## wp18274

## Source details

**Canonical URL:** [wp18274](https://www.imf.org/-/media/files/publications/wp/2018/wp18274.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2018/wp18274.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2018/wp18274.pdf.json)

---

### News dataset
- Source: Factiva news archives; focused on Reuters articles in English.
- Coverage:
  - close to 4.5 million daily articles covering 25 countries - 9 Advanced Economies (AEs) and 16 Emerging Markets (EMs) - from 1991 to 2015.
  - US-related articles represent one fourth of the sample (200 articles per day).
  - Distribution of non-US country tags: an average of 97,000 articles per country over the whole sample (20 articles per day).
  - About 60% of the corpus consists of local news (~2.5 million articles).
- Selection criteria:
  - Articles tagged with either “Economic News” or “Financial Market News” and at least one country in the sample.
  - Each article contains a timestamp, a title, a main text, and Factiva-generated topics and locations tags.
- Topic coverage: Economic policy, government finance, commodity markets, equity markets, forex, corporate news, political news; distribution of topics similar across AEs and EMs.
- Corpus composition: two article types — local news (mentions one country) and multi-country news.

### News-sentiment measures: dictionaries and construction
- Dictionaries combined:
  - Loughran and McDonald (2011) for financial texts.
  - Young and Soroka (2012) for political and economic texts.
- Expanded dictionary:
  - 7,217 negative words.
  - 3,250 positive words.
- Article-level sentiment measure s_j defined as:
  - s_j = (sum_wi j p_ij * w_ij − sum_wi j n_ij * w_ij) / (sum_wi j t_ij * w_ij),
    where p_ij = occurrences of positive word i in article j, n_ij = occurrences of negative word i, t_ij = occurrences of word i, and w_ij = weight for word i.
- Weighting schemes:
  - Baseline: w_ij = 1.
  - TF-IDF robustness check: w_i,j = log( N / N_i ), where N is the number of articles in the corpus and N_i is the number of articles in which word i is present.
- Implementation:
  - Daily sentiment index for country c is the average of s_j across articles on that day that include country c in location tags.
  - Number of articles used to estimate sentiment varies with the sentiment definition (unweighted, TF-IDF).

### Example illustrating dictionary approach
- Example article:
  - Title: Argentina’s Peronists defend Menem’s labor reforms.
  - Timestamp: 1996-09-02
  - Demonstrates that s_j provides a good indication of overall tone despite not capturing all textual nuance.

### Asset prices and related variables
- Equity returns:
  - Computed using each country’s main stock market index in local currency.
  - Daily world returns used as global control computed using the Dow Jones World Index.
- Market liquidity proxy:
  - Daily equity trading volumes from local stock exchanges.
  - Detrended log of daily trading volume: rolling average of past 60 days of log volume used as trend.
- Volatility proxy:
  - Stock market volatility measured by de-meaning each daily stock return, squaring the residual, then subtracting the past 60-day moving average of the squared residual.
- Other global proxies:
  - S&P Goldman Sachs Commodity index to measure daily changes in commodity prices (in %).
  - CBOE VIX to proxy for global volatility.

### Capital flows (EPFR equity flows)
- Data:
  - Daily equity fund flows from EPFR (“equity country flows” dataset) reporting estimated amount of equity funding in US dollars that came (or left) each country each day due to international funds’ portfolio reallocation.
  - Dataset covers 16 emerging markets between 2005 and 2015.
- Rationale and coverage:
  - As of 2013, EPFR global collected information from more than 29,000 equity funds and 18,000 fixed-income funds representing US$20 trillion of assets invested in over 80 advanced economies and EMs.
- Use:
  - EPFR flows act as a high-frequency proxy for foreign capital inflows to EMs and correlate well with Balance of Payments recorded capital flows into EMs.
- Distinctions:
  - Behavior of local vs. foreign equity funds distinguished using the fund’s official domicile.

### Empirical framework
- Estimation method:
  - Jordà’s (2005) local projection method for cumulative responses of asset prices to daily sentiment shocks, applied to individual and panel regressions.
- Baseline specification (Equation (1)) estimates cumulative equity returns Cum_R_i,t,t+h (in %) between day t and t+h as a function of:
  - country fixed effects μ_i,h;
  - lagged daily equity returns R_i,t and up to J=8 lags;
  - news-based sentiment index GoodNews_i,t and up to J=8 lags;
  - number of articles Art_i,t and up to J=8 lags;
  - de-trended log-trading volumes Vlm_i,t and up to J=8 lags;
  - volatility Vol_i,t and up to J=8 lags;
  - global controls Glob_t (daily world equity returns (in %), VIX, changes in commodity prices (in %)) and up to J=8 lags;
  - outlier and day-of-week dummies D_i,t.
- Implementation details:
  - Equal weighting scheme used in main results; TF-IDF reported in Appendix and robustness checks.
  - Lags up to 8 days; horizons estimated up to 20 days.
  - OLS estimation with Newey and West (1987) standard errors.
  - Include residual from horizon h−1 estimation as a regressor at horizon h to mitigate efficiency loss and address potential bias.

### Results — Benchmark (US and panel)
- US-only replication:
  - Regression on US news and daily US Dow Jones Industrial Index returns between 1991 and 2015.
  - Sample size: 6,260 observations.
  - Finding: Good news — one standard positive deviation in sentiment — generates positive but transitory returns.
    - Response peaks slightly below 10 basis points.
    - Not statistically different from zero after one calendar week.
- Panel results:
  - Panel covering 25 countries; estimation based on 101,170 observations.
  - Controls include World Dow Jones Index daily return, daily VIX, daily changes in commodity prices; dummies for outliers; winsorizing at 0.05%; day-of-week fixed effects.
  - Findings:
    - Positive sentiment shock implies positive and economically significant increase in equity returns.
    - Peak magnitude similar to US — peak estimated at 9 basis points — but the reversal is not complete.
    - Results robust to removing the US from the sample.

### Distinguishing local and multi-country news: identification strategy
- Corpus composition: 40% of corpus contains multi-country articles; remaining 60% implied to be local-focused.
- Highly restrictive local-news filter removed 1.5 million articles across countries to retain genuinely local news.
- Two-step procedure:
  - Re-compute daily country news sentiment excluding any article mentioning any other country.
  - Extract a common factor (“global news sentiment”) from the initial sentiment series using a Kalman filter and include it in regressions.
- Model specification (Equation 2) includes country fixed effects, lags of returns, liquidity and volatility proxies, global controls, global and local good-news sentiment indices, article counts, and outlier/day-of-week dummies.

### Effects on equity returns: local vs global sentiment shocks
- Controlling for global sentiment reduces estimated cumulative response to local news sentiment shocks:
  - Local cumulative response roughly halved: 5 basis points (with global control) versus 10 basis points (previous estimates without global control).
  - Local gains completely vanish after a week when controlling for global sentiment.
- Economic significance:
  - Median absolute deviation in the sample is 70 basis points, for both AEs and EMs.
- Global news sentiment effects:
  - Stronger and more permanent effect on returns.
  - Estimated total permanent impact around 25 basis points — roughly 5 times the size of the local sentiment shock — reached after 10 to 15 days.
  - Result only marginally affected by exclusion of the GFC.
- Interpretation:
  - Local sentiment shocks consistent with temporary “noise (or liquidity) traders” effects.
  - Global news shocks likely contain genuinely new information about fundamentals that is slowly incorporated into stock prices; alternative explanation is persistent drifts absent new hard information.

### Effects on equity fund flows
- Data and sample:
  - Daily equity flows from international mutual funds tracked by EPFR between 2005 and 2015 for 16 EMs; sample size is 23,720 observations.
- Model (Equation (3)) projects cumulative equity fund flows (CumF_i,t,t+h) in % of initial allocation at t-1 and includes analogous controls as for prices.
- Main findings:
  - Local news optimism attracts equity fund flows only temporarily.
    - Statistically significant cumulative increase at the peak of 0.01%.
    - Cannot exclude a full reversal after a week (at the 5% significance level).
  - Global news optimism generates a larger and permanent inflow into EM equity:
    - Peaks after 2 weeks around 0.1%.
    - Robust to exclusion of the GFC.
  - Heterogeneity by investor domicile:
    - Results almost entirely driven by foreign investors (investors domiciled outside the country).
    - Response of local equity investors is not significantly different from zero at all horizons and for both types of shocks (at 5% significance level).
- Magnitude note:
  - Percentages expressed as ratio of Assets Under Management before the shock (at t-1).
  - A 0.01% increase means the equity fund industry increased its stock of equity assets in country c by 0.01%.
  - Average mean deviation of equity flows in the sample is around 0.01%.

### Properties of the Global News Sentiment Index
- State dependence:
  - Effect of global news sentiment is stronger when the global equity market is down.
  - Impact of global sentiment shocks roughly four times stronger in global bear markets versus bull markets.
  - Asymmetry not observed for local news-sentiment shocks.
- Comparison with VIX:
  - Global news sentiment and VIX correlation: -0.35.
  - A one standard deviation shock in the VIX has a 1% impact on equity markets around the world — roughly six times the impact of a one-standard deviation shock in global news sentiment.
  - VIX shocks of that magnitude happen much less often; global news sentiment shocks account for more variance than VIX shocks at most horizons.
- News composition:
  - Global news sentiment generally driven by multi-country news.
  - Share of multi-country news increases significantly when global news sentiment is more extreme.
  - Multi-country news vs local news (averages reported):
    - Roughly twice as long.
    - Cover twice as many topics.
    - Use twice as many “tonal words” (positive or negative).
    - Use more extreme language based on TFIDF measures.
  - Directional differences:
    - Strongly positive global news sentiment: coverage tilts toward positive financial and corporate news in advanced economies, especially the US (except Greece-related news overrepresented in low sentiment periods).
    - Strongly negative global news sentiment: coverage tilts strongly toward economic and political news in emerging markets.

### Robustness and extensions
- Stability:
  - Findings relatively stable over time and across countries (EMs and AEs); not driven by a single episode or distinct group of countries.
  - No evidence that the effect of global news is significantly stronger now than in the 1990s, or that it affects more EMs than AEs.
- Not driven by extremes or key countries:
  - Results robust to exclusion of the Global Financial Crisis.
  - Results robust to exclusion of the US (available on request).
- Alternative sentiment measure:
  - Robust to a TF-IDF based sentiment measure.
- Appendix figures:
  - Figure 8 and Figure 9 show consistent responses of equity prices to local and global news sentiment across samples, TF-IDF measures, and country/time splits; standard errors corrected using Newey and West (1987).

### Conclusion (section summary)
- News sentiment has pervasive impact on short-term equity prices worldwide between 1991 and 2015.
- Distinct impacts:
  - Local news: temporary, smaller effects on prices and flows.
  - Global news: larger, more persistent effects, substantial role in international co-movement of asset prices.
- Role of foreign investors emphasized in transmitting sentiment shocks.

*Source: IMF Working Paper wp18274 — 1. News articles; section 2 excerpt.*

### 1. News articles

### wp18274 - 1. News articles

### News dataset
- Source: Factiva news archives; focused on Reuters articles in English.
- Coverage:
  - close to 4.5 million daily articles covering 25 countries - 9 Advanced Economies (AEs) and 16 Emerging Markets (EMs) - from 1991 to 2015.
  - US-related articles represent one fourth of the sample (200 articles per day).
  - The distribution of non-US country tags: an average of 97,000 articles per country over the whole sample (20 articles per day).
- Selection criteria:
  - Articles tagged with either “Economic News” or “Financial Market News” and at least one country in the sample.
  - Each article contains a timestamp, a title, a main text, and Factiva-generated topics and locations tags.
- Topic coverage:
  - Economic policy, government finance, commodity markets, equity markets, forex, corporate news, political news.
  - Distribution of topics similar across AEs and EMs.
- Corpus composition:
  - Two article types: local news (mentions one country) and multi-country news.
  - About 60% of the corpus consists of local news (~2.5 million articles).

### News-sentiment measures: dictionaries and construction
- Dictionaries combined:
  - Loughran and McDonald (2011) for financial texts.
  - Young and Soroka (2012) for political and economic texts.
- Expanded dictionary (include inflections present in corpus):
  - 7,217 negative words.
  - 3,250 positive words.
- Article-level sentiment measure s_j defined as:
  - s_j = (sum_wi j p_ij * w_ij − sum_wi j n_ij * w_ij) / (sum_wi j t_ij * w_ij),
    where p_ij = occurrences of positive word i in article j, n_ij = occurrences of negative word i, t_ij = occurrences of word i, and w_ij = weight for word i.
- Weighting schemes:
  - Baseline: w_ij = 1 (each word contributes proportionally to its frequency in the article).
  - TF-IDF robustness check: w_i,j = log( N / N_i ), where N is the number of articles in the corpus and N_i is the number of articles in which word i is present (gives more weight to rarer words across documents).
- Implementation notes:
  - After computing s_j for each article, a daily sentiment index for country c is the average of s_j across articles on that day that include country c in location tags.
  - Number of articles used to estimate sentiment varies with the sentiment definition (unweighted, TF-IDF); the sample of articles is adjusted accordingly.

### Example illustrating dictionary approach
- Example article (metadata preserved from corpus):
  - Title: Argentina’s Peronists defend Menem’s labor reforms.
  - Timestamp: 1996-09-02
  - Text excerpt discusses President Menem’s proposals for more flexible labor laws, unemployment insurance, suspending collective wage deals, and replacing redundancy payouts with unemployment insurance.
  - Topics: Labor/Personnel Issues, Corporate/Industrial News, Economic/Monetary Policy, Economic News, Political/General News, Labor Issues, Domestic Politics
  - Locations: Argentina, Latin America, South America
- Purpose: demonstrates that while s_j does not capture all textual nuance, it provides a good indication of overall tone.

### Asset prices and related variables
- Equity returns:
  - Computed using each country’s main stock market index in local currency.
  - Daily world returns used as global control are computed using the Dow Jones World Index.
- Market liquidity proxy:
  - Daily equity trading volumes from local stock exchanges.
  - Detrended log of daily trading volume: rolling average of past 60 days of log volume used as trend (Campbell, Grossman, and Wang (1993); Tetlock (2007)).
- Volatility proxy:
  - Stock market volatility measured by de-meaning each daily stock return, squaring the residual, then subtracting the past 60-day moving average of the squared residual.
- Other global proxies:
  - S&P Goldman Sachs Commodity index to measure daily changes in commodity prices (in %).
  - CBOE VIX to proxy for global volatility.

### Capital flows (EPFR equity flows)
- Data:
  - Daily equity fund flows from EPFR (“equity country flows” dataset) reporting estimated amount of equity funding in US dollars that came (or left) each country each day due to international funds’ portfolio reallocation.
  - Dataset of equity flows covers 16 emerging markets between 2005 and 2015.
- Rationale and coverage:
  - EPFR global dataset tracks asset allocation of equity and debt funds domiciled in developed countries and key offshore financial centers.
  - As of 2013, the EPFR global was collecting information from more than 29,000 equity funds and 18,000 fixed-income funds representing US$20 trillion of assets invested in over 80 advanced economies and EMs.
  - EPFR flows have been used widely as a high-frequency proxy for foreign capital inflows to EMs and correlate well with Balance of Payments recorded capital flows into EMs.
- Distinctions:
  - After estimating effect of sentiment shocks on total equity inflows (or outflows), behavior of local vs. foreign equity funds is distinguished using the fund’s official domicile.

### Empirical framework
- Estimation method:
  - Jordà’s (2005) local projection method for cumulative responses of asset prices to daily sentiment shocks, applied to individual and panel regressions.
  - Preferred because of uncertainty about timing, strength, and linearity of responses over a long sample spanning AEs and EMs.
- Baseline specification (Equation (1)) estimates cumulative equity returns Cum_R_i,t,t+h (in %) between day t and t+h as a function of:
  - country fixed effects μ_i,h;
  - lagged daily equity returns R_i,t and up to J=8 lags;
  - news-based sentiment index GoodNews_i,t and up to J=8 lags;
  - number of articles Art_i,t and up to J=8 lags;
  - de-trended log-trading volumes Vlm_i,t and up to J=8 lags;
  - volatility Vol_i,t and up to J=8 lags;
  - global controls Glob_t (daily world equity returns (in %), VIX, changes in commodity prices (in %)) and up to J=8 lags;
  - outlier and day-of-week dummies D_i,t.
- Implementation details:
  - Equal weighting scheme used to create the sentiment index in main results; TF-IDF results reported in Appendix and robustness checks.
  - Lags up to 8 days; horizons estimated up to 20 days.
  - OLS estimation with Newey and West (1987) standard errors to correct for serial autocorrelation and heteroskedasticity.
  - Include residual from horizon h−1 estimation as a regressor at horizon h to mitigate efficiency loss and address potential bias (Jordà (2005); Teulings and Zubanov (2014)).
- Controls and deviations from prior work (Tetlock 2007):
  - Control for number of articles published each day to separate volume of news from average tone.
  - Include global proxies (global returns, VIX, commodity changes) to capture global co-movement.
  - Estimate cumulative responses up to 20 days (vs. 5 trading days in Tetlock (2007)).

### Results — Benchmark (US and panel)
- US-only replication:
  - Regression on US news and daily US Dow Jones Industrial Index returns between 1991 and 2015.
  - Sample size: 6,260 observations (original Tetlock (2007) used 4,000).
  - Finding: Good news — measured by a one standard positive deviation in sentiment — generates positive but transitory returns.
    - Response peaks slightly below 10 basis points.
    - Not statistically different from zero after one calendar week.
- Panel results:
  - Panel covering 25 countries; estimation based on 101,170 observations.
  - Controls: World Dow Jones Index daily return, daily VIX, daily changes in commodity prices; dummies for outliers; winsorizing at 0.05%; day-of-week fixed effects.
  - Findings:
    - Effect of news sentiment on asset prices is pervasive beyond the US.
    - Positive sentiment shock implies positive and economically significant increase in equity returns.
    - Peak magnitude similar to US — peak estimated at 9 basis points — but the reversal is not complete.
    - Results robust to removing the US from the sample (result not driven by US).
- Comparison to Tetlock (2007):
  - Tetlock (2007): one-standard deviation change in pessimism generated a drop of 8.1 basis points in Dow Jones returns the next day, almost completely reversed by the end of the trading week.

### Global vs. local news sentiment (summary)
- Corpus composition:
  - Local news (~60% of corpus, ~2.5 million articles) tends to convey specialized content and only mentions one country.
- Analysis approach:
  - The dataset distinguishes between local and multi-country news; country-level daily sentiment indexes are computed from articles that tag the country in location tags.
- Implication:
  - The paper examines how local versus global/multi-country news contribute to sentiment-driven asset price responses (details and formal results continued beyond provided excerpt).

*Source: IMF Working Paper wp18274 - 1. News articles (pdf).*

### section 2, discussing labor market laws in Argentina.

### wp18274 - section 2, discussing labor market laws in Argentina

### Distinguishing local and multi-country news: identification strategy
- Corpus composition: 40% of corpus contains multi-country articles; remaining 60% implied to be local-focused.
- Highly restrictive local-news filter: removed 1.5 million articles across countries to retain genuinely local (country-specific) news.
- Two-step procedure:
  - Re-compute daily country news sentiment excluding any article mentioning any other country.
  - Extract a common factor (“global news sentiment”) from the initial sentiment series using a Kalman filter and include it in regressions to contrast local vs global news effects and control residual cross-country correlation.
- Model specification: cumulative return projection equation (Equation 2) includes:
  - country fixed effects (휇푖,ℎ),
  - lags of returns (푅푖,푡−푗),
  - liquidity proxy (푉푙푚푖,푡−푗),
  - volatility proxy (푉표푙푖,푡−푗),
  - global controls (퐺푙표푏푡−푗),
  - global good-news and local good-news sentiment indices (퐺푙표푏푎푙_퐺표표푑푁푒푤푠, 퐿표푐푎푙_퐺표표푑푁푒푤푠),
  - article counts (퐴푟푡푖,푡−푗),
  - outlier and day-of-week dummies (퐷푖,푡ℎ).

### Effects on equity returns: local vs global sentiment shocks
- Controlling for global sentiment reduces the estimated cumulative response to local news sentiment shocks:
  - Local cumulative response roughly halved: 5 basis points (with global control) versus 10 basis points (previous estimates without global control).
  - Local gains completely vanish after a week when controlling for global sentiment.
- Economic significance:
  - Sentiment shocks remain economically significant despite smaller magnitude.
  - Median absolute deviation in the sample is 70 basis points, for both AEs and EMs.
- Global news sentiment effects:
  - Global news sentiment shocks have a stronger and more permanent effect on returns.
  - Estimated total permanent impact around 25 basis points — roughly 5 times the size of the local sentiment shock — reached after 10 to 15 days.
  - Result only marginally affected by exclusion of the GFC.
- Interpretation:
  - Local sentiment shocks consistent with temporary “noise (or liquidity) traders” effects.
  - Global news shocks likely contain genuinely new information about fundamentals that is slowly incorporated into stock prices; alternative explanation is persistent drifts absent new hard information.

### Effects on equity fund flows
- Data: daily equity flows from international mutual funds tracked by EPFR between 2005 and 2015 for 16 EMs; sample size is 23,720 observations.
- Model (Equation 3) projects cumulative equity fund flows (퐶푢푚퐹푖,푡,푡+ℎ) in % of initial allocation at t-1 and includes analogous controls as for prices.
- Main findings:
  - Local news optimism attracts equity fund flows only temporarily.
    - Statistically significant cumulative increase at the peak of 0.01%.
    - Cannot exclude a full reversal after a week (at the 5% significance level).
  - Global news optimism generates a larger and permanent inflow into EM equity:
    - Peaks after 2 weeks around 0.1%.
    - Robust to exclusion of the GFC.
  - Heterogeneity by investor domicile:
    - Results almost entirely driven by foreign investors (investors domiciled outside the country).
    - Response of local equity investors is not significantly different from zero at all horizons and for both types of shocks (at 5% significance level).
- Magnitude note:
  - Percentages expressed as ratio of Assets Under Management before the shock (at t-1).
  - A 0.01% increase means the equity fund industry increased its stock of equity assets in country c by 0.01%.
  - Average mean deviation of equity flows in the sample is around 0.01%.

### Properties of the Global News Sentiment Index
- Three key findings summarized:
  1. State dependence:
     - Effect of global news sentiment is stronger when the global equity market is down.
     - Impact of global sentiment shocks roughly four times stronger in global bear markets versus bull markets.
     - Asymmetry not observed for local news-sentiment shocks.
  2. Comparison with VIX:
     - Global news sentiment and VIX correlation: -0.35.
     - Spikes in VIX are always matched by synchronized drops in global news sentiment, but many movements in global sentiment are not matched by changes in the VIX.
     - Quantitative impact: a one standard deviation shock in the VIX has a 1% impact on equity markets around the world — roughly six times the impact of a one-standard deviation shock in global news sentiment.
     - Frequency: VIX shocks of that magnitude happen much less often; global news sentiment shocks account for more variance than VIX shocks at most horizons.
  3. News composition:
     - Global news sentiment is generally driven by multi-country news.
     - Share of multi-country news increases significantly when global news sentiment is more extreme (either positively or negatively).
     - Multi-country news vs local news (averages reported):
       - Roughly twice as long.
       - Cover twice as many topics.
       - Use twice as many “tonal words” (positive or negative).
       - Use more extreme language based on TFIDF measures.
     - Directional differences:
       - When global news sentiment is strongly positive, coverage tilts toward positive financial and corporate news in advanced economies, especially the US (except Greece-related news overrepresented in low sentiment periods).
       - When global news sentiment is negative, coverage tilts strongly toward economic and political news in emerging markets.

### Robustness and extensions
- Stability:
  - Findings relatively stable over time and across countries (EMs and AEs); not driven by a single episode or distinct group of countries.
  - No evidence that the effect of global news is significantly stronger now than in the 1990s, or that it affects more EMs than AEs.
- Not driven by extremes or key countries:
  - Results robust to exclusion of the Global Financial Crisis.
  - Results robust to exclusion of the US (available on request).
- Alternative sentiment measure:
  - Robust to a TF-IDF based sentiment measure.

### Conclusion (section summary)
- News sentiment has pervasive impact on short-term equity prices worldwide between 1991 and 2015.
- Distinct impacts:
  - Local news: temporary, smaller effects on prices and flows.
  - Global news: larger, more persistent effects, substantial role in international co-movement of asset prices.
- Role of foreign investors emphasized in transmitting sentiment shocks.

*Source: wp18274 - section 2, discussing labor market laws in Argentina (excerpt provided).*

### References

### References

### Key cited works
- Baker, S. R., Bloom, N., and Davis, S. J., 2016, Measuring Economic Policy Uncertainty, Quarterly Journal of Economics, 131 (4), 1593-1636.
- Baker, M., Wurgler, J. (2007). “Investor sentiment in the stock market.” Journal of Economic Perspectives 21, 129-151.
- Broner, Fernando, Tatiana Didier, Aitor Erce, and Sergio L. Schmukler, 2013, “Gross Capital Flows: Dynamics and Crises,” Journal of Monetary Economics Vol. 60, No. 1, pp. 113-33.
- Bruno, Valentina, and Hyun Song Shin. 2014. Cross-border banking and global liquidity. Review of Economic Studies 82(2): 535–564.
- Campbell, John, Sanford Grossman, and Jiang Wang, 1993, Trading volume and serial correlation in stock returns, Quarterly Journal of Economics 108, 905–939.
- Da, Z., Engelberg, J., Gao, P., (2014b). “The sum of all FEARS: investor sentiment and asset prices.” Review of Financial Studies 28, 1-32.
- Cerutti, Eugenio M., Stijn Claessens, & Damien Puy, 2015. "Push Factors and Capital Flows to Emerging Markets; Why Knowing Your Lender Matters More Than Fundamentals," IMF Working Papers 15/127.
- Fratzscher, Marcel, 2012, “Capital flows, Push versus pull factors and the global financial crisis” Journal of International Economics, Vol. 88, No. 2, pp. 341-356.
- Gabaix, Xavier, 2016, “Power Laws in Economics: an Introduction” Journal of Economic Perspectives, Volume 30, Number 1, Pages 185–206.
- Garcia, D., 2013, “Sentiment during recessions,” Journal of Finance, 68(3), 1267–1300.
- Gentzkow, Matthew, Bryan T. Kelly, and Matt Taddy. 2017. “Text as Data.” NBER Working Paper.
- Ghosh, Atish R., Mahvash Qureshi, Jun Il Kim, and Juan Zalduendo, 2014, “Surges,” Journal of International Economics, Vol. 92, No. 2, pp. 266-85.
- Jordà, O, 2005, “Estimation and Inference of Impulse Responses by Local Projections” American Economic Review 95(1), March 2005, 161-182.
- Jordà, O, M Schularick, A M Taylor, and F Ward (2018), “Global financial cycles and risk premiums”, CEPR Discussion Paper 12969.
- Jotikasthira, Chotibhak, Christian Lundblad, and Tarun Ramadorai, 2012, “Asset Fire Sales and Purchases and the International Transmission of Funding Shocks,” Journal of Finance, Vol. 67, No. 6, pp. 2015–050.
- Loughran, T., and B. McDonald, 2011, “When Is a Liability Not a Liability? Textual Analysis, Dictionaries, and 10-Ks,” Journal of Finance, 66(1), 35–65.
- Obstfeld, Maurice. 2015. Trilemmas and trade-offs: Living with financial globalization. BIS Working Paper 480.
- Puy, Damien, 2016. "Mutual funds flows and the geography of contagion," Journal of International Money and Finance, vol. 60(C), pages 73-93.
- Raddatz, Claudio, and Sergio Schmukler, L., 2012, “On the International Transmission of Shocks: Micro-evidence from Mutual Fund Portfolios,” Journal of International Economics, Vol. 88, No. 2, pp. 357-74.
- Rey, Hélène, 2013, “Dilemma not Trilemma: The Global Financial Cycle and Monetary Policy Independence,” Proceedings of the 2013 Federal Reserve Bank of Kansas City Economic Symposium at Jackson Hole, pp. 285-333.
- Shiller, Robert J., 2000, Irrational Exuberance, Princeton University Press, Princeton, NJ.
- Tetlock, P. C., 2007, “Giving content to investor sentiment: The role of media in the stock market,” Journal of Finance, 62(3), 1139–1168.
- Teulings C, and N. Zubanov, 2014 “Is Economic Recovery a Myth? Robust Estimation of Impulse Responses”, Journal of Applied Econometrics, 29: 497-514.
- Young, Lori, and Stuart Soroka. 2012. “Affective News: The Automated Coding of Sentiment in Political Texts.” Political Communication, 29: 205–231.

### Appendix I — News articles: stylized facts
- Country and time coverage: distribution of country tags reported for the corpus of articles (table referenced).
- Main topics covered (all countries):
  - Most frequent tags include: Commodity & Financial Markets News; Economic News; Commodity markets; Economic & Monetary Policy; Equity Markets; Government Finance; Money and Forex; Economic Performance; Derivative Securities; Trade and External Payments.
  - Note: tags are not mutually exclusive; articles can be tagged across several categories.
- Example global news article (title and excerpt preserved):
  - Title: Fears of Brazilian devaluation hit emerging markets
  - Timestamp: 1998-09-11
  - Key reported figures and observations from the article:
    - A huge exodus of dollars from Brazil's foreign exchange markets, estimated at over $2 billion, panicked the Sao Paulo stock market into a plunge of nearly 16 percent.
    - Brazil's central bank hiked its key interest rate overnight by 20 points to nearly 50 percent to try to halt the massive outflows.
    - Reserves reported "around $50 billion".
    - Potential widening of Budapest's 4.5 percent wide trading band for the forint discussed.
    - Mexican peso quoted at 10.65 bid to the dollar by 1215 GMT, just off Thursday's historic low of 10.685.
    - Multiple currency quotations and comparisons reported (preserved as in article): Chinese yuan CNY=) at 8.279 vs 8.2798; New Taiwanese dollar TWD=) 34.47 vs 34.4; Indonesian rupiah IDR=) 11,600 vs 11,900; Thai baht THB=TH) at 40.65 per dollar vs 40.7; Philippine peso PHP=) 43.4 per dollar vs 43.6; South Korean won KRW=) at 1,365 per dollar vs 1,367; Indian rupee INR=) 42.41 per dollar vs 42.4; Russian rouble RUB=) on MICEX Selt electronic trading system at 10.51/13.15 per dollar vs average rate of 12.375 on Thursday; Zloty 1.31 percent above target basket parity vs 3.97 percent at Thursday's fixing; Mark/Czech crown DEMCZK=) at 18.03 bid vs 17.838; Hungarian forint DEMHUF=) unchanged from Thursday at 2.25 percent below parity against a target basket; Slovak crown DEMSKK=) fixed at 5.35 percent below target basket vs 5.80 percent on Thursday; Ukrainian hryvnia UAH=) unchanged at 3.10 per dollar; Romanian leu ROL=) at 9,045 per dollar vs 9,025; Israeli shekel ILS=) 3.8508 bid on dollar from Thursday's 3.8568; South African rand ZAR=) 6.3 per dollar vs 6.2555; Kenyan shilling KES=) at 59.8 per dollar vs 59.9; Mexican peso MXN=) at 10.65 per dollar vs 10.48; Brazil's real BRL=) at 1.1786 per dollar vs 1.1789; Venezuela bolivar VEB=) unchanged at 586.9 per dollar.
  - Topics listed for article: Money/Forex Markets, Foreign Exchange News, Commodity/Financial Market News.
  - Locations tagged for article include: Africa, Argentina, Asia, Brazil, Central America, China, Emerging Market Countries, Eastern Asia, European Union Countries, Central/Eastern Europe, Europe, Hong Kong, Hungary, Indonesia, Japan, Latin America, Mexico, North America, Poland, Russia, South Africa, South America, Southeast Asia, Southern Africa, United Kingdom, United States, Arizona, CIS Countries, Western U.S., Western Europe.

### Appendix II — Asset prices coverage
- Stock indices coverage table (country, Sample_Start, Sample_End, Stock Index):
  - Argentina: 1/2/1991 — 12/31/2015 — ARGENTINA MERVAL - PRICE INDEX
  - Brazil: 1/2/1991 — 12/31/2015 — BRAZIL BOVESPA - TOT RETURN IND
  - Chile: 1/2/1997 — 12/30/2015 — CHILE SANTIAGO SE GENERAL (IGPA) - PRICE INDEX
  - China: 1/2/1991 — 12/31/2015 — SHANGHAI SE COMPOSITE - PRICE INDEX
  - Germany: 1/2/1997 — 12/31/2015 — DAX 30 PERFORMANCE - PRICE INDEX
  - Spain: 1/2/1997 — 12/31/2015 — IBEX 35 - PRICE INDEX
  - France: 1/1/1997 — 12/31/2015 — FRANCE CAC 40 - PRICE INDEX
  - Greece: 1/2/1997 — 12/31/2015 — ATHEX COMPOSITE - PRICE INDEX
  - Indonesia: 1/2/1991 — 12/31/2015 — IDX COMPOSITE - PRICE INDEX
  - Ireland: 1/2/1997 — 12/30/2015 — IRELAND SE OVERALL (ISEQ) - PRICE INDEX
  - India: 1/2/1991 — 12/31/2015 — S&P BSE (SENSEX) 30 SENSITIVE - PRICE INDEX
  - Italy: 1/2/1997 — 12/30/2015 — FTSE MIB INDEX - PRICE INDEX
  - Japan: 1/1/1991 — 12/31/2015 — NIKKEI 225 STOCK AVERAGE - PRICE INDEX
  - Korea: 1/3/1991 — 12/31/2015 — KOREA SE COMPOSITE (KOSPI) - PRICE INDEX
  - Mexico: 1/2/1991 — 12/31/2015 — MEXICO IPC (BOLSA) - PRICE INDEX
  - Malaysia: 1/2/1991 — 12/31/2015 — FTSE BURSA MALAYSIA KLCI - PRICE INDEX
  - Peru: 1/2/1997 — 12/31/2015 — S&P/BVL GENERAL(IGBVL) - PRICE INDEX
  - Philippines: 1/2/1991 — 12/30/2015 — PHILIPPINE SE I(PSEi) - PRICE INDEX
  - Poland: 1/1/1997 — 12/31/2015 — WARSAW GENERAL INDEX - TOT RETURN IND
  - Portugal: 1/2/1997 — 12/30/2015 — PORTUGAL PSI-20 - PRICE INDEX
  - Russia: 12/30/1991 — 12/31/2015 — RUSSIA RTS INDEX - PRICE INDEX
  - Thailand: 1/2/1991 — 12/30/2015 — BANGKOK S.E.T. - PRICE INDEX
  - Turkey: 1/2/1991 — 12/31/2015 — BIST NATIONAL 100 - PRICE INDEX
  - United States: 1/1/1991 — 12/31/2015 — DOW JONES INDUSTRIALS - PRICE INDEX
  - South Africa: 1/1/1997 — 12/31/2015 — FTSE RAFI

### Appendix III — Robustness and extensions (summary of methods and results)
- Robustness checks and extensions are derived using equation 2.
- Figure 8:
  - Plots response of local asset prices to local and global news sentiment using TF-IDF measures of news sentiment.
  - Results provided for the full sample (mirrors Figure 2.A).
  - Results are unchanged when excluding the GFC or restricting attention to AEs or EMs.
  - Note: t denotes the number of days after the shock.
  - Blue line: cumulative response of equity prices to local news sentiment shocks.
  - Green line: cumulative response to global news sentiment shocks.
  - Standard errors corrected for serial auto-correlation and heteroskedasticity using a Newey and West (1987) estimator.
- Figure 9:
  - Plots response across time and groups of countries using the standard measure of sentiment.
  - Note: t denotes the number of days.
  - Blue line: cumulative response of equity prices to local news sentiment shocks.
  - Green line: cumulative response to global news sentiment shocks.
  - Standard errors corrected for serial auto-correlation and heteroskedasticity using a Newey and West (1987) estimator.
  - Panels/splits shown: EM (1991-1999), EM (1999-2007), EM (2007-2015), AE (1991-1999), AE (1999-2007), AE (2007-2015).

*Italic: IMF Working Paper — References and Appendix (wp18274).*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2018/wp18274.pdf_