## _wp13122

## Source details

**Canonical URL:** [_wp13122](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13122.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13122.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13122.pdf.json)

---

### Introduction — background and policy motivation
- The paper studies whether foreign investors are a “stabilizing” or “destabilizing” influence in an emerging equity market, focusing on behavior in extreme events (the tails) rather than average-day behavior.
- Four policy-relevant questions:
  1. Do foreign investors exacerbate a domestic crisis by withdrawing capital on a large scale during the crisis?
  2. Is there asymmetric behavior, with different responses to very good versus very bad days in the local economy?
  3. Are foreign investors “big fish in a small pond” — do large transactions kick off substantial temporary mean-reverting distortions in equity or currency markets of an illiquid emerging market?
  4. When there is stress in the global financial system, do foreign investors withdraw capital on a large scale and act as a vector of crisis transmission?
- Methodological contribution: adapt the non-parametric event study to focus on extreme events (tail-focused event study) and use bootstrap inference to avoid parametric assumptions.
- Empirical illustration: India (dataset and results summarized below).
- Causality caveat: temporal directionality consistent with Granger causality does not imply exogenous causal identification; both variables are endogenous and may respond to common shocks.

### Methodology and data — data coverage, event definition, and inference
- Dataset:
  - Period: 15 February 2000 to 29 July 2011.
  - Observations (N): 2,035 observations of daily data (text also refers to “2,036 observations” when discussing tails).
- Main daily series:
  - Indian stock market returns (Nifty; expressed as log differences).
  - Net FII inflow (foreign institutional investor activity), scaled as FII/MktCap (vertical axis units multiplied by 10,000 in Figure 2) to address non-stationarity.
  - US S&P 500 returns.
  - Call money rate (weighted by volume of trades, expressed as a percentage per annum); first differences used.
  - Nominal Rupee – US Dollar exchange rate (percent changes used; negative changes denote nominal appreciation of the Rupee versus the US Dollar).
- Institutional context for FII data:
  - Foreign investors invest via registered “foreign institutional investors” (FIIs).
  - Default foreign ownership limit in a given firm: 24%; shareholder resolution can raise it up to 98%.
  - No more than 20 firms at any time where FIIs lack headroom; selling and repatriation unrestricted.
  - Raw net FII time-series obtained from custodian-supplied aggregate data; scaled by market capitalization to obtain stationary FII/MktCap.
- Event definition:
  - Events are dates with extreme values (tail observations) of returns or flows.
  - Main tail cutoffs: 5% tails (upper and lower 5%); comparisons include 2.5% tails.
  - Baseline event window: five market days before and after the event (11-day window); robustness check uses 20-day window (10 days on each side).
  - Two treatments of clustered tail values:
    1. Fuse consecutive extreme events of the same direction into a single clustered event (primary analysis).
    2. Analyze only uncontaminated events (no other event within the event window) as robustness.
  - Formal uncontaminated-event definition uses quantiles Q(x,q) and window width W; uncontaminated upper-tail set E_u+ defined by absence of other extreme events within ±W days.
  - Fusion for runs: returns on days of a run are summed to create a single event return; event-window values adjusted accordingly.
- Inference:
  - Bootstrap-based event-study procedure sampling at the event level with replacement.
  - Bootstrap repetition: 1000 resamples to obtain bootstrap distribution and percentile confidence intervals.
- Relation to VARs:
  - Tail event-study is analogous to a tail-VAR impulse response; Monte Carlo shows tail event-study can detect relationships present only in tails that average VARs miss.
- Research questions tested via separate event studies for domestic extreme days, extreme FII-flow days, and extreme S&P 500 days.

### Results — summary of empirical findings (India)
- FII flows around very positive Nifty days:
  - On the event date, FII investment has a statistically significant positive value.
  - Location estimator: +1 basis points (of the overall market capitalization) net FII purchase on the event date; null of 0 rejected.
  - In the four trading days after the event, point estimator adds another 2 basis points.
  - Evidence consistent with slow or positive feedback trading by FIIs after very good Nifty days.
  - Point estimate is 1.4 times the standard deviation of the daily FII time-series; 95% confidence band from 0.26 to 2.56 standard deviations.
- FII flows around very bad Nifty days:
  - No evidence of FII selling prior to the event date.
  - No evidence of positive feedback trading after the event; slight evidence of positive foreign investment in the event window.
  - No evidence that FII outflows in response to very bad days trigger a crisis over the sample period.
- Nifty returns responding to extreme FII flows:
  - Positive FII-flow extremes:
    - Nifty returns rise before very good FII-flow days (joint response to good news).
    - Large FII inflows associated with an extremely positive Nifty day.
    - Post-event profile is flat; null of a flat profile cannot be rejected — consistent with an efficient response (no overshooting/undershooting).
  - Very bad FII-flow days:
    - Not preceded by large drops in Nifty returns.
    - Do not trigger further negative returns; cumulative profile relatively flat after the sharp outflow.
    - On the day of sharp outflows, there is a drop in Nifty of one standard deviation of the daily series, but not exacerbated subsequently.
- FII flows responding to S&P 500 extremes:
  - Very good S&P 500 days:
    - Associated with strong FII inflows into India worth at least twice its standard deviation, both before and after the event.
    - Response stronger than for domestic Nifty positive events — FIIs transmit good S&P 500 news into India.
  - Very bad S&P 500 days:
    - No marked impact on FII flows; pattern similar to response to very bad Nifty days.
    - Asymmetry: FIIs react strongly to very good S&P 500 days but not symmetrically to very bad S&P 500 days.
  - Nifty responds on the event date in both directions with an approximately flat trajectory thereafter.
- Knock-on impacts on interest rates and exchange rates:
  - Call money rate:
    - Exceptionally large FII inflows have imperceptible effects on the call money rate.
    - FII outflows appear followed by some rise in interest rates, but confidence intervals are very wide; effect uncertain.
    - Endogenous policy responses (central bank intervention, sterilization) could influence observed patterns.
  - Exchange rate (INR/USD):
    - Large FII inflows associated with Rupee appreciation; large FII outflows associated with Rupee depreciation.
    - For outflows, depreciation appears to precede the extreme event.
    - Impacts correspond to 1.4 and 2.85 times the standard deviation for appreciation and depreciation respectively.
    - Results do not support the view that large inflows or outflows trigger a panic in the currency.
- Robustness:
  - Using VIX instead of S&P 500 yields qualitatively similar patterns: declines in VIX associated with increases in foreign investment into India; increases in VIX do not show statistically significant FII responses.
- Main empirical conclusion:
  - In the India sample, foreign investors are not a vector of crisis transmission into the Indian equity market over the sample period; extreme events yield relatively benign outcomes.

### Key statistics and event distributions (preserved exactly as in source)
- Observations (N): 2036 for all series listed.
- Table 1 — selected summary statistics:
  - Daily net FII inflow (bps to COSPI):
    - Minimum: -10.87
    - 5%: -2.24
    - 25%: -0.36
    - Median: 0.48
    - Mean: 0.59
    - 75%: 1.41
    - 95%: 3.55
    - Maximum: 30.16
    - Standard deviation: 2.12
    - IQR: 1.77
  - Daily Nifty returns (per cent):
    - Minimum: -13.97
    - 5%: -3.28
    - 25%: -0.93
    - Median: 0.16
    - Mean: 0.06
    - 75%: 1.08
    - 95%: 2.95
    - Maximum: 16.23
    - Standard deviation: 1.98
    - IQR: 2.01
  - Daily returns on S&P 500 (per cent):
    - Minimum: -9.47
    - 5%: -2.44
    - 25%: -0.68
    - Median: 0.09
    - Mean: 0.00
    - 75%: 0.72
    - 95%: 2.25
    - Maximum: 12.40
    - Standard deviation: 1.57
    - IQR: 1.40
  - Daily returns on INR/USD exchange rate (bps):
    - Minimum: -417.09
    - 5%: -62.76
    - 25%: -13.63
    - Median: 0.00
    - Mean: 0.07
    - 75%: 12.29
    - 95%: 69.10
    - Maximum: 287.02
    - Standard deviation: 43.56
    - IQR: 25.92
  - Daily change in the call money rate (bps):
    - Minimum: -18040.30
    - 5%: -1735.37
    - 25%: -170.10
    - Median: 0.00
    - Mean: -1.04
    - 75%: 140.81
    - 95%: 1671.22
    - Maximum: 27677.69
    - Standard deviation: 1899.29
    - IQR: 310.91
- Event counts and runs at 5% tails (preserved counts as in Tables 2 and 3):
  - Totals across 2000–2011 (Table 5):
    - S&P 500: 5% Good days total 67; 5% Bad days total 59
    - Nifty: 5% Good days total 66; 5% Bad days total 60
    - FII (bps change): 5% Good days total 80; 5% Bad days total 70
    - INR (% change): 5% Good days total 60; 5% Bad days total 66
    - Call rate (bps change): 5% Good days total 36; 5% Bad days total 30
  - Runs distribution (lower tail run lengths Two, Three, Four, Five):
    - Nifty: 3, 0, 0, 0
    - S&P 500: 3, 0, 0, 0
    - FII: 7, 4, 0, 3
    - INR: 8, 0, 0, 0
    - Call rate: 0, 1, 0, 0
  - Runs distribution (upper tail run lengths Two, Three, Four, Five):
    - Nifty: 4, 0, 1, 0
    - S&P 500: 1, 1, 0, 0
    - FII: 12, 3, 0, 0
    - INR: 6, 1, 0, 0
    - Call rate: 3, 0, 0, 0
- Quantile values for 5% tails (Table 4 — preserved exactly):
  - Lower tail quantiles (Min., 25%, Median, Mean, 75%, Max.):
    - Nifty: Min. -13.97; 25% -5.24; Median -4.21; Mean -3.67; 75% -3.84; Max. -3.43
    - S&P 500: Min. -7.16; 25% -4.09; Median -3.05; Mean -4.85; 75% -2.71; Max. -2.44
    - FII: Min. -0.11; 25% -0.04; Median -3.21; Mean -3.76; 75% -0.03; Max. -0.02
    - INR: Min. -4.17; 25% -1.16; Median -0.90; Mean -1.17; 75% -0.75; Max. -0.63
    - Call rate: Min. -93.65; 25% -27.36; Median -22.15; Mean -27.22; 75% -18.70; Max. -17.45
  - Upper tail quantiles (Min., 25%, Median, Mean, 75%, Max.):
    - Nifty: Min. 2.96; 25% 3.26; Median 3.98; Mean 3.32; 75% 4.96; Max. 17.30
    - S&P 500: Min. 2.27; 25% 2.45; Median 2.90; Mean 4.75; 75% 3.52; Max. 9.91
    - FII: Min. 0.04; 25% 0.04; Median 4.81; Mean 5.69; 75% 0.06; Max. 0.17
    - INR: Min. 0.69; 25% 0.80; Median 0.96; Mean 1.15; 75% 1.16; Max. 3.40
    - Call rate: Min. 16.76; 25% 18.03; Median 22.91; Mean 162.59; 75% 26.46; Max. 4607.69

### Conclusions, caveats, and directions for future research
- Main conclusion for India sample:
  - Tail-focused event study with bootstrap inference yields relatively benign results: extreme events do not support concerns that foreign investors produced destabilizing panic behavior leading to crises over the sample period.
- Caveats:
  - Causality is not established; temporal ordering does not imply exogeneity.
  - Results apply to the aggregated equity market (Nifty); heterogeneity across firms, smaller markets, and illiquid securities may reveal different patterns.
  - Short-term debt flows (local-currency and foreign-currency) may display different risks and warrant separate study.
- Suggested extensions:
  - Apply the methodology to other countries, especially smaller and less integrated markets.
  - Examine other asset classes including debt capital.
  - Investigate cross-sectional variation across multiple firms to identify where FIIs might be “big fish in small ponds.”
  - Further refinements of event-study econometrics and applications to short-term debt flows.
- Reproducibility:
  - Full source code developed for the paper has been released into the public domain to assist replication and downstream research.

*Source: _wp13122 — IMF working paper (selected sections: 1. Introduction; 3. Methodology and Data; 4. Results; References).*

### 1. Introduction

### 1. Introduction

### Background and policy motivation
- The impact of international capital flows on emerging markets has occupied the attention of policy makers and economists for many decades.
- While developing countries have eased capital controls in recent decades, the debate is not settled, and many policy makers continue to be concerned about the problems associated with financial globalization.
- These concerns have become more prominent after the global crisis, with the suggestion by the IMF that capital controls should be viewed more favorably under certain situations.
- A significant international finance literature explores whether foreign investors are a ‘stabilizing’ or ‘destabilizing’ influence in emerging equity markets:
  - Stabilizing: foreign investors forecast prices better than domestic investors and enhance market efficiency.
  - Destabilizing: foreign investors trade in a manner that pushes prices away from fundamental value.
- Existing literature yields mixed results and focuses largely on average relationships rather than behavior in extreme events.

### Four policy-relevant questions for emerging market policy makers
- From the viewpoint of policy makers in emerging markets, four questions about the financial stability implications of foreign investment flows loom large:
  1. Do foreign investors exacerbate a domestic crisis by withdrawing capital on a large scale during the crisis?
  2. In this, is there asymmetric behavior, with different responses to very good versus very bad days in the local economy?
  3. Are foreign investors `big fish in a small pond’ – do their large transactions kick off substantial temporary mean-reverting distortions in the equity or currency market of an illiquid emerging market?
  4. When there is stress in the global financial system, do foreign investors withdraw capital on a large scale, and thus act as a vector of crisis transmission?
- These four questions are almost exclusively about behavior in extreme events (the tails), not average-day behavior.

### Limitations of linear/average approaches in the literature
- Many studies estimate linear relationships (e.g., VARs and VECMs) whose parameters reflect overall average relationships across all observations.
- There may be an ordinary regime (behavior on ordinary days) and different behavior in the tails.
- Averaging across all observations may underplay extreme behavior in the tails and give misleadingly reassuring answers to policy makers.
- Linear models estimated using overall data may mask nonlinearities in the tail response which are policy-relevant.

### Methodological contribution: tail-focused event study
- Contribution: adapt the event study (a workhorse of empirical finance) to directly address the four policy-relevant questions by focusing on extreme events.
- The methodology:
  - Focuses on extreme events and allows for the possibility that behavior under stressed market conditions may differ from day-to-day outcomes.
  - Measures relationships of interest under stressed conditions.
  - Avoids parametric assumptions by extending the non-parametric event study methodology.
- Example implementation:
  - Identify events consisting of extreme movements of the domestic stock market index.
  - Treat these dates as events and conduct an event study to measure how foreign investment behaves surrounding these dates.
  - Yields evidence about inter-linkages between foreign investment and stock market fluctuations in the tails without parametric assumptions about tail behavior.

### Empirical illustration and main findings (India)
- The paper illustrates the proposed methodology using data for one large emerging market, India.
- Main findings for India (described as relatively benign):
  - On very good days in the local economy, foreign investors seem to exacerbate the boom by bringing in additional capital.
  - There is asymmetric behavior: on very bad days in the local economy, no significant effects are found.
  - Foreign investors do not seem to be `big fish in a small pond’: extreme days of foreign investment in India do not kick off short-term price distortions with mean-reversion in following days, either on the currency market or on the equity market.
  - Very positive days on the S&P 500 trigger additional capital flowing into India, but there is no evidence of the reverse: international crises (with very poor days for the S&P 500) do not trigger exit by foreign investors.
  - Conclusion: foreign investors are not a vector of crisis transmission into the Indian equity market.
- Causality caveat noted:
  - One has to be careful about causality in interpreting results: even though there is temporal directionality consistent with so-called “Granger causality,” both variables are endogenous and subject to simultaneous exogenous effects.
  - A comment notes apparent inconsistency with large FII outflows and Indian stock market price fall in 2008-09; possible reasons include different aggregation and time scales and that the procedure averages across different extreme events rather than focusing on a single episode.

### Directions for future research and data availability
- Suggested extensions:
  - (a) Explore relationships observed in other countries, particularly relatively small countries and those less integrated into financial globalization.
  - (b) Examine other asset classes including debt capital, which have been an important source of concern in international financial crises.
  - (c) Investigate cross-sectional variation between multiple firms traded in the Indian equity market; foreign investors may not be `big fish in a small pond’ for the overall index but could affect illiquid securities or small-country indexes differently.
- The stance of foreign investors towards debt securities may differ considerably.
- The full source code developed for this paper has been released into the public domain to assist replication, downstream research, and methodological advances.

### Organization of the paper
- Section 2: summarize several strands of literature relevant for the analysis; discuss key studies analyzing the relationship between foreign institutional investment and stock market performance in India using linear parametric methods; discuss recent event studies in international trade and capital flows; relate the approach to literature on international information transmission.
- Section 3: overview of the data and the event study methodology.
- Section 4: presents results.
- Section 5: concludes.

*Source: _wp13122 - 1. Introduction*

### 3. Methodology and Data

### 3. Methodology and Data

### Data coverage and key series
- Dataset period: 15 February 2000 to 29 July 2011.
- Number of daily observations reported: 2,035 observations of daily data.
- Main daily series used:
  - Indian stock market returns (Nifty; expressed as log differences).
  - Net FII inflow (foreign institutional investor activity).
  - US S&P 500 returns.
  - Call money rate (domestic interest rate; weighted by volume of trades, expressed as a percentage per annum).
  - Nominal Rupee – US Dollar exchange rate (percent changes used).
- Summary statistics for all series are presented in Table 1 (in source).

### FII data: institutional context and construction
- Institutional constraints and practical convertibility:
  - Foreign investment can only be undertaken by “foreign institutional investors” (FIIs) required to register with the securities regulator.
  - Ownership by all foreign investors in a given firm cannot exceed 24% by default, but shareholder resolution can raise it up to 98%.
  - Of the over 5000 listed companies, at any point in time, there are no more than 20 firms where FIIs lack headroom for additional purchases; hence for almost all firms the 24 per cent limit has not been reached or the shareholder resolution has raised the limit.
  - There are no restrictions on selling or repatriating capital.
- Data source and processing:
  - FIIs are required to settle trades through custodian banks; custodian banks supply aggregate data to the government — this is the source of the daily FII time-series.
  - Raw net FII time-series is non-stationary (reflecting growth of India’s equity market and dollar value of foreign investment).
  - To correct non-stationarity, net FII is divided by the market capitalisation of the CMIE Cospi index, producing the scaled series FII/MktCap (vertical axis units multiplied by 10,000 in Figure 2).
  - Standard tests confirm the scaled series (FII/MktCap) is stationary.
- Note: Chakrabarti (2006) points to a structural break in net FII around April 2003; rescaling addresses this issue.

### Stock market data: timing and alignment
- Indian index: Nifty (dominant for index derivatives and index funds).
- US index: S&P 500.
- Trading hours do not overlap; causal relationship of interest runs from the US market to the Indian market.
- Alignment choice: previous calendar day of US data is lined up with the Indian data.

### Interest rate and exchange rate variables
- Interest rate: call money rate (weighted average of reported trades as calculated by the Reserve Bank of India), expressed as a percentage per annum. First differences are used to address non-stationarity.
- Exchange rate: nominal Rupee – US Dollar rate. Percent changes are used to avoid non-stationarity (negative changes denote nominal appreciation of the Rupee versus the US Dollar).
- Policymaker interest: response of interest rates and exchange rates to sharp movements in FII flows is examined.

### Event definition and identification
- Event-centric approach:
  - Events are defined as dates on which extreme values (tail observations) of returns or flows are observed.
  - Example: scanning S&P 500 one-day returns and identifying dates with returns in distribution tails.
- Tail probability choice:
  - Tradeoff: identifying truly extreme events (favoring deeper tails) versus having adequate sample size and statistical precision.
  - Dataset size described alternately as “large dataset of 2,036 observations” (text uses 2,036 in discussing tail reach), which permits reaching into the tails while retaining power.
- Tail cutoffs used in analysis:
  - Main event definition uses 5% tails (upper and lower 5%).
  - Comparison and notes mention 2.5% tails (more extreme) and differences in time patterns when using 2.5% tails.
- Time distribution of extreme events:
  - Table 2 summarizes time pattern; examples:
    - Of 67 upper-tail S&P 500 events, 8 occurred in 2000; median return of those 8 was +3.19%.
    - A large proportion of extreme values cluster in 2008 and 2009, especially for return variables.
    - FII tail values are more evenly distributed across years (e.g., 2005: 4 Nifty return tails vs 16 FII flow tails).
  - Few extreme rupee events observed prior to 2007 due to a structural break in the exchange rate regime on 23 March 2007 (Zeileis, Shah and Patnaik 2010).

### Event window, clustering, and uncontaminated events
- Event window choice:
  - Baseline pre-event and post-event windows: five market days each (a calendar week each side).
  - Robustness check: 20-day window (10 days on each side) — results not qualitatively modified.
- Clustering of tail values:
  - Tail events may cluster (consecutive extreme days). Two analysis paths are used:
    1. Primary analysis: fuse consecutive extreme events of the same direction into a single clustered event. In event time, date +1 is the first day after the run; date −1 is the last day prior to the start of the run. This preserves important crisis observations characterized by runs.
    2. Robustness check: analyze only uncontaminated events (no other event within the event window).
- Example statistics (right-hand top of Table 3):
  - For FII returns, lower 5% tail contains 102 extreme events; of these, 56 are uncontaminated and 41 are clustered.
- Summary count:
  - Table 3 summarizes the number of events (including clustered events) for Nifty returns, S&P 500 returns, normalized net FII flows, and three subsidiary variables.
  - Roughly 60% of the 99 or 100 extreme events in each tail are uncontaminated (no another tail event in the pre-event or post-event window).

### Formal definitions: uncontaminated events (methodology #1)
- Notation:
  - x_t, t = 1, ..., T denotes the Nifty returns time-series (log differences).
  - For quantile q, Q(x,q) such that Pr(x_t > Q(x,q)) = q.
  - Upper-tail dates: E+ = {i} s.t. x_i > Q(x,q).
  - Lower-tail dates: E−. All extreme events E = E+ ∪ E−.
- Uncontaminated upper-tail set E_u+:
  - E_u+ = {j in E+ such that (j+k) not in E for k in {-W,...,-1,1,...,W}}.
  - W is the event-window width in days prior to and after the event.

### Runs / clustered events (methodology #2)
- Runs definition:
  - An uncontaminated run of length R: {n, n+1, ..., n+R} ⊂ E+ with (j+k) not in E for k in {-W,...,-1,1,...,W}.
- Fusion of returns:
  - Returns on days of the run are fused into a single event return: sum_{t=n}^{n+R} x_t (displayed in source with summation notation).
  - Event window values become: x_{n-W}, ..., x_{n-1}, sum_{t=n}^{n+R} x_t, x_{n+R+1}, ..., x_{n+R+W}.
- Contaminated windows that satisfy the run properties are merged into the analysis alongside uncontaminated events for the second event study.

### Inference: bootstrap-based event-study procedure
- Motivation:
  - Classical inference in event studies can be unreliable due to distributional assumptions (normality) and serial correlation — especially relevant for FII flows.
  - Bootstrap avoids imposing normality and is robust to serial correlation.
- Bootstrap algorithm (as used):
  1. Suppose there are N events. Each event is represented as a time-series of cumulative returns (CR) or cumulative quantities (for FII) within the event window. The overall summary statistic is the average of all CR time-series (denote as the statistic of interest).
  2. Construct bootstrap samples by sampling with replacement at the event level: draw N events with replacement from the dataset of N events; for each drawn event, take its CR time-series to form one bootstrap realization of the statistic.
  3. Repeat step 2 one thousand times (1000) to obtain the bootstrap distribution of the statistic. Percentiles of this distribution are used to construct bootstrap confidence intervals shown in graphs.
- Bootstrap approach used is based on Davison, Hinkley, and Schectman (1986).

### Comparison with VAR impulse responses and Monte Carlo illustration
- Analogy: event-study tail response is analogous to a tail impulse response function (tail-VAR IRF).
- Monte Carlo experiment described:
  - Case I: two white-noise series with no tail relationship. VAR impulse response and tail event-study produce similar results; event-study 95% confidence interval is wider because it uses only a small number of tail events.
  - Case II: data-generating process contains a relationship only in the tails. VAR impulse response (an overall average relationship) fails to pick up the tail relationship (95% CI wider and null not rejected). Tail event-study picks up the tail response clearly.
- Sample size used in the simulation is stated as identical to the size of the dataset (text references the dataset size when describing Monte Carlo).
- The Monte Carlo details and source code are available from the authors on request.

### Relation to existing literature and methodological positioning
- Connections:
  - Related to Broner et al., 2010 in question focus (but Broner et al. use annual data).
  - Related to Lasfer et al., 2003 (event studies of indexes after extreme days).
  - Cumperayot et al., 2006 analyze linkages between extreme stock and currency days (not event-study framework).
  - Lasfer et al., 2012 study Chinese evidence and foreign investor effects on different market segments (A vs B markets).
- Distinction:
  - Use of high-frequency (daily) data enables efficient identification of causal effects not visible with low-frequency data.
  - Approach builds on event-study econometrics rather than regression frameworks for events.

### Research questions and how they are tested
- Question 1: Do foreign investors exacerbate a domestic crisis by withdrawing capital on a large scale?
  - Tested via event study measuring foreign investors’ behavior surrounding extreme domestic stock market index events.
- Question 2: Is there asymmetric behavior (different responses to very good vs very bad local days)?
  - Tested by separate event studies for very positive and very negative days for the local stock market index.
- Question 3: Are foreign investors “big fish in a small pond” — do their large transactions trigger substantial temporary mean-reverting distortions in an illiquid emerging market?
  - Tested by defining extreme events as days with very positive or very negative foreign capital inflows and observing outcomes for the domestic stock market index.
- Question 4: During stressed global financial conditions, do foreign investors withdraw capital on a large scale and thus act as a vector of crisis transmission?
  - Tested by event studies focusing on extreme S&P 500 days and examining outcomes for foreign capital inflows and the domestic stock market index.

*Source: 3. Methodology and Data (from the supplied IMF PDF chapter).*

### 4. Results

### 4. Results

### FII flows and extreme Nifty returns
- Method: For each type of 11-day window (the event and five days before and after), a confidence interval for cumulative values is constructed beginning from the first day of the 11-day window.
- Very positive Nifty days (left pane of Figure 4):
  - Prior to large positive Nifty returns, there is no unusual activity in FII investment.
  - On the event date, FII investment has a statistically significant positive value.
  - The location estimator shows +1 basis points (of the overall market capitalization) as the net purchase of FIIs on the event date, and the null hypothesis of 0 can be rejected.
  - Interpretation: both foreign investors and Nifty may be responding to positive news.
  - In the four trading days after the event date, the point estimator adds another 2 basis points (of the overall market capitalization).
  - Evidence of slow or positive feedback trading by FIIs in the days after the event date.
- Very bad Nifty days (right pane of Figure 4):
  - No evidence of foreign investors selling prior to the event date.
  - Unusually large negative Nifty returns are not due to prior FII selling for exogenous reasons.
  - After the event date, there is no evidence of positive feedback trading; slight evidence of foreign investment being positive in the event window.
- Overall interpretation:
  - There is not a simple relationship between Nifty returns and FII flows; patterns differ by sign of movements.
  - From a policymaker perspective, there is no evidence in this data that FII outflows in response to very bad days trigger a crisis over the sample period.
  - Concern about positive feedback: the point estimate is 1.4 times the standard deviation of the daily time-series of FII investment, with a 95% confidence band from 0.26 to 2.56 standard deviations.

### Nifty returns responding to extreme FII flows (Figure 5)
- Event-study setup: y-axis pertains to cumulative returns; efficient markets response is a step response on the event date with a flat profile thereafter. Rejections could indicate under-reaction or over-reaction.
- Positive FII-flow extremes:
  - Nifty returns rise before very good days for FII flows, likely reflecting joint responses to good news.
  - Large FII inflows are associated with an extremely positive day for Nifty.
  - After the event date, the null of a flat profile cannot be rejected — consistent with an efficient response (no evidence of overshooting or undershooting).
- Very bad FII-flow days (large outflows):
  - Not preceded by large drops in Indian stock market returns.
  - Do not trigger further negative returns; cumulative graph is relatively flat after the sharp outflow.
  - On the day of sharp outflows, there is a drop in Nifty of one standard deviation of the daily series, but it is not exacerbated in subsequent days.
- Interpretation: sharp declines in FII flows do not appear to trigger large and persistent domestic stock market declines.

### FII flows responding to global shocks: S&P 500 extremes (Figures 6 and 7)
- Rationale: S&P 500 returns are used to capture global shocks and aggregate global information; the US market is unlikely to be affected by the Indian market or by FII flows in and out of India.
- FII response to very good S&P 500 days (Figure 6):
  - Very good days on the S&P 500 are associated with strong FII inflows into India worth at least twice its standard deviation, both before and after the event.
  - The response is much stronger than that observed for domestic (Nifty) positive events.
  - Interpretation: FIIs are a vector of transmission of good news on the S&P 500 into India.
- FII response to very bad S&P 500 days:
  - No marked impact on FII flows; pattern similar to the response to very bad Nifty days.
  - Asymmetry: FIIs communicate extremely good S&P 500 days into India via large purchases, but do not behave symmetrically for large negative S&P 500 days.
- Nifty response to S&P 500 extremes (Figure 7):
  - Good news for the S&P 500 is likely good news for India and vice versa.
  - Impact on Nifty on event date in both directions, with an approximately flat trajectory thereafter.

### Knock-on impacts on interest rates and exchange rates (Figures 8 and 9)
- Interest rates (call money rate, Figure 8):
  - Exceptionally large FII inflows have imperceptible effects on the call money rate.
  - FII outflows appear to be followed by some rise in interest rates, but the confidence interval grows very wide; the effect is therefore uncertain.
  - Endogenous policy responses (e.g., central bank intervention, sterilization) could be part of the observed phenomenon; effects vary with exchange rate regime and extent of sterilization.
- Exchange rate (Figure 9):
  - Large FII inflows are associated with appreciation of the Rupee; large outflows are associated with depreciation.
  - For outflows (“very bad” days for FII flows), the depreciation appears to precede the extreme event.
  - The impact is 1.4 and 2.85 times the standard deviation for appreciation and depreciation respectively.
  - Interpretation: capital flows affect the exchange rate, but results do not support the view that large inflows or outflows trigger a panic with respect to the currency.

### Event construction and uncontaminated events
- Main results fuse a cluster of extreme events (of the same sign) into one; date +1 is always the first date after the last extreme event.
- Results for uncontaminated events (single-day extreme events with no other extreme day in the event window) are also present in the figures.

### Robustness checks (Section 5)
- Robustness: global shocks measured by S&P 500 returns in main results; alternative using VIX (implied volatility index).
- Figure 10 (VIX): patterns are qualitatively similar to those seen in Figure 6 for the S&P 500.
  - Extreme events with a decline of VIX are associated with increases in foreign investment into India.
  - The reverse (response to an increase in VIX) does not have a statistically significant response.

### Conclusion (selected results and implications)
- The tail-focused event study with bootstrap inference imposes no functional form on the response profile and focuses on relationships in the tails rather than average relationships.
- Application to equity investment by foreign investors in India yields relatively benign results: average tail or extreme events do not support concerns that foreign investors have produced destabilizing panic behavior leading to crises in the sample period.
- Caveats and further research directions:
  - Refinements of the event study econometrics.
  - Application across multiple countries, especially smaller and less liquid markets.
  - Application across firms to uncover heterogeneity of foreign investor impacts.
  - Application to short-term debt flows (local-currency and foreign-currency) where risks may differ and could reveal a hierarchy of capital flows.

*Source: _wp13122 - 4. Results*

### References

### _wp13122 - References

### Major components
- References list containing cited works on systemic risk, event-study methodology, foreign institutional investment, asset market linkages, bootstrap methods, and related empirical finance literature.
- Tables and Figures sections summarizing empirical statistics, distributions of extreme events, runs distribution, quantile values for 5% tails, and yearly distribution of extreme values (5% tails).
- Figures include: Net FII Flows; Normalized Net FII Flows; Monte Carlo experiment (Case (I), Case (II)); Extreme event on Nifty and response of FII; Extreme event on FII and response of Nifty; Extreme event on S&P 500 and response of FII; Extreme event on S&P 500 and response of Nifty; Extreme event on FII and response of call money rate; Extreme event on FII and response of INR; Extreme event on VIX and response of FII.

### Key statistics (Table 1: Summary Statistics)
- Observations (N): 2036 for all series listed.
- Daily net FII inflow (bps to COSPI) (selected statistics):
  - Minimum: -10.87
  - 5%: -2.24
  - 25%: -0.36
  - Median: 0.48
  - Mean: 0.59
  - 75%: 1.41
  - 95%: 3.55
  - Maximum: 30.16
  - Standard deviation: 2.12
  - IQR: 1.77
- Daily Nifty returns (per cent):
  - Minimum: -13.97
  - 5%: -3.28
  - 25%: -0.93
  - Median: 0.16
  - Mean: 0.06
  - 75%: 1.08
  - 95%: 2.95
  - Maximum: 16.23
  - Standard deviation: 1.98
  - IQR: 2.01
- Daily returns on Nikkei 225 (per cent):
  - Minimum: -17.53
  - 5%: -2.89
  - 25%: -0.96
  - Median: 0.04
  - Mean: -0.03
  - 75%: 0.96
  - 95%: 2.60
  - Maximum: 13.23
  - Standard deviation: 1.85
  - IQR: 1.92
- Daily returns on S&P 500 (per cent):
  - Minimum: -9.47
  - 5%: -2.44
  - 25%: -0.68
  - Median: 0.09
  - Mean: 0.00
  - 75%: 0.72
  - 95%: 2.25
  - Maximum: 12.40
  - Standard deviation: 1.57
  - IQR: 1.40
- Daily returns on INR/USD exchange rate (bps):
  - Minimum: -417.09
  - 5%: -62.76
  - 25%: -13.63
  - Median: 0.00
  - Mean: 0.07
  - 75%: 12.29
  - 95%: 69.10
  - Maximum: 287.02
  - Standard deviation: 43.56
  - IQR: 25.92
- Daily change in the call money rate (bps):
  - Minimum: -18040.30
  - 5%: -1735.37
  - 25%: -170.10
  - Median: 0.00
  - Mean: -1.04
  - 75%: 140.81
  - 95%: 1671.22
  - Maximum: 27677.69
  - Standard deviation: 1899.29
  - IQR: 310.91

### Distribution of events at 5% (Table 2)
- Lower tail (counts, totals shown as in table):
  - Nifty: Un-clustered 57; Clustered Used 6; Clustered Not-used 39; Total-used 45; Total 63; Total (column) 102
  - S&P 500: Un-clustered 56; Clustered Used 6; Clustered Not-used 40; Total-used 46; Total 62; Total (column) 102
  - FII: Un-clustered 56; Clustered Used 41; Clustered Not-used 5; Total-used 46; Total 97; Total (column) 102
  - INR: Un-clustered 58; Clustered Used 16; Clustered Not-used 28; Total-used 44; Total 74; Total (column) 102
  - Call rate: Un-clustered 29; Clustered Used 3; Clustered Not-used 70; Total-used 73; Total 32; Total (column) 102
- Upper tail:
  - Nifty: Un-clustered 61; Clustered Used 12; Clustered Not-used 29; Total-used 41; Total 73; Total (column) 102
  - S&P 500: Un-clustered 65; Clustered Used 5; Clustered Not-used 32; Total-used 37; Total 70; Total (column) 102
  - FII: Un-clustered 65; Clustered Used 33; Clustered Not-used 4; Total-used 37; Total 98; Total (column) 102
  - INR: Un-clustered 53; Clustered Used 15; Clustered Not-used 34; Total-used 49; Total 68; Total (column) 102
  - Call rate: Un-clustered 33; Clustered Used 6; Clustered Not-used 63; Total-used 69; Total 39; Total (column) 102

### Runs distribution (Table 3)
- Lower tail run lengths (Two, Three, Four, Five) by series:
  - Nifty: 3, 0, 0, 0
  - S&P 500: 3, 0, 0, 0
  - FII: 7, 4, 0, 3
  - INR: 8, 0, 0, 0
  - Call rate: 0, 1, 0, 0
- Upper tail run lengths (Two, Three, Four, Five) by series:
  - Nifty: 4, 0, 1, 0
  - S&P 500: 1, 1, 0, 0
  - FII: 12, 3, 0, 0
  - INR: 6, 1, 0, 0
  - Call rate: 3, 0, 0, 0

### Quantile values (per cent) for tail at 5% (Table 4)
- Lower tail quantiles (Min., 25%, Median, Mean, 75%, Max.):
  - Nifty: Min. -13.97; 25% -5.24; Median -4.21; Mean -3.67; 75% -3.84; Max. -3.43
  - S&P 500: Min. -7.16; 25% -4.09; Median -3.05; Mean -4.85; 75% -2.71; Max. -2.44
  - FII: Min. -0.11; 25% -0.04; Median -3.21; Mean -3.76; 75% -0.03; Max. -0.02
  - INR: Min. -4.17; 25% -1.16; Median -0.90; Mean -1.17; 75% -0.75; Max. -0.63
  - Call rate: Min. -93.65; 25% -27.36; Median -22.15; Mean -27.22; 75% -18.70; Max. -17.45
- Upper tail quantiles (Min., 25%, Median, Mean, 75%, Max.):
  - Nifty: Min. 2.96; 25% 3.26; Median 3.98; Mean 3.32; 75% 4.96; Max. 17.30
  - S&P 500: Min. 2.27; 25% 2.45; Median 2.90; Mean 4.75; 75% 3.52; Max. 9.91
  - FII: Min. 0.04; 25% 0.04; Median 4.81; Mean 5.69; 75% 0.06; Max. 0.17
  - INR: Min. 0.69; 25% 0.80; Median 0.96; Mean 1.15; 75% 1.16; Max. 3.40
  - Call rate: Min. 16.76; 25% 18.03; Median 22.91; Mean 162.59; 75% 26.46; Max. 4607.69

### Yearly distribution of extreme values (5% tails) (Table 5) — totals and notable yearly counts
- Totals across 2000–2011:
  - S&P 500: 5% Good days total 67; 5% Bad days total 59
  - Nifty: 5% Good days total 66; 5% Bad days total 60
  - FII (bps change): 5% Good days total 80; 5% Bad days total 70
  - INR (% change): 5% Good days total 60; 5% Bad days total 66
  - Call rate (bps change): 5% Good days total 36; 5% Bad days total 30
- Selected yearly examples (number of events and medians as in table):
  - Year 2000:
    - S&P 500: No. 8, Median 3.19 (5% Good days); No. 7, Median -2.63 (5% Bad days)
    - Nifty: No. 3, Median 4.04 (5% Good days); No. 9, Median -5.02 (5% Bad days)
    - FII: No. 10, Median 4.52 (5% Good days); No. 12, Median -3.37 (5% Bad days)
    - INR: No. 4, Median 0.86 (5% Good days); No. 1, Median -0.80 (5% Bad days)
    - Call rate: No. 5, Median 49.12 (5% Good days); No. 5, Median -18.66 (5% Bad days)
  - Year 2008:
    - S&P 500: No. 8, Median 3.17 (5% Good days); No. 11, Median -2.98 (5% Bad days)
    - Nifty: No. 15, Median 4.83 (5% Good days); No. 16, Median -4.69 (5% Bad days)
    - FII: No. 0 (5% Good days); No. 14, Median -3.47 (5% Bad days)
    - INR: No. 12, Median 1.53 (5% Good days); No. 9, Median -1.22 (5% Bad days)
    - Call rate: No. 11, Median 23.23 (5% Good days); No. 5, Median -27.44 (5% Bad days)
  - Year 2011:
    - S&P 500: No. 1, Median 2.30 (5% Good days); No. 1, Median -2.52 (5% Bad days)
    - Nifty: No. 2, Median 4.40 (5% Good days); No. 1, Median -4.13 (5% Bad days)
    - FII: No. 0 (5% Good days); No. 3, Median -3.70 (5% Bad days)
    - INR: No. 7, Median 0.87 (5% Good days); No. 3, Median -0.75 (5% Bad days)
    - Call rate: No. 1, Median 22.57 (5% Good days); No. 0 (5% Bad days)

*Content unit: _wp13122 - References*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2013/_wp13122.pdf_