## 6. ICAPM Estimation Results

## Source details

**Canonical URL:** [6. ICAPM Estimation Results](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16177.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16177.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16177.pdf.json)

---

### Main objectives and empirical approach
- Investigate whether low-income frontier markets (FMs) resemble emerging markets (EMs) in exposure to private capital flows and volatility versus the rest of low-income developing countries (LIDCs).
- Two-pronged empirical strategy:
  - Difference-in-differences (DiD) on an annual panel of 76 countries, 2000–14, to assess net portfolio investment to FMs relative to EMs (lower-quartile EMs, LQ-EMs) and non-FM LIDCs (NFM-LIDCs).
  - International Capital Asset Pricing Model (ICAPM) analysis using monthly sovereign bond returns for FMs over 2000–14 to test for changes in linkages with international capital markets before and after 2008.

### Key hypotheses
- Hypothesis 1 (H1): Catching-up on capital flows — in a low interest rate environment and similar macro risks, increases in private capital flows to FMs will be comparable to or exceed those to EMs (search for yield → Δ[CF/Y]FM ≥ Δ[CF/Y]LQ-EM).
- Hypothesis 2 (H2): Increased sensitivity of FMs to global economic and financial developments — if H1 holds, FMs become more exposed to external shocks and investor sentiment, implying sharper decreases in flows in turmoil.
- Hypothesis 3 (H3): Greater integration of FMs into global financial markets — conditional on H1, expect increased co-movement between FMs’ bond returns and global bond returns; in ICAPM, FMs’ market beta increases and becomes significant after 2008.

### Stylized facts (2000–14)
- Large private capital flows to LIDCs over 2000–14, especially FDI; cumulative net FDI increased more than six fold during the period.
- Cumulative net portfolio flows:
  - Surge in gross non-FDI private flows (as percent of GDP) to LIDCs by 2015 comparable to EMs; most of these flows went to FMs.
  - Increase in flows to FMs was far below that to EMs in absolute US$ terms, but significantly larger as a share of recipient countries’ GDP.
- Price dispersion: FM assets were priced lower than EM assets prior to repricing, consistent with segmentation or higher risk premia; search for yield and lower risk aversion likely contributed to increased integration and flows to FMs after 2008.

### Data, country groupings, and variables
- Annual panel: 76 countries, 2000–14; net portfolio investment (NPI) series from IMF WEO, scaled by GDP (NPI/Y).
- FM definition: IMF taxonomy identifies 14 LIDCs as FMs; time-varying composition used in robustness checks; alternative FM grouping combining IMF and market analysts’ classifications used in Appendix C.
- Financial integration data: JP Morgan NEXGEM index for FM country bond indices (NEXGEM includes 18 countries); global benchmark JP Morgan Global Aggregate Bond Index (GABI) for global bond returns.
- Push variables: U.S. 10-Year Treasury Bond Yield; VIX (FRED, Bloomberg).
- Pull variables: country-level controls from WEO; Financial Risk rating from ICRG used in robustness checks (available for ~50% of sample).

### Selected summary statistics (preserved exactly)
- Net Portfolio Investment (% GDP) — group statistics before/after 2008:
  - LIDCs
    - 2000-08: Mean 0.6931 Std.Dev 1.129 Min 0.124 Max 0.329
    - 2008-14: Mean 0.354 Std.Dev 0.207 Min 0.105 Max 0.67
  - FMs
    - 2000-08: Mean 0.135 Std.Dev 0.446 Min -0.444 Max 1.277
    - 2008-14: Mean 1.007 Std.Dev 0.807 Min -0.108 Max 2.397
  - NFM-LIDCs
    - 2000-08: Mean 0.046 Std.Dev 0.719 Min -0.054 Max 0.218
    - 2008-14: Mean 0.13 Std.Dev 0.124 Min -0.063 Max 0.335
  - EMs (Control group)
    - 2000-08: Mean 0.077 Std.Dev 0.331 Min -0.507 Max 0.686
    - 2008-14: Mean 0.386 Std.Dev 0.317 Min 0.005 Max 0.843

- Selected bivariate correlations (with p-values in parentheses; * denotes significance in source):
  - Portfolio Investment/GDP vs VIX: -0.0672* (0.0302)
  - Portfolio Investment/GDP vs U.S. 10-Y Treasury Bond Yield: -0.1303* (0.0000)
  - Current-Account/GDP vs Portfolio Investment/GDP: -0.1189* (0.0001)
  - Exports/GDP vs Portfolio Investment/GDP: 0.0463 (0.1357)

### DiD empirical strategy and identification
- Baseline DiD regression (dependent variable NPI/Y):
  - NPIi,t/Yi,t = β0 LIDCi,t + β1 Crisist + β2 FMi,t + β3 FMi,t * Crisist + β4 NFMi,t * Crisist + γ*Xi,t + υi + ξt + ui,t
  - Crisis dummy: =1 if observation is after 2008, 0 otherwise.
  - Key coefficient: β3 (interaction FM × Crisis) tests H1.
  - Country and year fixed effects included; control variables include global push and country pull factors.
- Additional specifications:
  - Triple interaction (FM × Crisis × VIX) to test H2 (sensitivity to global risk aversion).
  - Year-by-FM interactions (FM × year dummies) to inspect year-specific deviations (focus on post-crisis years and episodes of heightened stress: 2008–09, 2013).
  - Robustness: time-varying FM group; inclusion of ICRG governance indicator (reduces sample by ~50%).

### DiD estimation results — portfolio flows (selected findings preserved)
- Aggregate/average effects (2000–14):
  - Portfolio flows to FMs exceeded those to LQ-EMs by 1.4 percentage points of GDP (unconditional baseline result).
  - When controlling for pull factors, FMs’ portfolio flows exceed EMs’ by 0.62 percent of GDP.
  - Using time-varying FM composition, FMs received more than LQ-EMs by roughly 0.85 percentage point (robustness discussion).
- Non-FM LIDCs (NFM-LIDCs) did not exhibit substantial or significant changes relative to LQ-EMs.
- Year-specific dynamics:
  - Net portfolio flows to FMs increased post-GFC overall but dried out in years of heightened global risk aversion — notably 2008–09 and 2013 (the taper tantrum).
  - Interactions for post-crisis years generally positive and significant except for years of heightened stress (2008, 2009, 2013), supporting H2.
- Triple interaction (FM × Crisis × VIX):
  - The triple interaction coefficient is not statistically significant, indicating that after the crisis FMs’ exposure to changes in investor sentiment has become comparable to that of the EM control group (LQ-EMs).
- Robustness (selected preserved coefficients and test statistics):
  - Time-varying FM estimates reported in tables: baseline 1.417**, reduced to 0.623** when controls added; alternative time-varying estimates around 0.876***, 0.845***, 0.844*** in columns 6–8.
  - Inclusion of governance proxy (ICRG) on smaller sample shows Crisis_t * Frontier_t coefficients: 0.674* (s.e. 0.374), 0.649* (s.e. 0.374), 0.632** (s.e. 0.279), 0.630** (s.e. 0.281), 0.647** (s.e. 0.273), 0.638** (s.e. 0.276) across columns (1)–(6).
  - Selected year-by-FM interaction coefficients (Appendix G, Table 1): (year==2007)*FM_i,t = 1.243** (s.e. 0.530); (year==2010)*FM_i,t = 1.169** (s.e. 0.530); (year==2012)*FM_i,t = 2.045*** (s.e. 0.531).

### ICAPM framework and empirical implementation
- Hypothesis H3: larger capital flows after 2008 increased FMs’ financial integration with global markets; expect global market beta (β) to increase and become significant after 2008.
- Empirical approach:
  - Single-factor ICAPM: R_i,t − R_t^f = α_i + β*(R_t^Mkt − R_t^f) + γ_i + ε_i,t.
  - Two-factor ICAPM (partial-integration): R_i,t − R_t^f = α_i + β1*(R_t^Mkt − R_t^f) + β2*(R_t^local − R_t^f) + γ_i + ε_i,t.
  - Global market proxy: Barclays Global Bond Index (R_Mkt); risk-free rate proxied by 3-month US T-bill rate (R_f).
  - Local index proxies: for FMs R_nexgem; for EMs R_embi.
  - Sample: 2000–14, monthly frequency; FMs (12) and EMs (10) in ICAPM sample.

### ICAPM estimation results (preserved numeric highlights)
- Single-factor ICAPM — Frontier Markets (Panel A; monthly observations 824 before, 839 after)
  - BEFORE beta estimates: 0.454 (0.312); BEFORE with FE 0.499 (0.322)
  - AFTER beta estimates: 1.716*** (0.391); AFTER with FE 1.716*** (0.402)
  - R-squared: single-factor BEFORE 0.008; AFTER 0.081; with FE BEFORE 0.020; AFTER 0.088
- Two-factor ICAPM — Frontier Markets (columns 5–6)
  - BEFORE global beta 0.179 (0.282); AFTER global beta 1.121*** (0.323)
  - R_nexgem − R_f coefficients: AFTER 0.473*** (0.127); BEFORE 0.339** (0.159)
  - R-squared: two-factor BEFORE 0.140; AFTER 0.315
- Single-factor ICAPM — Emerging Markets (Panel B; monthly observations 702 before, 730 after)
  - BEFORE beta estimates: 1.373*** (0.332); AFTER 1.669*** (0.439)
  - WITH FE: BEFORE 1.312*** (0.331); AFTER 1.656*** (0.450)
  - R-squared: single-factor BEFORE 0.079; AFTER 0.059; with FE BEFORE 0.135; AFTER 0.072
- Two-factor ICAPM — Emerging Markets (columns 5–6)
  - BEFORE β1 0.541** (0.219); AFTER β1 0.713** (0.327)
  - R_embi − R_f coefficients: AFTER 0.744*** (0.0509); BEFORE 0.681*** (0.0413)
  - R-squared: two-factor BEFORE 0.641; AFTER 0.531
- Interpretation of magnitudes:
  - For FMs, post-2008 single-factor market beta ~ 1.7: a 1 percent change in international bond returns translates into a 1.7 percent change in FM bond returns after 2008.
  - Two-factor results show FMs moved from pricing mainly local factors before 2008 to pricing both local and global factors after 2008 (partial-integration).

### Interpretation: integration, vulnerabilities, and convergence
- Evidence supports H3: FMs moved from segmentation before 2008 toward greater integration with global bond markets after 2008, as indicated by positive and significant global betas.
- Combined with DiD results, FMs have become more similar to EMs than to other LIDCs in exposure to volatile private capital flows and global market shocks.
- Economic significance: market betas around 1.7 post-2008 imply FMs can experience larger return and price responses to global bond market shocks than advanced markets.

### Principal conclusions (preserved quantitative summaries)
- Three main empirical results:
  - After the GFC, average annual net portfolio investment to FMs outstripped those to EMs by about 0.6 percentage point of GDP in one specification and by larger magnitudes in other specifications (1.4 percentage points in baseline; 0.62 percent of GDP when full pull factors included; roughly 0.85 percentage point in time-varying FM robustness).
  - Net portfolio flows to FMs increased post-GFC overall but fell sharply in years of heightened global risk aversion (2008–09, 2013).
  - ICAPM evidence shows increased market betas for FMs after 2008, consistent with greater financial integration and comovement with global markets.
- Overall interpretation: FMs have become more similar to EMs (and less like other LIDCs) in their exposure to volatile private capital flows and to global financial market shocks.

### Policy implications
- Greater vulnerability to adverse global financial developments points to the importance of frameworks for FMs to:
  - manage vulnerabilities to capital flow reversals, and
  - cope with reversals that could jeopardize macroeconomic performance.
- Risk-adjusted flows: when level of risk aversion is taken into account, portfolio flows to FMs are not statistically different from those to EMs, implying similar vulnerability during market stress (examples cited: 2008, 2009, 2013).

### Caveats and suggested areas for further research
- Findings are based on changes during a relatively short period of unusually lax monetary conditions in major advanced economies and may not reflect a permanent shift.
- FM identification may entail sample selection bias; regressions control for other determinants of portfolio flows to mitigate bias.
- Bond return data limitations: not all FMs could be included in ICAPM regressions; a larger sample would improve robustness.
- The study does not quantify implications for FMs’ financing needs or prescribe specific resilience frameworks—areas for further research.

*Source: Excerpt from the chapter "6. ICAPM Estimation Results" in the provided IMF working paper (2000–14 sample; tables and appendices as cited in the source).*

### REFERENCES ____________________________________________________________25

### _wp16177 - REFERENCES ____________________________________________________________25

### FIGURES

- 1. Capital Flows to LIDCs, 2000-2014 ___________________________________________6
- 2. Portfolio Flows to Developing Countries Sub-groups, 2000-2014 ___________________7
- 3. Annual Portfolio Flows to Developing Countries Subgroups _______________________8
- 4. EM and FM Bond Index Levels ______________________________________________9

### TABLES

- 1. Summary Statistics _______________________________________________________11
- 2. Correlation Matrix _______________________________________________________12
- 3. Illustration of DiD estimation _______________________________________________15
- 4. Hypothesis 1 Results: Catching up on Capital Flows _____________________________18
- 5. FM Sensitivity to Global Risk Aversion _______________________________________19

*Source: _wp16177 - REFERENCES ____________________________________________________________25*

### 6. ICAPM Estimation Results _________________________________________________22

### 6. ICAPM Estimation Results

### Main objectives and approach
- Investigate whether low-income frontier markets (FMs) resemble emerging markets (EMs) in exposure to private capital flows and volatility, versus the rest of low-income developing countries (LIDCs).
- Two-pronged empirical strategy:
  - Difference-in-differences (DiD) estimation on annual panel data covering a panel of 76 countries during 2000–14 to assess the magnitude of net portfolio investment to FMs relative to other subgroups (EMs and non-FM LIDCs).
  - International Capital Asset Pricing Model (ICAPM) analysis using monthly sovereign bond returns for FMs over 2000–14 to test for changes in linkages between FMs and international capital markets before and after the Global Financial Crisis (GFC).

### Key hypotheses
- Hypothesis 1 (H1): Catching-up on capital flows — in a low interest rate environment and similar macro risks, increases in private capital flows to FMs will be comparable to or exceed those to EMs (search for yield → Δ[CF/Y]FM ≥ Δ[CF/Y]LQ-EM).
- Hypothesis 2 (H2): Increased sensitivity of FMs to global economic and financial developments — if H1 holds, FMs become more exposed to external shocks and investor sentiment, implying sharper decreases in flows in turmoil.
- Hypothesis 3 (H3): Greater integration of FMs into global financial markets — conditional on H1, expect increased co-movement between FMs’ bond returns and global bond returns; in ICAPM, FMs’ market beta increases and becomes significant after 2008.

### Stylized facts (2000–14)
- Over 2000–14 LIDCs received large private capital flows, especially FDI; cumulative net FDI increased more than six fold during the period (reflecting the commodities supercycle).
- Cumulative net portfolio flows:
  - Surge in gross non-FDI private flows (as percent of GDP) to LIDCs by 2015 comparable to EMs (Araujo et al. 2015); most of these flows went to FMs.
  - Increase in flows to FMs was far below that to EMs in absolute US$ terms, but significantly larger as a share of recipient countries’ GDP.
- Price dispersion: FM assets were priced lower than EM assets prior to repricing, suggesting either segmentation or higher risk premia; search for yield and lower risk aversion likely contributed to increased integration and flows to FMs after 2008.

### Data and country groupings
- Source and sample:
  - Annual panel of 76 countries, 2000–14, including all LIDCs and lower-quartile EMs (LQ-EMs) as control group.
  - Net portfolio investment series from IMF WEO, scaled by GDP (reporting net flows).
- FM definition:
  - IMF taxonomy (IMF, 2014 policy paper) identifies 14 LIDCs as FMs; time-varying composition used in robustness checks.
  - Alternative FM grouping exercise combining IMF and market analysts’ classifications confirms robustness (Appendix C).
- Financial integration data:
  - JP Morgan NEXGEM index used for FM country bond indices (NEXGEM includes 18 countries); global benchmark JP Morgan Global Aggregate Bond Index (GABI) used for global bond returns.
- Push and pull variables:
  - Push: U.S. 10-Year Treasury Bond Yield, VIX (FRED, Bloomberg).
  - Pull: country-level controls from WEO; Financial Risk rating from ICRG used in robustness checks (available for ~50% of sample).

### Summary statistics (selected, preserved exactly as reported)
- Net Portfolio Investment (% GDP) — group statistics before/after 2008 as presented:
  - LIDCs
    - 2000-08: Mean 0.6931 Std.Dev 1.129 Min 0.124 Max 0.329
    - 2008-14: Mean 0.354 Std.Dev 0.207 Min 0.105 Max 0.67
  - FMs
    - 2000-08: Mean 0.135 Std.Dev 0.446 Min -0.444 Max 1.277
    - 2008-14: Mean 1.007 Std.Dev 0.807 Min -0.108 Max 2.397
  - NFM-LIDCs
    - 2000-08: Mean 0.046 Std.Dev 0.719 Min -0.054 Max 0.218
    - 2008-14: Mean 0.13 Std.Dev 0.124 Min -0.063 Max 0.335
  - EMs (Control group)
    - 2000-08: Mean 0.077 Std.Dev 0.331 Min -0.507 Max 0.686
    - 2008-14: Mean 0.386 Std.Dev 0.317 Min 0.005 Max 0.843

- Bivariate correlations (selected pairs; p-values in parentheses; * denotes significance as in source):
  - Portfolio Investment/GDP vs VIX: -0.0672* (0.0302)
  - Portfolio Investment/GDP vs U.S. 10-Y Treasury Bond Yield: -0.1303* (0.0000)
  - Current-Account/GDP vs Portfolio Investment/GDP: -0.1189* (0.0001)
  - Exports/GDP vs Portfolio Investment/GDP: 0.0463 (0.1357)

### Empirical strategy details (DiD)
- Baseline DiD regression (dependent variable NPI/GDP):
  - NPIi,t/Yi,t = β0 LIDCi,t + β1 Crisist + β2 FMi,t + β3 FMi,t * Crisist + β4 NFMi,t * Crisist + γ*Xi,t + υi + ξt + ui,t
  - Crisis dummy: =1 if observation is after 2008, 0 otherwise.
  - Key coefficient: β3 (interaction FM × Crisis) tests H1.
  - Country and year fixed effects included; control variables include global push and country pull factors.
- Additional DiD specifications:
  - Triple interaction (FM × Crisis × VIX) to test H2 (sensitivity to global risk aversion).
  - Year-by-FM interactions (FM × year dummies) to inspect year-specific deviations, focusing on post-crisis years and episodes of heightened stress (2008–09, 2013).
  - Robustness: time-varying FM group based on qualification dates; inclusion of ICRG governance indicator (reduces sample by ~50%).

### DiD estimation results — portfolio flows (selected preserved findings)
- Aggregate/average effects (2000–14):
  - Portfolio flows to FMs exceeded those to LQ-EMs by 1.4 percentage points of GDP (unconditional baseline result).
  - When controlling for pull factors, FMs’ portfolio flows exceed EMs’ by 0.62 percent of GDP (preserved exactly as reported).
  - Using time-varying FM composition, FMs received more than LQ-EMs by roughly 0.85 percentage point (reported in robustness discussion).
- Non-FM LIDCs (NFM-LIDCs) did not exhibit substantial or significant changes in portfolio flows relative to LQ-EMs.
- Year-specific dynamics:
  - Net portfolio flows to FMs increased post-GFC overall but dried out in years of heightened global risk aversion — notably 2008–09 and 2013 (the taper tantrum).
  - Interactions for post-crisis years are generally positive and significant except for years of heightened stress (2008, 2009, 2013), supporting H2.
- Triple interaction test (FM × Crisis × VIX):
  - The triple interaction coefficient is not statistically significant, indicating that after the crisis FMs’ exposure to changes in investor sentiment has become comparable to that of the EM control group (LQ-EMs).
- Robustness checks:
  - Results hold when FM group is time-varying (columns reported: baseline 1.417**, reduced to 0.623** when controls added; alternative time-varying estimates around 0.876***, 0.845***, 0.844*** in columns 6–8 — numbers preserved from reported table).
  - Inclusion of governance proxy (ICRG) on smaller sample still shows increases in portfolio flows to FMs (magnitude preserved in text: about 0.85 percentage point in time-varying case).

### ICAPM / financial integration results (summary as presented)
- Asset-pricing analysis finds noteworthy change in FMs’ market betas with respect to global market returns after 2008.
  - Finding interpreted as an increase in financial integration and comovement of returns post-2008.
- Combined implications:
  - FMs have become more similar to EMs than to other LIDCs in terms of vulnerability to adverse global financial market conditions.
  - Increased co-movement and higher market beta imply greater spillover risk for FMs from global market shocks.

### Principal conclusions
- Three main empirical results documented:
  - After the GFC, average annual net portfolio investment to FMs outstripped those to EMs by about 0.6 percentage point of GDP in one specification and by larger magnitudes in other specifications (1.4 percentage points in baseline; 0.62 percent of GDP when full pull factors included; roughly 0.85 percentage point in time-varying FM robustness).
  - Net portfolio flows to FMs increased post-GFC overall but fell sharply in years of heightened global risk aversion (2008–09, 2013).
  - ICAPM evidence shows increased market betas for FMs after 2008, consistent with greater financial integration and comovement with global markets.
- Overall interpretation: FMs have become more similar to EMs (and less like other LIDCs) in their exposure to volatile private capital flows and to global financial market shocks.

*Source: Excerpt from the chapter "6. ICAPM Estimation Results" in the provided IMF working paper (2000–14 sample; tables and appendices as cited in the source).*

### 0.6 percent of GDP relative to LQ-EMs (Appendix G, Table 2). Our reading of the evidence is

### _wp16177 - 0.6 percent of GDP relative to LQ-EMs (Appendix G, Table 2). Our reading of the evidence is

### Robustness of DiD results and portfolio flow comparisons
- Results are highly robust across different specifications, although some results lose significance in some specifications.
- DiD regressions comparing all LIDCs (FMs and NFM-LIDCs) as a group to LQ-EMs (Appendix G, Table 3) show coefficients of interest are not statistically significant when comparing LQ-EMs with the whole LIDC universe.
- No evidence of increased portfolio flows to LIDCs as a whole that would suggest convergence towards, or greater similarity with, EMs in terms of these flows.
- Initial findings of a similar trajectory between FMs and EMs are specific to FMs: after 2008, private capital flows to FMs have been on a different trajectory from those to the rest of the LIDC group.
- Quantitative finding: portfolios flows to FMs have exceeded those to EMs by about 0.6 percentage points of GDP a year.

### Financial Integration: hypothesis (H3) and empirical framework
- Hypothesis H3: As a result of FMs’ larger capital flows after 2008, their financial integration with global markets increased; increased economic and financial integration implies local assets are more affected by world shocks and correlations (betas) would increase.
- Empirical approach: calculate global market betas using an ICAPM model (single-factor and two-factor CAPM specifications).
- Sample: 2000–14, monthly frequency, covering 12 FMs and 10 EMs.
- Global market proxy: Barclays Global Bond Index (R_Mkt); risk-free rate proxied by 3-month US T-bill rate (R_f).
- Dependent variable: R_i,t − R_t^f where R_i,t is country i bond index return at month t; all returns in US$ prices.
- Single-factor ICAPM (Eq. 4): R_i,t − R_t^f = α_i + β*(R_t^Mkt − R_t^f) + γ_i + ε_i,t.
- Two-factor ICAPM (Eq. 5, partial-integration): R_i,t − R_t^f = α_i + β1*(R_t^Mkt − R_t^f) + β2*(R_t^local − R_t^f) + γ_i + ε_i,t.
  - For FMs, local index proxied by JP NEXGEM index (R_nexgem); for EMs, local index proxied by EMBI (R_embi).
- Interpretation: insignificant global beta implies segmentation; significant global beta implies integration. Expect significant global β for FMs after 2008 if H3 holds.

### ICAPM estimation results (single-factor and two-factor findings)
- Single-factor ICAPM results:
  - For FMs:
    - Before 2008: global market beta insignificant (segmentation).
    - After 2008: global market beta positive and significant; post-2008 market beta for FMs ~ 1.7 (comparable to EMs).
    - Numeric example: a change in international bond returns of 1 percent translates into a 1.7 percent change in FM bond returns after 2008; bond prices move opposite to returns, so FM bond prices will decrease more than advanced markets for a negative shock.
  - For EMs:
    - Global market beta positive and significant in both periods, with a slight increase after 2008.
    - Post-2008 market beta for EMs also approximately 1.7.
- Table 6 highlights (preserving reported numbers exactly):
  - Panel A: ICAPM Frontier Markets (monthly observations 824 before, 839 after)
    - Single-factor beta estimates (columns 1–4): BEFORE 0.454 (0.312); AFTER 1.716*** (0.391); BEFORE with FE 0.499 (0.322); AFTER with FE 1.716*** (0.402)
    - Two-factor (columns 5–6): BEFORE 0.179 (0.282); AFTER 1.121*** (0.323); R_nexgem − R_f coefficients: AFTER 0.473*** (0.127) and BEFORE 0.339** (0.159) as reported
    - R-squared: single-factor BEFORE 0.008; AFTER 0.081; with FE BEFORE 0.020; AFTER 0.088; two-factor BEFORE 0.140; AFTER 0.315
  - Panel B: ICAPM Emerging Markets (monthly observations 702 before, 730 after)
    - Single-factor beta estimates: BEFORE 1.373*** (0.332); AFTER 1.669*** (0.439); with FE BEFORE 1.312*** (0.331); AFTER 1.656*** (0.450)
    - Two-factor (columns 5–6): BEFORE 0.541** (0.219); AFTER 0.713** (0.327); R_embi − R_f coefficients: AFTER 0.744*** (0.0509) and BEFORE 0.681*** (0.0413) as reported
    - R-squared: single-factor BEFORE 0.079; AFTER 0.059; with FE BEFORE 0.135; AFTER 0.072; two-factor BEFORE 0.641; AFTER 0.531
- Two-factor ICAPM results:
  - Support partial-integration: both local and global factors are priced separately for FMs and EMs.
  - For FMs: before 2008 only local factors are priced; after 2008 both local and global factors are priced.
  - R^2 for FMs increases from 8.8 percent to 31.5 percent between single-factor and two-factor (post-2008), indicating improved explanatory power but also substantial unexplained variance reflecting omitted risk factors and idiosyncratic risks.
  - For EMs: local and global factors significant both before and after 2008, consistent with partial-integration already in the 2000s.

### Interpretation: integration, vulnerabilities, and convergence
- Findings support financial integration hypothesis H3: following monetary easing and increased capital flows after 2008, FMs moved from full-segmentation to greater integration with global markets.
- Consequence: FMs now resemble EMs in terms of vulnerability to global shocks and external risks.
- Economic significance: with market betas around 1.7 post-2008, FMs can experience larger return (and therefore price) responses to global bond market shocks than advanced markets.

### Policy implications and conclusions
- Change in capital flow landscape to developing countries may have policy implications for FMs.
- Finding of greater vulnerability to adverse global financial developments points to importance of frameworks for FMs to:
  - manage vulnerabilities to capital flow reversals, and
  - cope with reversals that could jeopardize macroeconomic performance.
- Empirical summary:
  - DiD results: since the 2008 crisis, FMs resemble EMs in capital flow patterns and differ from the rest of LIDCs.
  - Quantitative summary: portfolios flows to FMs have exceeded those to EMs by about 0.6 percentage points of GDP a year.
  - Risk-adjusted flows: when level of risk aversion is taken into account, portfolio flows to FMs are not statistically different from those to EMs, implying similar vulnerability during market stress (examples cited: 2008, 2009, 2013).
  - ICAPM results: FMs were segmented before 2008 but show post-2008 positive and significant market betas, indicating increased integration.

### Caveats and areas for further research
- Conclusions are based on changes during a relatively short period of unusually lax monetary conditions in major AEs and may not reflect a permanent shift.
- Identification of FMs in the sample may entail sample selection bias; regressions control for other determinants of portfolio flows to mitigate bias.
- Bond return data limitations: not all FMs could be included in ICAPM regressions; a larger sample would improve robustness.
- The study does not quantify implications for FMs’ financing needs or prescribe specific frameworks for strengthening resilience—areas suggested for further research.

*Source: IMF working paper content (excerpt provided).*

### Appendix B. Sub-Groups

### Appendix B. Sub-Groups

### Subgroup definitions and methodology
- LIDCs group divided into two subgroups for analytical purposes: FMs (Frontier Markets) and the rest of the LIDCs, based on characteristics that are key drivers of economic performance (IMF, 2014).
- FM classification is time-varying (drawn from De Lira and Nkusu (2015)).
- Control group consists of EMs (Emerging Markets).

### Frontier Markets (FMs) and EMs (Control Group)
- FMs (with year in which each met the FMs classification criteria):
  - Bangladesh (2010)
  - Bolivia (1997)
  - Côte d’Ivoire (2010)
  - Ghana (1997)
  - Kenya (1996)
  - Mongolia (2008)
  - Mozambique (1999)
  - Nigeria (1998)
  - Papua New Guinea (2003)
  - Senegal (2009)
  - Tanzania (2010)
  - Uganda (2010)
  - Vietnam (2008)
  - Zambia (1998)
  - (Honduras)
  - Ukraine
- EMs (Control Group):
  - Angola
  - Armenia
  - Egypt
  - El Salvador
  - Fiji
  - Georgia
  - Guatemala
  - Indonesia
  - Kosovo
  - Morocco
  - Paraguay
  - Sri Lanka
  - Swaziland
  - Syria
  - Tunisia

### Appendix C — Deriving a Frontier Market Group Combining IMF and Market Analysts’ Classifications
- Classification scheme:
  - For each country that does not exceed the income threshold separating LMIC from UMIC, each market analyst and the IMF assign discrete values: -1 (EM), 1 (FM), 0 (neither).
  - The IMF’s classification is assigned the same weight as all market analysts together; market analysts are assigned equal weights individually.
  - If the overall weighted average rating (WaR) is at least 0.5, Classification = FM (WaR≥0.5).
  - Yellow highlights indicate country is not FM; green highlights indicate country is FM after combining IMF and market analysts' classifications.
- Selected country entries (Country | IMF | Next Eleven | FTSE | MSCI | Russell | NEX | GEM | EMBI | Weighted average rating (WaR) | Classification, FM=1 if WaR≥0.5):
  - Indonesia -1 -1 -1 -1 -1 -1 -0.917 0
  - Egypt -1 1 1 1 1 1 -0.667 0
  - Angola -1 0 0 0 0 1 -1 -0.5000 0
  - El Salvador -1 0 0 0 0 1 -1 -0.5000 0
  - Fiji -1 0 0 0 0 0 0.5000 1
  - Guatemala -1 0 0 0 0 1 -0.5000 0
  - Kosovo -1 0 0 0 0 0 0.5000 1
  - Paraguay -1 0 0 0 0 1 -0.5000 0
  - Swaziland -1 0 0 0 0 0 0.5000 1
  - Syria -1 0 0 0 0 0 0.5000 1
  - Armenia -1 0 0 0 0 1 0.4170 0
  - Georgia -1 0 0 0 0 1 0.4170 0
  - Ukraine - 1 0 0 1 1 0.4170 0
  - Morocco -1 0 1 1 1 0.3330 0
  - Tunisia -1 0 1 1 1 0.3330 0
  - Sri Lanka -1 0 1 1 1 1 0.2500 0
  - Bolivia 1 0 0 0 0 1 0.5001 1
  - Mongolia 1 0 0 0 0 1 0.5001 1
  - Papua New Guinea 1 0 0 0 0 0 0.5001 1
  - Tanzania 1 0 0 0 0 1 0.5001 1
  - Uganda 1 0 0 0 0 0 0.5001 1
  - Mozambique 1 0 0 0 0 1 0.5831 1
  - Senegal 1 0 0 1 1 0.5831 1
  - Zambia 1 0 0 1 1 0.5831 1
  - Honduras 0 0 0 0 0 1 0.0830 0
  - Bangladesh 1 -1 1 1 1 0 0.6671 1
  - Cote d'Ivoire 1 0 1 0 1 1 0.6671 1
  - Nigeria 1 -1 1 1 1 -1 0.6671 1
  - Vietnam 1 -1 1 1 1 -1 0.6671 1
  - Ghana 1 0 1 1 1 -1 0.7501 1
  - Kenya 1 0 1 1 1 0 0.8331 1

### Appendix D — Definition of Variables
- Dependent variable:
  - NPI/Y: Net portfolio investment scaled by GDP, where NPI_i,t is the net portfolio investment and Y_i,t is country i’s nominal GDP. Source: WEO.
- Key explanatory variables — Push factors:
  - U.S. 10-year Treasury bond yield. Source: Federal Reserve Economic Data – FRED – St. Louis Fed.
  - VIX (risk aversion): Chicago Board Options Exchange Market Volatility Index; Source: Bloomberg.
- Key explanatory variables — Pull factors:
  - Growth: Output Growth. Source: IMF.
  - Total Debt-to-GDP Ratio: Sovereign Debt scaled by GDP. Source: WEO.
  - Fiscal-Balance-to-GDP Ratio: Fiscal Balance scaled by GDP. Source: WEO.
  - Governance: Country Financial Governance Index. Source: International Country Risk Guide (ICRG).
- Others:
  - R_Mkt - R_f: Excess Returns. Source: J.P Morgan and Federal Reserve Economic Data – FRED – St. Louis Fed.
  - R_f_t: Risk-free rate. Source: Federal Reserve Economic Data – FRED – St. Louis Fed.
  - R_i,t : ∆lnP_i,t, where P_i,t is the bond index of country i in month t. Source: Barclays.

### Appendix E — Financial Integration, CAPM, Sub-Groups
- Data obtained from NEXGEM and EMBIG JP Morgan Indices.
- FMs listed in this appendix:
  - Bolivia, Ghana, Côte d’Ivoire, Senegal, Tanzania, Vietnam, Zambia, Nigeria, Kenya, Mongolia, Honduras, Mozambique
- EMs listed in this appendix:
  - Angola, Egypt, El Salvador, Georgia, Indonesia, Morocco, Paraguay, Sri Lanka, Tunisia, Guatemala, Ukraine

### Appendix F — Econometric Framework and Identification Strategy
- Estimation strategy:
  - Uses difference-in-differences (DD) comparing FMs and NFM-LIDCs (treatment groups) with lower bound EMs (control group).
  - Estimators account for time-invariant differences between groups via inclusion of FM_i,t, LIDC_i,t and NFM_i,t and common trends via Crisis_t.
  - Inclusion of controls X_i,t reduces endogeneity concerns and improves estimator efficiency (Roberts and Whited, 2012).
- Implications for estimated crisis effects:
  - (E4) - (E3) = effect of the crisis on the EMs control group = β1.
  - (E6) - (E5) = effect of the crisis on the FMs = β1` β3.
  - (E8) - (E7) = effect of the crisis on the NFM-LIDCs = β1 + β4.
  - DD estimator for FMs is obtained by differentiating (E6 - E5) and (E4 - E3), yielding β3.

### Appendix G — Additional Regression Tables: Key reported estimates and statistics
- Table 1. Investigating Treatment Effects per year (Dependent variable: NPI/Y; Observations between 2000 and 2014; country and year fixed effects; standard errors clustered at country level)
  - Selected year-by-FM interaction coefficients (columns (1)–(4)):
    - (year==2001)*FM_i,t = 0.304 (s.e. 0.532)
    - (year==2007)*FM_i,t = 1.243** (s.e. 0.530)
    - (year==2010)*FM_i,t = 1.169** (s.e. 0.530)
    - (year==2011)*FM_i,t = 1.115** (s.e. 0.531)
    - (year==2012)*FM_i,t = 2.045*** (s.e. 0.531)
    - (year==2014)*FM_i,t = 1.254** (s.e. 0.531)
  - Observations: 1,040 (columns 1–2), 1,002 (columns 3–4).
  - R-Squared: 0.233 (cols 1–2), 0.232 (col 3), 0.242 (col 4).
  - F tests on equality of coefficients examples:
    - F-stat 2008=2011 = 2.91, p-val 2011 = 0.0549 (cols 1–2)
    - F-stat 2008=2012 = 9.88, p-val 2012 = 0.0001 (cols 1–2)
    - F-stat 2008=2013 = 0.21, p-val 2013 = 0.8137 (cols 1–2)
- Table 2. Robustness Check: Controlling for Governance (Observations between 2000 and 2014; country and year fixed effects; standard errors clustered at country level)
  - Crisis_t * Frontier_t coefficients across columns (1)–(6):
    - 0.674* (s.e. 0.374)
    - 0.649* (s.e. 0.374)
    - 0.632** (s.e. 0.279)
    - 0.630** (s.e. 0.281)
    - 0.647** (s.e. 0.273)
    - 0.638** (s.e. 0.276)
  - Crisis_t * NFM_t coefficients range: -0.284 to -0.315 (s.e. range 0.275–0.287).
  - VIX_t coefficients include -0.0299*** (s.e. 0.00811) in one specification.
  - U.S 10-Y TB yield_t coefficients include -0.379** (s.e. 0.147) and -0.279*** (s.e. 0.101).
  - Growth_t coefficients up to 1.270** (s.e. 0.571).
  - Debt/GDP_t-1 reported as -1.57e-06*** (s.e. 5.16e-07) in one column and similar magnitudes in others.
  - Governance_t coefficients reported as 0.00774 (s.e. 0.0158) and 0.00146 (s.e. 0.0164) in specified columns.
  - Observations: 731, 682, 580, 580, 571, 571 across columns.
  - R-squared: 0.164, 0.199, 0.188, 0.221, 0.188, 0.222 across columns.
- Table 3. Robustness Check: Investigating the Difference of all LIDCs with EMs (Dependent variable: NPI/Y; Observations between 2000 and 2014; country and year fixed effects)
  - Crisis_t * LIDC_t coefficients across columns (1)–(5):
    - 1.569 (s.e. 1.380)
    - 1.569 (s.e. 1.380)
    - 0.0118 (s.e. 0.236)
    - -0.00923 (s.e. 0.234)
    - 0.0544 (s.e. 0.226)
  - VIX_t coefficients include -0.0549 (s.e. 0.0404) and -0.0208*** (s.e. 0.00769).
  - Export/GDP_t coefficient reported as -0.0155** (s.e. 0.00695) in one column.
  - Observations: 1,220 (cols 1–2), 896 (cols 3–5).
  - R-squared: 0.561 (cols 1–2), 0.313–0.318 (cols 3–5).

*Source: _wp16177 - Appendix B. Sub-Groups*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp16177.pdf_