removed: outside trading hours (3.00 a.m. ñ 5.00 p.m. EST for Mexican and Brazilian bonds; 2.00 a.m. ñ 5.00 p.m. EST for Turkish and Russian bonds), weekends, major U.S. and U.K. public holidays, days with ≥95 percent zero 10-minute returns.
  - Additional cleaning: exclude non-positive bids/asks, absolute bid/ask price change > 10% of previous price, and technical jumps in U.S. treasury data (~3 percent of sample).
  - Overnight returns replaced with unconditional means.
- Summary statistics (Table 1 highlights; preserve values exactly):
  - Number of trading days (varies by bond): Brazil 325 (10m), 325 (1min sample days vary); Mexico 297; Russia 324; Turkey 325; U.S. 340.
  - Proportion of 8:00–9:00 quotes: Brazil 9.92%, Mexico 10.77%, Russia 11.06%, Turkey 10.16%, U.S. 8.27%.
  - Number of observations (example values): Brazil 27,300 (1min) and 19,500 (10min); U.S. 30,600 (1min) and 20,340 (10min).
  - Liquidity (average number of bid/ask quotes per trading day, period November 6, 2006 ñ February 20, 2008):
    - Brazil: 166 (10m) and 18 (1min)
    - Mexico: 210 and 25
    - Russia: 236 and 29
    - Turkey: 230 and 26
    - U.S.: 659 and 55
  - Liquidity before vs during the subprime crisis:
    - Brazil before crisis: 60 and 9; during crisis: 246 and 25.
    - Mexico before crisis: 129 and 21; during crisis: 276 and 29.
    - Russia before crisis: 188 and 27; during crisis: 272 and 31.
    - Turkey before crisis: 170 and 18; during crisis: 276 and 32.
    - U.S. before crisis: 561 and 52; during crisis: 751 and 58.
    - Note: For all assets, differences in liquidity before and during the subprime crisis are significant at the 5% level.
  - Returns (examples):
    - Mean returns: 0.000 for all 1- and 10-minute series.
    - Standard deviation (examples): Brazil 1-minute 0.019; Brazil 10-minute 0.011; Mexico 10-minute 0.013; U.S. 1-minute 0.003.
    - Skewness and kurtosis indicate negatively skewed distributions (except some series) and excess kurtosis across series (e.g., Brazil kurtosis 19.662 for 1min, 37.326 for 10min; U.S. kurtosis 14.501 for 1min, 53.033 for 10min).
    - First-order autocorrelation generally small and negative for returns (examples: Brazil 1min −0.171; U.S. 10min 0.055).
  - Absolute returns:
    - Mean absolute returns (examples): Brazil 1min 0.009; U.S. 1min 0.004.
    - U.S. treasury note returns are the least volatile with average absolute return equal to 0.2 basis points (noted qualitatively).
- Macroeconomic announcements and expectations:
  - Data on announcements and market expectations obtained from Bloomberg; selection guided by timeliness, economy-wide relevance, frequency (≥4 announcements in sample), and availability of analyst forecasts.
  - Measures of surprise: standardized surprise S_k,t = (Actual_k,t − Expectation_k,t)/bσ_k where bσ_k is standard deviation of surprises over the sample.
  - Use of release time indicators A_k,t as dummy measures of information arrival (motivated by Andersen et al. findings and to handle releases lacking forecasts).
- Announcement counts (Table 2 highlights; preserve values exactly for illustrative items):
  - CPI announcements (counts of Obs. and Exp. where provided): Brazil 336 Obs., 288 Exp.; Mexico 23 Obs., 23 Exp.; Russia 64 Obs., 55 Exp.; Turkey 15 Obs., 15 Exp.; U.S. 32 Obs., 32 Exp.; Germany 32 Obs., 32 Exp.
  - Interest rate announcements counts: Brazil 9 Obs., 9 Exp.; Mexico 13 Obs., 13 Exp.; Russia 15 Obs., 15 Exp.; Turkey 15 Obs., 15 Exp.; U.S. 8 Obs., 7 Exp.; Germany 17 Obs., 17 Exp.
  - Total number of releases varies by country: Turkey 152 releases (implied in text), Brazil 529 releases (range stated).

### IV. Two-Stage Modeling of Returns and Volatility
- Estimation approach: two-step weighted least-squares (WLS) following Andersen and Bollerslev (1998) and Andersen et al. (2003, 2007).
- Conditional mean model (equation (2)):
  - R_t = ψ_0 + Σ_{i=1}^I ψ_i R_{t−i} + Σ_{k=1}^K Σ_{j=0}^J ψ_kj S_{k,t−j} + ε_t, where ε_t ~ (0, σ_t^2).
  - Lag length I and response length J determined by model selection criteria.
- Volatility model (multiplicative specification, equations (3) and (4)):
  - ε_t volatility: σ_t = h(deterministic volatility) × g(stochastic volatility) × u_t, u_t i.i.d. mean 1, variance 1.
  - Log absolute residual model: ln|ε_t| = β + ζ(t) + σ_d(t) + Σ_{i=1}^{I0} ϕ_i ln|ε_{t−i}| + Σ_{k=1}^K Σ_{j=0}^{J0} ϕ_kj A_{k,t−j} + ln u_t.
  - Deterministic intraday seasonality ζ(t) modeled with cubic splines (hourly knots + extra knot at 8:30 a.m. to capture U.S. opening); separate splines for Mondays and Fridays; allow for structural break at onset of subprime crisis.
  - Long-run daily volatility σ_d(t) estimated via average of 10-minute absolute return innovations over previous day (chosen over GARCH(2,2) forecast due to better AIC/BIC fit).
  - Short-run persistence captured by lags of ln|ε_t|.
- Inference and robustness:
  - Use reciprocal of fitted volatility series as WLS weights for conditional mean to obtain correct standard errors.
  - Volatility innovations u_t tested to be i.i.d.; HAC (Newey-West) standard errors used with truncation L = trun[ (1/2) (T×N)^{1/3} ] for robustness.
- Regime detection:
  - Markov-switching VAR MS(2)-VAR(4,2) used to endogenously identify structural break in returns and volatility; sample January 2006–February 2008 (544 observations).
  - Joint transition probability matrix estimated: P = [0.8104 0.7041; 0.1896 0.2959].
  - June 5, 2007 identified as statistically significant date of structural break (start of U.S. subprime crisis period).

### V. Price Dynamics in Emerging Markets
- Model modifications for emerging markets:
  - Deterministic intraday seasonality is important; multiplicative volatility model fits emerging markets better than additive models used for mature markets.
  - Cubic splines with hourly knots (extra knot at 8:30 a.m.) capture intraday volatility peaks at U.S. opening; Mondays and Fridays require separate splines (largest volatility on Fridays, least on Mondays).
  - Intraday volatility displays inverse U-shaped pattern; spikes at U.S. and U.K. openings, especially for euro-denominated Russian and Turkish bonds.
  - Higher moments (skewness, kurtosis) show time-of-day seasonality; close-time skewness/kurtosis vary by bond (examples provided).
- Crisis effects:
  - Intraday patterns differ before vs during crisis: pre-crisis U.S. opening associated with increased volatility; during crisis, differences between U.K. and U.S. openings diminish.
  - Average daily volatility increased across all emerging markets during the financial crisis.
  - Crisis interaction terms included to control for changing deterministic volatility patterns.
- Lag structure selection:
  - Conditional mean AR lags and volatility ARCH lags tested up to six hours; model selection via AIC/BIC (F-test to resolve conflicts).
  - Chosen response lengths: J = 0 for returns (no lagged news response), J0 = 3 (30 minutes) for volatility equations.
  - No allowance for individual news coefficients to change during financial turbulence in preferred specification.
- AR-ARCH final specifications (Table 3 summary; preserve structure exactly):
  - Brazil: AR(9); ARCH(13) (coefficients allowed to change during crisis indicated by marker in original table).
  - Mexico: AR(16) (with crisis coefficient changes); ARCH(18) (with crisis coefficient changes).
  - Russia: AR(7) (with crisis coefficient changes); ARCH(8) (with crisis coefficient changes).
  - Turkey: AR(8) (with crisis coefficient changes); ARCH(18) (with crisis coefficient changes).
  - U.S.: AR(1) (with crisis coefficient changes); ARCH(9) (with crisis coefficient changes).
- Model adequacy:
  - Average correlation between observed and fitted volatility series: 0.35.
  - Volatility innovations u_t are approximately i.i.d.; remaining autocorrelation in residuals addressed with HAC standard errors.

### VI. News Effects in Emerging Bond Markets
- Repricing (price impact):
  - Little systematic evidence in 10-minute return data that macroeconomic surprises produce distinctive price shifts.
  - Few indicators significantly explain the conditional mean; effects are not consistent across countries.
  - Possible reasons:
    - Ambiguous economic interpretation of some announcements for sovereign bonds (e.g., trade balance surprises).
    - Larger margin of error in emerging market statistics and weaker forecast accuracy.
    - Lower prominence of short-term arbitrage trading in emerging bond markets.
    - Longer, more protracted responses by long-term investors and cross-over investors.
  - 1-minute returns are used for a focused event study on U.S. releases between 8:00 and 9:00 a.m. EST to capture very short-lived price reactions that 10-minute intervals may miss.
- Repositioning (volatility impact):
  - Volatility response is more pronounced than price response in emerging markets.
  - Volatility can remain elevated up to six times the preannouncement level for up to one and a half hours after announcement — about three times longer than in mature markets.
  - International (U.S.) news generally as important as domestic news for volatility in emerging markets.
  - Asymmetries documented: stronger volatility responses to negative news than to positive news.
  - Large surprises (upper/lower quantiles) produce disproportionately large volatility impacts.
  - During the subprime crisis, average volatility increased and the impact of U.S. macroeconomic news became more muted, possibly due to other news importance (e.g., financial sector performance) or perceptions of emerging economies' resilience.
- Nonlinearities and cross-border spillovers:
  - Evidence of heterogeneous responses across countries and indicators.
  - International and regional news cause significant spillovers into emerging sovereign bond prices and volatility.

*Source: _wp09147 - References . . . . . . . . . . . . . . . . . . . . . 28*

### 9.00 a.m. EST. Table 4 reports the contemporaneous and total impact of surprises in U.S. news, obtained

### _wp09147 - 9.00 a.m. EST. Table 4 reports the contemporaneous and total impact of surprises in U.S. news, obtained

### Model and estimation approach
- Estimated news-response model for 1-minute returns:
  - Rt = φ0 + φ1 Rt-1 + sum(k=1..K) sum(j=0..3) φkj Sk,t-j + εt (equation (5) in source).
  - Rt is the 1-minute log-return on quotes posted between 8.00 and 9.00 a.m. EST.
  - Sk,t is the standardized news corresponding to a U.S. macroeconomic announcement k made at 8.30 a.m. EST.
  - ContE§ denotes contemporaneous impact (φk0). TotE§ denotes total impact over 3 minutes, calculated as Σj=0..3 φkj and tested with a χ2 Wald statistic.
- Impulse-response estimates obtained using Monte Carlo simulations and account for parameter estimation uncertainty.
- Sample period: October 1, 2006 ñ February 20, 2008.
- Data sources: Bloomberg, Tullett Prebon.

### Price-level (returns) findings (Table 4 and Figure 3)
- Main qualitative finding:
  - Price responses to positive surprises in announcements of real activity indicators are negative (price declines), consistent with a period near the peak of the economic cycle and rising price pressures.
- Magnitude and timing:
  - "The magnitude of contemporaneous effects on returns varies between 8 and 108 basis points," consistent with prior empirical work (Almeida, Goodhart, and Payne (1998)).
  - "The most significant impact of news is observed within three minutes of the announcement"; total effect reported as percentage change in return during four minutes (Σj=0..3 φkj in model specification).
- Illustrative country-specific contemporaneous and total effects on 1-minute returns (from Table 4; values are percentage-point changes):
  - GDP (Advance): Brazil ContE§ -0.160 TotE§ -1.517; Mexico ContE§ -0.040 TotE§ -3.367; Russia ContE§ -0.051 TotE§ -3.674; Turkey ContE§ 0.153 TotE§ -2.511; U.S. ContE§ -0.086 TotE§ -1.453.
  - Core CPI (U.S. surprise impact on 1-minute returns): Brazil ContE§ -0.735 TotE§ -3.802; Mexico ContE§ 0.060 TotE§ -4.893; Russia ContE§ 0.012 TotE§ -3.735; Turkey ContE§ 0.116 TotE§ -3.458; U.S. ContE§ -0.185 TotE§ -2.255.
  - Other items reported in Table 4 include Personal consumption (Adv), GDP (Pre), Personal consumption (Pre), Current account, Trade balance, Durable goods orders, Retail sales, Personal spending, Personal income, Housing starts, Building permits, CPI, PPI, Unemployment — see table for exact contemporaneous (ContE§) and total (TotE§) values by country.

### Volatility-impact findings (Figures 4–5, Table 5)
- Overall patterns:
  - The effect of macroeconomic news on volatility is "much more significant than its impact on price levels."
  - Releases of domestic and U.S. macroeconomic data increase volatility by one and a half times on average, with responses lasting for up to one and a half hours (and even longer for some types of news).
  - Prolonged volatility in emerging bond markets reflects a longer portfolio reallocation process owing to lower liquidity, not necessarily market inefficiency.
- Heterogeneity across countries and news origin:
  - U.S. news reaction is large and mostly homogeneous across emerging markets, with a smaller volatility reaction in Brazilian bonds.
  - Domestic news triggers more muted and differentiated responses: Brazilian bonds react instantaneously and significantly to a gamut of domestic macro announcements; Russia shows the weakest volatility response to domestic news; Turkish bonds show a protracted pickup in volatility.
  - Mexico and Russia exhibit volatility reversals for some announcements (contrary to Admati and Päeiderer (1988) model predictions).
- Cross-sectional quantitative examples from Table 5 (ContE§ and TotE§ denote percentage change in volatility; observation window = 40 minutes — Σj=0..3 φkj0):
  - All Domestic News (ContE§ / TotE§):
    - Brazil: 46.11 / 51.69
    - Mexico: 38.281 / 36.44
    - Russia: 11.89 / -10.29
    - Turkey: -8.20 / 56.70
  - CPI (domestic news) (ContE§ / TotE§):
    - Brazil: 165.19 / 236.41
    - Mexico: 124.19 / 185.53
    - Russia: 163.97 / 614.42
    - Turkey: 61.09 / 113.86
  - All U.S. News (ContE§ / TotE§):
    - Brazil: 25.54 / 49.88
    - Mexico: 33.70 / 92.89
    - Russia: 27.59 / 62.21
    - Turkey: 15.89 / 77.87
    - U.S.: 33.123 / 68.91
  - Interest Rate (U.S. news) (ContE§ / TotE§):
    - Brazil: 119.71 / 361.92
    - Mexico: 279.94 / 645
    - Russia: -63.69 / -238.14
    - Turkey: -5.77 / -41.10
    - U.S.: 218.51 / 506.45
- Notes on interpretation:
  - Coefficients in bold in tables indicate significance at the 5% level (HAC standard errors).
  - For Russia, no meaningful monetary policy interest rate series is available; some domestic releases occur at irregular times and/or are preannounced in speeches, potentially muting responses.

### Asymmetries, nonlinearity, and conditional patterns (Sections B, Tables 6–8, Table 7)
- Asymmetry of sign (good vs bad news):
  - Negative local real activity news produces more volatility over the total observation window than positive news.
  - Negative U.S. real economic shocks elicit stronger responses than positive shocks, particularly for Mexican, Turkish, and U.S. bonds.
  - Inflation news responses are more balanced; positive U.S. inflation surprises (signalling possible loosening of U.S. policy or lower future issuance) cause greater volatility in emerging market bonds.
- Size of surprise (big vs small):
  - Classify surprises by absolute magnitude; the study used a 70th percentile cutoff for cross-country comparability.
  - Table 7 evidence: larger surprises in U.S. data trigger a more sizable and more immediate volatility reaction than large surprises in domestic news.
  - Examples from Table 7 (ContE§ / TotE§ for lower and upper 0.70 quantiles):
    - Brazil, Domestic news: Lower quantile 89.86 / 67.32; Upper quantile 38.93 / 78.57.
    - U.S. news, Upper quantile: Brazil 32.35 / 51.08; Mexico 64.96 / 121.47; Russia 53.45 / 131.86; Turkey 35.49 / 143.52; U.S. 52.23 / 92.21.
- Regime dependence (before vs during subprime crisis):
  - With onset of financial turbulence (June 2007), response to U.S. macro releases became less pronounced; domestic news in Brazil and Turkey gained importance.
  - Table 8 selected examples (ContE§ / TotE§ before vs during crisis):
    - All Domestic News (Brazil): Before crisis -7.50 / 27.12; During crisis 45.12 / 56.05.
    - All U.S. News (Brazil): Before crisis 47.18 / 90.68; During crisis 5.90 / 7.09.
  - Interpretation: investors shifted attention away from U.S. (and German) news during the early stages of the U.S. financial crisis; intraday volatility rose but aggregate effect of U.S. surprises on emerging bond volatility became less consistent and weaker.

### Mechanisms, interpretation, and market-structure considerations
- Timing and information content:
  - Indicators released more timely and frequently tend to elicit stronger responses (e.g., weekly trade balance in Brazil vs monthly).
  - GDP, though released late, has large marginal information content for emerging markets.
- Reliability and preannouncement:
  - Perceived data reliability and forecast precision affect market reaction; high revision/noisiness dampens immediate reaction.
  - Early guidance (preliminary/advance figures, policy decision rules, preannouncement) would reduce magnitude of surprises and market reaction.
- Market-structure drivers of prolonged volatility:
  - Prolonged elevated volatility in emerging market bonds likely due to greater information asymmetries, lower liquidity, larger share of international investors, and OTC market features—leading to longer and more expensive portfolio reallocations.

### Key conclusions (Section VII)
- Immediate price re-pricing is nearly instantaneous (absorbed within a five-minute period), similar to mature bond markets.
- The repricing process is accompanied by a prolonged period of elevated volatility and trading activity in emerging bond markets—volatility remains elevated for more than one and a half hours (about twice as long as in mature markets).
- International and regional news (notably U.S. inflation and monetary policy announcements) is at least as important as local news for emerging market external bonds.
- Strong asymmetric effects are present: bad news often matters more than good news; large surprises have disproportionately large impacts.
- During the early subprime crisis period, attention shifted away from U.S. (and German) macro news toward more timely/specific domestic indicators in some emerging markets, though no general change in response patterns was identified.

*Source: _wp09147 - 9.00 a.m. EST. Table 4 reports the contemporaneous and total impact of surprises in U.S. news, obtained*

### References

### _wp09147 - References

### References

- Admati, A. R. and P. Páeiderer (1988). A Theory of Intraday Patterns: Volume and Price Variability. Review of Financial Studies 1(1), 3ñ40.
- Almeida, A., C. Goodhart, and R. Payne (1998). The E§ects of Macroeconomic News on High Frequency Exchange Rate Behavior.Journal of Financial and Quantitative Analysis 33(3), 383ñ408.
- Andersen, T. G. and T. Bollerslev (1997). Intraday Periodicity and Volatility Persistence in Financial Markets.Journal of Empirical Finance 4, 115 Ω Uñ158.
- Andersen, T. G. and T. Bollerslev (1998). Deutsche Mark-Dollar Volatility: Intraday Activity Patterns, Macroeconomic Announcements, and Longer Run Dependencies.Journal of Finance 53(1), 219ñ265.
- Andersen, T. G., T. Bollerslev, and J. Cai (2000). Intraday and Interday Volatility in the Japanese Stock Market.Journal of International Financial Markets, Institutions and Money 10, 107ñ130.
- Andersen, T. G., T. Bollerslev, F. X. Diebold, and C. Vega (2003). Micro E§ects of Macro Announcements: Real-Time Price Discovery in Foreign Exchange.American Economic Review 93(1), 38ñ62.
- Andersen, T. G., T. Bollerslev, F. X. Diebold, and C. Vega (2007). Real-time Price Discovery in Global Stock, Bond and Foreign Exchange Markets.Journal of International Economics 73, 251ñ277.
- Andersson, M. (2007). Using Intraday Data to Gauge Financial Market Responses to Fed and ECB Monetary Policy Decisions. Working Paper 726, European Central Bank.
- Andritzky, J. R., G. J. Bannister, and N. T. Tamirisa (2007). The Impact of Macroeconomic Announcements on Emerging Market Bonds.Emerging Markets Review 8, 20ñ37.
- Balduzzi, P., E. J. Elton, and C. Green (2001). Economic News and Bond Prices: Evidence from the U.S. Treasury Market.Journal of Financial and Quantitative Analysis 36(4), 523ñ543.
- Bauwens, L., W. B. Omrane, and P. Giot (2005). News Announcements, Market Activity and Volatility in the Euro/Dollar Foreign Exchange Market.Journal of International Money and Finance 24, 1108ñ1125.
- Bernanke, B. S. and K. N. Kuttner (2005). What Explains the Stock Marketís Reaction to Federal Reserve Policy?Journal of Finance 60(3), 1221ñ1257.
- Bollerslev, T. (2001). Financial Econometrics:  Past Developments and Future Challenges.Journal of Econometrics 100, 41ñ51.
- Bollerslev, T., J. Cai, and F. M. Song (2000). Intraday Periodicity, Long Memory Volatility, and Macroeconomic Announcement E§ects in the U.S. Treasury Bond Market.Journal of Empirical Finance 37, 37ñ55.
- Chang, Y. and S. J. Taylor (1998). Intraday E§ects of Foreign Exchange Intervention by the Bank of Japan. Journal of International Money and Finance 17, 191ñ210.
- Chang, Y. and S. J. Taylor (2003). Information Arrivals and Intraday Exchange Rate Volatility.Journal of International Financial Markets, Institutions and Money 13, 85ñ112.
- Cutler, D. M., J. M. Poterba, and L. H. Summers (1989). What Moves Stock Prices?Journal of Portfolio Management 15(3), 4ñ12.
- Danielsson, J. and R. Payne (2002). Real Trading Patterns and Prices in Spot Foreign Exchange Markets. Journal of International Money and Finance 21, 203ñ222.
- De Boor, C. (1978).A Practical Guide to Splines. New York: Springer-Verlag.
- DeGennaro, R. P. and R. E. Shrieves (1997). Public Information Releases, Private Information Arrival and Volatility in the Foreign Exchange Market.Journal of Empirical Finance 4(4), 295ñ315.
- Dominguez, K. M. (2006). When Do Central Bank Interventions Ináuence Intra-daily and Longer-term Exchange Rate Movements?Journal of International Money and Finance 25(7), 1051ñ1071.
- Dwyer, G. P. and R. Hafer (1989). Interest Rates and Economic Announcements.Review,Federal Reserve Bank of St. Louis48, 34ñ46.
- Ederington, L. H. and J. H. Lee (1993). How Markets Process Information: News Releases and Volatility. Journal of Finance 48(4), 1161ñ1191.
- Emerging Market Traders Association (2008a). Fourth Quarter 2007 Debt Trading Volume Survey.
- Emerging Market Traders Association (2008b). Fourth Quarter 2008 Debt Trading Volume Survey.
- Engle, R. F. and J. R. Russell (1998). Autoregressive Conditional Duration: A New Model for Irregularly Spaced Transaction Data.Econometrica 66(5), 1127ñ1162.
- Eubank, R. L. (1988).Spline Smoothing and Nonparametric Regression, Volume 90 ofStatistics, Textbooks and Monographs. New York: Marcel Dekker.
- Fleming, M. J. and E. M. Remolona (1999). Price Formation and Liquidity in the U.S. Treasury Market: The Response to Public Information.Journal of Finance 54(5), 1901ñ1915.
- Forsberg, L. and E. Ghysels (2007). Why Do Absolute Returns Predict Volatility So Well?Journal of Financial Econometrics 5, 31ñ67.
- Gau, Y.-F. and M. Hua (2007). Intraday Exchange Rate Volatility: ARCH, News and Seasonality E§ects. Quarterly Review of Economics and Finance 13, 135ñ158.
- Goodhart, C., T. Ito, and R. Payne (1996). One Day in June, 1993: A Study of the Workings of Reuters D2000-2 Electronic Foreign Exchange Trading System. In J. Frankel, G. Galli, and A. Giovannini (Eds.),The Microstructure of Foreign Exchange Markets, pp. 107ñ179. Chicago: University of Chicago Press.
- Gosnell, T. F., A. J. Keown, and J. M. Pinkerton (1996). The Intraday Speed of Stock Price Adjustment to Major Dividend Changes: Bid-Ask Bounce and Order Flow Imbalances.Journal of Banking and Finance 20(2), 247ñ266.
- Green, C. T. (2004). Economic News and the Impact of Trading on Bond Prices.Journal of Finance 54(3), 1201ñ1233.
- Hakkio, C. S. and D. K. Pearce (1985, October). The Reaction of Exchange Rates to Economic News. Economic Inquiry 23(4), 621ñ636.
- Hamilton, J. D. (1989). A New Approach to the Economic Analysis of Nonstationary Time Series and the Business Cycle.Econometrica 57, 357ñ384.
- Hamilton, J. D. (1994).Time Series Analysis. Princeton, New Jersey: Princeton University Press.
- Huang, R. D. and H. R. Stoll (1994). Market Microstructure and Stock Return Predictions.Review of Financial Studies 7(1), 179ñ213.
- JPMorgan (2008). Emerging Markets Bond Index Monitor (February).
- Kim, O. and R. E. Verrecchia (1991a). Market Reaction to Anticipated Announcements.Journal of Financial Economics 30, 273ñ309.
- Kim, O. and R. E. Verrecchia (1991b). Trading Volume and Price Reactions to Public Announcements. Journal of Accounting Research 29, 302ñ321.
- Kim, O. and R. E. Verrecchia (1997). Pre-announcement and Event-period Private Information.Journal of Accounting and Economics 24, 395ñ419.
- Li, L. and R. F. Engle (1998). Macroeconomic Announcements and Volatility of Treasury Futures. Discussion Paper 98-27, Department of Economics, University of California at San Diego.
- Mandelbrot, B. (1963). The Variation of Certain Speculative Prices.Journal of Business 36, 394ñ419.
- Melvin, M. and X. Yin (2000). Public Information Arrival, Exchange Rate Volatility, and Quote Frequency. The Economic Journal 110(465), 644ñ661.
- Mizrach, B. and C. J. Neely (2007). The Microstructure of the U.S. Treasury Market. Working Papers 2007-052, Federal Reserve Bank of St. Louis.
- Newey, W. K. and K. D. West (1987). A Simple, Positive Semi-deÖnite, Heteroskedasticity and Autocor- relation Consistent Covariance Matrix.Econometrica 55(3), 703ñ08.
- OíHara, M. (1995).Market Microstructure Theory. Cambridge, MA: Blackwell Publishers.
- Rigobon, R. and B. Sack (2006). Noisy Macroeconomic Announcements, Monetary Policy, and Asset Prices. Working Paper 12420, National Bureau of Economic Research.
- Robitaille, P. and J. Roush (2006). How Do FOMC Actions and U.S. Macroeconomic Data Announcements Move Brazilian Sovereign Yield Spreads and Stock Prices? International Finance Discussion Papers 868, Board of Governors of the Federal Reserve System.
- Veredas, D. (2006). Macroeconomic Surprises and Short-term Behaviour in Bond Futures.Empirical Eco- nomics 30, 843ñ866.
- Wongswan, J. (2006). Transmission of Information across International Equity Markets.Review of Finan- cial Studies 19(4), 1157ñ1189.

*Source: _wp09147 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09147.pdf_