## _wp09111

## Source details

**Canonical URL:** [_wp09111](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09111.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09111.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09111.pdf.json)

---

### Key findings on EM equity market performance
- After more than a year of relatively small spillovers from advanced-economy financial turmoil, equity prices in emerging markets (EM) "succumbed to the dramatic worsening of financial distress in mid-September 2008."
- In spring 2009, despite steep and abrupt declines, emerging equity prices as a group were still above their level in 2003, "which marked the beginning of their steep rally."
- Domestic/fundamental factors and global/external conditions both play important and roughly equal roles in driving EM equity valuations.
- Despite increased financial integration and higher foreign investor participation, "to date, the more open EM economies or those with higher foreign investor participation have not been affected disproportionately by the global financial turbulence."
- Current-cycle declines (October 2007–August 2008) have been "shallower and more differentiated across EM countries" compared with the previous cycle’s declines associated with generalized EM crises in the 1990s.
- Stock market correlations with the United States increased from "0.17 during the period January 1998–December 2002" to "0.91 during the period January 2003–May 2008."

### Policy implications and recommendations
- Build and sustain resilient capital markets, particularly equity markets, by:
  - Fostering deeper capital markets.
  - Introducing legal, regulatory, and accounting reforms that conform to international standards.
  - Developing a well-functioning securities market with supporting infrastructure.
- Emphasize measures that make equity markets more resilient when equity prices decline.
- Specific recommendations to enhance structural resilience:
  - Foster deeper and more liquid capital markets with diverse institutional investors (domestic and foreign; buy-and-hold and active participants).
  - Establish funded pension schemes and develop domestic insurance sector to broaden local institutional investor base and demand for long-term instruments.
  - Develop more diverse local financial products and extend yield curves on sovereign and corporate fixed-income securities as well as equities.
  - Weigh benefits of discretionary interference by authorities in structured, formally regulated markets against possible reputational costs for market development.
  - Ensure reforms in legal, regulatory, prudential, and accounting systems are consistent with international standards.
  - Improve firm-level governance, transparency, and risk management.
  - Build robust securities market infrastructure and institutions (securities exchange and clearing systems) and supporting markets (repo, margin trading, securities lending, derivatives).
  - Sequence enhancements to securities market infrastructure and introduction of new financial instruments with appropriate oversight.
  - Develop derivatives markets within solid product design, regulation, and market infrastructure and oversight.
  - Ensure a well-functioning stock-lending system as prerequisite for proficient short selling; caution that introducing shorting in a bearish market could exacerbate downturn.
- Operational note: careful implementation is important at each stage to reap benefits of innovation while managing financial stability risks and ensuring proper market functioning.

### Empirical approach and data
- Sample and frequency:
  - Monthly observations, "January 2001 to May 2008."
  - Cross-section: "30 EM economies."
- Dependent variable:
  - Log change in the U.S. dollar equity price indices (panel regressions); log level used in the panel cointegration.
- Two groups of explanatory variables:
  - Domestic/fundamental factors:
    - ΔGDP = the change of GDP growth
    - ΔINT = Interest rate differential
    - ΔEXCHRATE = Exchange rate expectation
    - ΔINFLATION = Inflation rate differential
    - ΔMARRATIO = Equity market capitalization to GDP ratio
    - ΔPRICRED = Credit to the private sector
  - Global/financial factors:
    - ΔLIQ = Global excess liquidity (the difference between broad money growth and estimates for money demand in the euro area, Japan, and the United States)
    - ΔCREDRISK = Credit risk premium (level of the 10-year U.S. dollar swap spread)
    - ΔMARRISK = Market risk premium (implied volatility of the S&P 500 index (VIX))
    - ΔEQINF = Portfolio equity inflow
    - ΔMSCI = Morgan Stanley Capital International world index
- Two panel specifications:
  - First specification: market capitalization/GDP included; panel cointegration for log levels of equity price index, forward exchange rate, market capitalization.
  - Second specification: private credit replaces market capitalization; MSCI added to global factors.
- Panel cointegration framework used to capture long-run relationships and relax exogeneity assumptions.

### Unit root and panel cointegration evidence
- Unit root tests indicate nonstationarity of:
  - log equity price index, log market capitalization, and log forward exchange rate.
- Unit root statistics (Table 2):
  - Levin-Lin rho-stat: 4.13; 1.99; 4.76
  - Levin-Lin t-rho-stat: 5.39; 3.73; 6.81
  - Levin-Lin ADF-stat: 3.57; 3.87; 4.57
  - IPS ADF-stat: 4.30; 1.14; 5.22
- Pedroni heterogeneous panel cointegration test results (Table 3) for log price indices, log forward exchange rate, log market capitalization:
  - Panel v-stat: 3.47**
  - Panel rho-stat: -3.67**
  - Panel pp-stat: -2.16**
  - Panel adf-stat: 1.34
  - Group rho-stat: -4.28**
  - Group pp-stat: -3.69**
  - Group adf-stat: 1.48
- Interpretation: residual e_it = SM_it - α_it - β_1i,t EXCHRATE_1i,t - β_2i,t MARCAP_2i,t is stationary (e_it ~ I(0)) when the three series are cointegrated.

### Error correction and long-run adjustment
- Error correction term interpretation:
  - A 1 percent reduction in the error correction term is associated with a 0.05 percent increase in equity prices (reflecting adjustment toward the long-run relationship).
- Error correction term coefficients (Table 4, first specification):
  - Error correction term = -0.0519 (30 economies), -0.0381 (Asia), -0.0451 (EMEA), -0.0811 (Latin America); p-values indicate significance.
- Error correction term coefficients (Table 5, second specification):
  - Error correction term = -0.0358 (30 economies), -0.0222 (Asia), -0.0305 (EMEA), -0.0660 (Latin America).

### Panel regression key coefficients and model fit (monthly Jan. 2001–May 2008)
- Sample sizes and fit:
  - Time-series sample (monthly): 2001m1-2008m5
  - No. of cross-section economies: 30 (full), 12 (Asia), 12 (EMEA), 6 (Latin America)
  - No. of observations: 2293 (30 economies); Asia 891; EMEA 875; Latin America 527
  - Adjusted R^2 (first specification): 0.3498 (30 economies); Asia 0.4465; EMEA 0.3225; Latin America 0.5754
- First specification — Selected domestic factors and coefficients (coefficients; p-values):
  - GDP growth: 0.9849 (0.00000)***; Asia 1.3615 (0.00081)***; EMEA 0.5364 (0.11908); Latin America 0.5282 (0.06237)*
  - Interest rate differential: -0.0671 (0.00258)***; Asia -0.1848 (0.02039)**; EMEA 0.0557 (0.46182); Latin America -0.0201 (0.45839)
  - Forward exchange rate: 0.0419 (0.00000)***; Asia 0.0146 (0.04899)**; EMEA 0.4994 (0.00000)***; Latin America 0.3771 (0.00000)***
  - Inflation rate differential: 0.0252 (0.37505); Latin America 0.1917 (0.00348)***
  - Market capitalization/GDP ratio: 0.2580 (0.00000)***; Asia 0.3778 (0.00000)***; EMEA 0.1759 (0.00000)***; Latin America 0.3071 (0.00000)***
- First specification — Selected external factors:
  - Excess global liquidity: 1.4400 (0.00000)***; Asia 1.8083 (0.00000)***; EMEA 0.8258 (0.01487)**; Latin America 1.2565 (0.00007)***
  - Credit risk premium: -12.5266 (0.00000)***; Asia -10.9920 (0.00000)***; EMEA -9.0441 (0.00017)***; Latin America -11.3054 (0.00000)***
  - Market risk premium: -0.3064 (0.00000)*** (similar significance across regions)
  - Foreign equity inflow: 0.0000 (0.52484) — not significant
- Second specification — Selected domestic factors:
  - Credit growth: 0.7124 (<0.0001)***; Asia 0.6777 (<0.0001)***; EMEA 0.4711 (<0.0001)***; Latin America 0.3456 (<0.0001)***
  - GDP growth: 0.3777 (0.0893)*; region-specific p-values reported in source
  - Forward exchange rate: 0.0361 (0.0002)***; EMEA 0.3816 (<0.0001)***; Latin America 0.6384 (<0.0001)***
- Second specification — Selected external factors:
  - Global liquidity: 0.9203 (<0.0001)***; Asia 1.5011 (<0.0001)***; EMEA 0.3663 (0.3334); Latin America 0.8726 (0.0235)**
  - Market risk premium: -0.2746 (<0.0001)***; credit risk premium: -6.7535 (<0.0001)***
  - MSCI: 0.1141 (0.0187)** for full sample
  - Foreign equity inflow: not significant for full sample; significant for EMEA in this specification (0.0004, p-value (0.0169)**)

### Scenarios of external shocks and quantified effects (Table 6)
- Scenario 1: 10 percent decline in global excess liquidity (from current level of 4.5 percent) and 10 percent increase in both credit and market risk premia:
  - Global excess liquidity declines by 10 percent: -0.65 (percent change in equity prices)
  - U.S. credit risk premium increases by 10 percent: -0.74
  - U.S. market risk premium increases by 10 percent: -0.55
  - Three combined scenarios: -1.94
- Scenario 2: Larger shock (zero excess liquidity and historical peak risk premia):
  - Zero global excess liquidity: -6.52
  - U.S. credit risk premium at historical peak (May 2000): -10.02
  - U.S. market risk premium at historical peak (August 1998): -8.10
  - Three combined scenarios: -24.64
- Historical reference levels used in Scenario 2:
  - Credit risk premium: May 2000 = 1.39 versus 0.59 in May 2008
  - Market risk premium: August 1998 = 44 versus 18 in May 2008

### Contribution analysis: global versus domestic factors (Table 7 and method)
- Method (four steps) summary:
  - Step 1: Multiply historical value of each variable by estimated coefficient ("model values"), domestic factors weighted by each economy's annual GDP.
  - Step 2: Sum model values across 30 economies and 89 periods to get total change.
  - Step 3: Calculate each variable’s share in total change.
  - Step 4: Sum shares of domestic and external factors to obtain contributions (global factors averaged over 30 countries).
- Result (Table 7):
  - Share of global factors: 45.8 (In percent)
  - Share of domestic factors: 54.2 (In percent)
- Interpretation: Global and domestic factors explain equity price increases during Jan 2001–May 2008 in roughly equal measure, conditional on model, country and period coverage, and specification.

### Vector autoregression (VAR) analysis — approach and summarized findings
- Seven separate eight-variable structural VARs estimated for: Argentina, Chile, China, Romania, Russia, Singapore, South Africa.
- Country selection based on Chinn-Ito Financial Openness Indicator (least open: Argentina, China, Russia; most open: Chile, Romania, Singapore, South Africa).
- VAR structure:
  - Exogenous foreign block: global excess liquidity, U.S. credit and market risk premia.
  - Economy-specific block: GDP growth, forward exchange rate, interest rate and inflation rate differentials, market capitalization to GDP ratio.
  - Lag length: Schwarz's BIC points to one lag except Romania (2 lags).
- Identification and estimation:
  - Granger Causality and Block Exogeneity Wald Tests: global excess liquidity, credit and market risk premia generally exogenous.
  - Cholesky decomposition (degrees-of-freedom adjusted) used for impulses; different orderings do not materially change significance of spillovers.
- VAR results (summary):
  - Equity price impulse responses generally align with panel regression results.
  - Three observations across countries:
    - (i) Most equity price responses to global indicator shocks are in expected direction and same sign as panel results.
    - (ii) Equity price response typically tapers off after three months.
    - (iii) Smaller countries tend to have slightly larger responses overall.
  - Statistical significance: In most cases shocks are not statistically different from zero except response to market risk in the first month for all but Argentina and China.

### Comparative observations and stylized facts
- EM equity market trends and capital flows:
  - EM equity prices rose significantly beginning in 2003; associated with a concomitant rise of EM capital inflows.
  - EM cross-border outward investments grew and often masked high gross capital inflows in net terms.
  - In some mature EMs, stock market capitalization-to-GDP ratios are "now approaching that of advanced economies," though sustainability is uncertain.
  - Total equity market returns in many EM economies increased at a much faster pace than in advanced economies during the upturn.
- Foreign investor presence:
  - "Foreign holdings of EM equity have increased overall since 2003," though not necessarily as a proportion of total equity value in all cases.
  - The presence of foreign investors "does not seem to be associated with larger equity losses since the October 2007 peak."
- Cycle comparisons:
  - Current equity corrections in both EMs and advanced economies have been "shallower and more gradual" than many past episodes; current upturns in advanced economies were more modest relative to past bubbles (e.g., Japan in the late 1980s, U.S. dot-com bubble).

### Methodological issues (Annex I) and data notes (Annex II)
- Rationale for panel cointegration:
  - Pedroni panel cointegration tests allow complete endogeneity, heterogeneous dynamics and cointegrating vectors; robust to endogeneity, omitted variables, simultaneity and measurement error.
  - Framework permits α_i and β_i to vary across panel members, accommodating heterogeneous short- and long-run dynamics.
- Panel cointegration regression specification:
  - y_it = α_i + β_1i,t x_1i,t + β_2i,t x_2i,t +...+ β_ki,t x_ki,t + γ_i t + e_it
  - If y_it and x_ki,t are I(1), they are cointegrated if e_it is I(0).
- Data and variable implementation notes:
  - Monthly sample of 30 economies from January 2001 to May 2008.
  - Panel regressions run on specified country lists for Asia, EMEA, and Latin America.
  - GDP growth uses monthly consensus forecast for annual GDP growth rate in the Consensus Forecast Database.
  - Interest rate differential: spread between 1-year domestic and 6-month U.S. Treasury rates.
  - Exchange rate expectation: log change in forward exchange rates (including NDFs where forward markets unavailable); log level in cointegration tests.
  - Equity market capitalization to GDP ratio calculated as (1+growth of market capitalization)/(1+ GDP growth) in regressions; log level in cointegration.
  - Global excess liquidity: difference between broad money growth and estimates for money demand in the euro area, Japan, and the United States.
  - Credit risk premium: level of the 10-year U.S. dollar swap spread.
  - MSCI added in second specification; only Hong Kong SAR and Singapore overlap with panel to minimize endogeneity.
  - Period since 2001 chosen for long upward trend starting from a low point and improved data availability.

_Source: _wp09111 - References and Annexes (IMF PDF)._

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

### _wp09111 - References..............................................................................................................

### Key findings on EM equity market performance
- After more than a year of relatively small spillovers from advanced-economy financial turmoil, equity prices in emerging markets (EM) "succumbed to the dramatic worsening of financial distress in mid-September 2008."
- In spring 2009, despite steep and abrupt declines, emerging equity prices as a group were still above their level in 2003, "which marked the beginning of their steep rally."
- The paper finds that "domestic/fundamental factors, such as growth and exchange rate expectations, and global/external conditions, such as excess liquidity and credit and market risk premia, both play important and roughly equal roles" in driving EM equity valuations.
- Despite increased financial integration and higher foreign investor participation, "to date, the more open EM economies or those with higher foreign investor participation have not been affected disproportionately by the global financial turbulence."
- Current-cycle declines (October 2007–August 2008) have been "shallower and more differentiated across EM countries" compared with the previous cycle’s declines associated with generalized EM crises in the 1990s.
- Stock market correlations with the United States increased from "0.17 during the period January 1998–December 2002" to "0.91 during the period January 2003–May 2008."

### Policy implications and recommendations
- EM countries should build and sustain resilient capital markets, particularly equity markets, by:
  - Fostering deeper capital markets.
  - Introducing legal, regulatory, and accounting reforms that conform to international standards.
  - Developing a well-functioning securities market with supporting infrastructure.
- Emphasis is placed on measures that make equity markets more resilient when equity prices decline.

### Empirical approach and data
- Sample and frequency:
  - Monthly observations, "January 2001 to May 2008."
  - Cross-section: "30 EM economies."
- Two groups of explanatory variables:
  - Domestic/fundamental factors:
    - ΔGDP = the change of GDP growth
    - ΔINT = Interest rate differential
    - ΔEXCHRATE = Exchange rate expectation
    - ΔINFLATION = Inflation rate differential
    - ΔMARRATIO = Equity market capitalization to GDP ratio (captures increases in volume of shares and provides a proxy for equity market depth)
    - ΔPRICRED = Credit to the private sector
  - Global/financial factors:
    - ΔLIQ = Global excess liquidity (the difference between broad money growth and estimates for money demand in the euro area, Japan, and the United States)
    - ΔCREDRISK = Credit risk premium (level of the 10-year U.S. dollar swap spread)
    - ΔMARRISK = Market risk premium (implied volatility of the S&P 500 index (VIX))
    - ΔEQINF = Portfolio equity inflow
    - ΔMSCI = Morgan Stanley Capital International world index
- Model specifications:
  - The model is specified in terms of (log) differences of all fundamental/domestic and all global/external variables.
  - Two fixed-effects panel specifications are defined (explicit algebraic forms provided in the source), both including an error correction term from panel cointegration.
  - The error correction term ε represents the deviation from the cointegrating relationship.
- Nonstationary/panel cointegration framework:
  - Unit root tests indicate that "log equity price index, log market capitalization, and log forward exchange rate" are nonstationary, while "all other variables used in the panel regressions are stationary."
  - The cointegrating relationship assessed is:
    - SM_it = log equity price index
    - EXCHRATE_1i,t = log forward exchange rate
    - MARCAP_2i,t = log market capitalization
    - The residual e_it = SM_it - α_it - β_1i,t EXCHRATE_1i,t - β_2i,t MARCAP_2i,t is stationary (e_it ~ I(0)) when the three series are cointegrated.
  - The panel cointegration framework allows relaxing exogeneity assumptions and captures long-run relationships robust to short-run dynamics and endogeneity.

### Comparative observations and stylized facts
- EM equity market trends and capital flows:
  - EM equity prices rose significantly beginning in 2003; this upturn was associated with a concomitant rise of EM capital inflows.
  - EM cross-border outward investments grew and, in net terms, often masked high gross capital inflows.
  - In some mature EMs, stock market capitalization-to-GDP ratios are "now approaching that of advanced economies," though sustainability is uncertain.
  - Total equity market returns in many EM economies increased at a much faster pace than in advanced economies during the upturn.
- Foreign investor presence:
  - "Foreign holdings of EM equity have increased overall since 2003," though not necessarily as a proportion of total equity value in all cases.
  - The presence of foreign investors "does not seem to be associated with larger equity losses since the October 2007 peak."
- Cycle comparisons:
  - Current equity corrections in both EMs and advanced economies have been "shallower and more gradual" than many past episodes, and current upturns in advanced economies were more modest relative to past bubbles (e.g., Japan in the late 1980s, U.S. dot-com bubble).

*Source: _wp09111 - References (IMF PDF).*

### conclusion that the variables are cointegrated (Table 3) (Pedroni 1995, 1999).

### _wp09111 - conclusion that the variables are cointegrated (Table 3) (Pedroni 1995, 1999).

### Unit root and cointegration evidence
- Unit root test results (Table 2) indicate nonstationarity of price indices, forward exchange rate, and market capitalization:
  - Levin-Lin rho-stat: 4.13; 1.99; 4.76
  - Levin-Lin t-rho-stat: 5.39; 3.73; 6.81
  - Levin-Lin ADF-stat: 3.57; 3.87; 4.57
  - IPS ADF-stat: 4.30; 1.14; 5.22
- Note: Critical values are –1.28 (10 percent level, denoted by *) and –1.64 (5 percent level, denoted by **). The positive unit root test results show that price indices, forward exchange rate and market capitalization are nonstationary.
- Economic interpretation: The non-stationarity of forward exchange rates during the sample period indicates the general appreciation trend, the global "catching up" and integration into global capital markets of most emerging economies.

- Pedroni heterogeneous panel cointegration tests (Table 3) for log price indices, log forward exchange rate, log market capitalization:
  - Panel v-stat: 3.47**
  - Panel rho-stat: -3.67**
  - Panel pp-stat: -2.16**
  - Panel adf-stat: 1.34
  - Group rho-stat: -4.28**
  - Group pp-stat: -3.69**
  - Group adf-stat: 1.48
- Note: First four tests are pooled within-dimension; last three are group mean between-dimension. Critical values: v-stat 1.28 (10 percent, *) and 1.64 (5 percent, **); others –1.28 (10 percent, *) and –1.64 (5 percent, **).

### Error correction and long-run relationship
- Error correction term from panel cointegration among log equity prices, log market capitalization, and log forward exchange rate:
  - A 1 percent reduction in the error correction term is associated with a 0.05 percent increase in equity prices, reflecting adjustment toward the long-run relationship (Table 4).
- The error correction term coefficients in panel regressions:
  - First specification (Table 4): Error correction term = -0.0519 (30 economies), -0.0381 (Asia), -0.0451 (EMEA), -0.0811 (Latin America); all with (0.00000)*** or significant p-values as shown.
  - Second specification (Table 5): Error correction term = -0.0358 (30 economies), -0.0222 (Asia), -0.0305 (EMEA), -0.0660 (Latin America); p-values indicated in table.

### Panel regression key coefficients and significance (monthly Jan. 2001–May 2008)
- First specification (Table 4) — Selected domestic factors (coefficients; p-values in parentheses):
  - GDP growth: 0.9849 (0.00000)***; Asia 1.3615 (0.00081)***; EMEA 0.5364 (0.11908); Latin America 0.5282 (0.06237)*
  - Interest rate differential: -0.0671 (0.00258)***; Asia -0.1848 (0.02039)**; EMEA 0.0557 (0.46182); Latin America -0.0201 (0.45839)
  - Forward exchange rate: 0.0419 (0.00000)***; Asia 0.0146 (0.04899)**; EMEA 0.4994 (0.00000)***; Latin America 0.3771 (0.00000)***
  - Inflation rate differential: 0.0252 (0.37505); Latin America 0.1917 (0.00348)***
  - Market capitalization/GDP ratio: 0.2580 (0.00000)***; Asia 0.3778 (0.00000)***; EMEA 0.1759 (0.00000)***; Latin America 0.3071 (0.00000)***
- First specification — Selected external factors:
  - Excess global liquidity: 1.4400 (0.00000)***; Asia 1.8083 (0.00000)***; EMEA 0.8258 (0.01487)**; Latin America 1.2565 (0.00007)***
  - Credit risk premium: -12.5266 (0.00000)***; Asia -10.9920 (0.00000)***; EMEA -9.0441 (0.00017)***; Latin America -11.3054 (0.00000)***
  - Market risk premium: -0.3064 (0.00000)***; similar significance across regions
  - Foreign equity inflow: 0.0000 (0.52484) — not significant
- Model fit and sample:
  - Adjusted R^2: 0.3498 (30 economies); Asia 0.4465; EMEA 0.3225; Latin America 0.5754
  - Time-series sample (monthly): 2001m1-2008m5; No. of cross-section economies: 30 (full), 12 (Asia), 12 (EMEA), 6 (Latin America)
  - No. of observations: 2293 (30 economies); Asia 891; EMEA 875; Latin America 527

- Second specification (Table 5) — Selected domestic factors:
  - Credit growth: 0.7124 (<0.0001)***; Asia 0.6777 (<0.0001)***; EMEA 0.4711 (<0.0001)***; Latin America 0.3456 (<0.0001)***
  - GDP growth: 0.3777 (0.0893)*; Asia 0.6727 (-0.1785) etc. (p-values and regional differences reported)
  - Forward exchange rate: 0.0361 (0.0002)***; EMEA 0.3816 (<0.0001)***; Latin America 0.6384 (<0.0001)***
- Second specification — Selected external factors:
  - Global liquidity: 0.9203 (<0.0001)***; Asia 1.5011 (<0.0001)***; EMEA 0.3663 (0.3334); Latin America 0.8726 (0.0235)**
  - Market risk premium: -0.2746 (<0.0001)***; credit risk premium: -6.7535 (<0.0001)***
  - MSCI: 0.1141 (0.0187)** for full sample
  - Foreign equity inflow: not significant for full sample; significant for EMEA in this specification (0.0004, p-value (0.0169)**)

### Scenarios of external shocks and their quantified effects (Table 6)
- Scenario 1: 10 percent decline in global excess liquidity (from current level of 4.5 percent) and 10 percent increase in both credit and market risk premia:
  - Global excess liquidity declines by 10 percent: -0.65 (percent change in equity prices)
  - U.S. credit risk premium increases by 10 percent: -0.74
  - U.S. market risk premium increases by 10 percent: -0.55
  - Three combined scenarios: -1.94
- Scenario 2: Larger shock (zero excess liquidity and historical peak risk premia):
  - Zero global excess liquidity: -6.52
  - U.S. credit risk premium at historical peak (May 2000): -10.02
  - U.S. market risk premium at historical peak (August 1998): -8.10
  - Three combined scenarios: -24.64
- Historical reference levels used in Scenario 2:
  - Credit risk premium: May 2000 = 1.39 versus 0.59 in May 2008
  - Market risk premium: August 1998 = 44 versus 18 in May 2008

### Contribution analysis: global versus domestic factors (Table 7 and method)
- Method summary (four steps):
  - Step 1: Multiply historical value of each variable by estimated coefficient ("model values"), domestic factors weighted by each economy's annual GDP.
  - Step 2: Sum model values across 30 economies and 89 periods to get total change.
  - Step 3: Calculate each variable’s share in total change.
  - Step 4: Sum shares of domestic and external factors to obtain contributions (global factors averaged over 30 countries).
- Result (Table 7):
  - Share of global factors: 45.8 (In percent)
  - Share of domestic factors: 54.2 (In percent)
- Interpretation: The global and domestic factors in the model explain equity price increases during Jan 2001–May 2008 in roughly equal measure; result is specific to this model’s coefficients, country and period coverage, and specification.

### Vector autoregression (VAR) analysis — approach and summarized findings
- Seven separate eight-variable structural VARs estimated for: Argentina, Chile, China, Romania, Russia, Singapore, South Africa.
- Country selection criterion: most and least open economies in each region based on Chinn-Ito Financial Openness Indicator (least open: Argentina, China, Russia; most open: Chile, Romania, Singapore, South Africa).
- VAR structure:
  - Exogenous foreign block: global excess liquidity, U.S. credit and market risk premia.
  - Economy-specific block: GDP growth, forward exchange rate, interest rate and inflation rate differentials, market capitalization to GDP ratio.
  - Lag length: Schwarz's BIC points to one lag except Romania (2 lags).
- Identification and estimation:
  - Granger Causality and Block Exogeneity Wald Tests: global excess liquidity, credit and market risk premia generally exogenous.
  - Cholesky decomposition (degrees-of-freedom adjusted) used for impulses; different orderings do not materially change significance of spillovers.
- VAR results (summary):
  - Equity price impulse responses generally align with panel regression results.
  - Three observations across countries:
    - (i) Most equity price responses to global indicator shocks are in expected direction and same sign as panel results.
    - (ii) Equity price response typically tapers off after three months.
    - (iii) Smaller countries tend to have slightly larger responses overall.
  - Statistical significance: In most cases shocks are not statistically different from zero except response to market risk in the first month for all but Argentina and China.

### Main conclusions and policy implications
- Empirical findings:
  - Global forces matter about as much as domestic fundamentals for emerging equity prices.
  - Evidence of spillovers via equity market channel:
    - Significant negative relationship of global credit and market risk premia with EM equity prices.
    - Positive relationship of global excess liquidity with EM equity prices — indicating integration with advanced economies.
  - Strong domestic growth and financial deepening (credit growth, market capitalization) are key drivers of equity prices — indicating part of high EM equity prices are underpinned by domestic fundamentals.
- Policy recommendations to enhance structural resilience of emerging equity markets:
  - Foster deeper and more liquid capital markets with diverse institutional investors (domestic and foreign; buy-and-hold and active participants).
  - Establish funded pension schemes and develop domestic insurance sector to broaden local institutional investor base and demand for long-term instruments.
  - Develop more diverse local financial products and extend yield curves on sovereign and corporate fixed-income securities as well as equities to deepen and diversify domestic financial markets.
  - Weigh benefits of discretionary interference by authorities in structured, formally regulated markets against possible reputational costs for market development.
  - Ensure reforms in legal, regulatory, prudential, and accounting systems are consistent with international standards (noting some standards may be reformed in light of crises).
  - Improve firm-level governance, transparency, and risk management.
  - Build robust securities market infrastructure and institutions (securities exchange and clearing systems) and supporting markets (repo, margin trading, securities lending, derivatives) to reduce transaction costs and foster liquidity.
  - Sequence enhancements to securities market infrastructure and introduction of new financial instruments with appropriate oversight.
  - Develop derivatives markets within solid product design, regulation, and market infrastructure and oversight.
  - Ensure a well-functioning stock-lending system as prerequisite for proficient short selling; caution that introducing shorting in a bearish market could exacerbate downturn.
- Supporting operational note:
  - Careful implementation is important at each stage to reap benefits of innovation while managing financial stability risks and ensuring proper market functioning.

*Source: _wp09111 - conclusion that the variables are cointegrated (Table 3) (Pedroni 1995, 1999).*

### Annex I. Methodological Issues

### Annex I. Methodological Issues

### Panel cointegration approach and rationale
- Dynamic panel techniques based on IV or GMM estimation cannot correct for endogeneity induced by latent heterogeneity; Pedroni panel cointegration tests are used to allow complete endogeneity, heterogeneous dynamics and cointegrating vectors.
- Panel cointegration is robust to endogeneity and many forms of omitted variables, simultaneity and measurement error, and can isolate long-run steady state relationships from short-run dynamics.
- Pedroni (1995,1999) relaxes homogeneity by allowing the slope coefficient β to vary across the i individual members of the panel.
- Allowing αi and various βi to vary across panel members permits considerable short- and long-run heterogeneity: dynamics and fixed effects can differ across individuals and the cointegration vector can differ across members under the alternative hypothesis (Richard Harris and Robert Sollis, 2003).
- The extra cross-sectional dimension combined with cointegrating relationships enables fixed effects and heterogeneous trend components to capture a broad class of unobserved mechanisms.
- The nonstationary panel framework relaxes strong exogeneity assumptions typical of cross-sectional approaches and allows isolation of long-run steady state relationships from short-run dynamics.
- Overall, the cointegration framework accommodates a broad set of channels that may explain stock prices across economies.

### Panel cointegration regression specification
- The panel cointegration regression:
  - y_it = α_i + β_1i,t x_1i,t + β_2i,t x_2i,t +...+ β_ki,t x_ki,t + γ_i t + e_it
- If y_it, t=1,....,T, for each economy has a unit root (y_it ~ I(1)), and x_ki,t, t=1,....,T, has a unit root (x_ki,t ~ I(1)), then x_ki,t and y_it are cointegrated if some linear combination e_it = y_it - α_i - β_ki,t x_ki,t is stationary (e_it ~ I(0)).

### Empirical implementation choices
- Unit root tests were performed for all variables to avoid using nonstationary variables while maintaining a relatively large sample.
- Variables found stationary are used in driving force panel specifications; nonstationary variables are subject to cointegration tests to uncover stochastic and nonstochastic long-run trending features.

### Annex II. Data and explanatory variables

### Sample and dependent variable
- Monthly sample of 30 economies from January 2001 to May 2008.
- Dependent variable: the log change in the U.S. dollar equity price indices.
- Panel cointegration utilizes the log level in the U.S. dollar equity price indices.
- Panel regressions run on the following 30 economies:
  - Asia: China, Hong Kong SAR, India, Indonesia, Korea, Malaysia, Pakistan, Philippines, Singapore, Sri Lanka, Thailand, and Vietnam.
  - Europe, Middle East, and Africa (EMEA): Bulgaria, Croatia, Czech Republic, Estonia, Hungary, Latvia, Lithuania, Nigeria, Romania, Russia, South Africa, and Turkey.
  - Latin America: Argentina, Brazil, Chile, Colombia, Mexico, and Peru.

### First panel specification — independent variables
- Domestic factors:
  - GDP growth: the change of the monthly consensus forecast for annual GDP growth rate in the Consensus Forecast Database, in local currency, as a proxy for macroeconomic fundamentals.
  - Interest rate differential: the spread between the 1-year domestic and the 6-month U.S. Treasury rates.
  - Exchange rate expectation: the log change in forward exchange rates (including non-deliverable forward rates (NDF) in 18 economies for which data were available); the log level in forward exchange rates in the panel cointegration.
  - Inflation rate differential: the difference between the domestic and the U.S monthly inflation rates.
  - Equity market capitalization: equity market capitalization to GDP ratio, calculated as (1+growth of market capitalization)/(1+ GDP growth) in the panel regressions, as a proxy of capital market size; the log level of equity market capitalization in the panel cointegration.
- Global or external factors:
  - Global excess liquidity: the difference between broad money growth and estimates for money demand in the euro area, Japan, and the United States.
  - Market volatility premium: the implied volatility of the S&P 500 index (VIX) as a proxy for market volatility.
  - Credit risk premium: the level of the 10-year U.S. dollar swap spread, which is the difference between the 10-year U.S. dollar swap rate and the 10-year U.S. Treasury bond, as a proxy for aggregate default risk.
  - Portfolio equity inflows: the growth of flows to Emerging Markets (Asia, EMEA and Latin America) Equity Funds, as a proxy for portfolio equity inflows.

### Second panel specification — alternative variables
- Same countries, data frequency, and explanatory variables as the first specification except:
  - Equity market capitalization is replaced by private credit in the domestic factors.
  - MSCI is added to the global factors.
- Specifically:
  - Domestic credit: the log change in the credit to the private sector, as proxy for financial deepening.
  - MSCI: the log change in the Morgan Stanley Capital International world index of 23 major stock markets.

### Additional implementation notes (as provided)
- The period since 2001 is chosen because 2001 marked the beginning of a long upward trend for EM equities starting from a low point and because data availability, especially portfolio equity inflows, improved from that date.
- Forecasted GDP growth is chosen as a proxy for fundamentals rather than corporate profits, dividend yield, or taxes for reasons of data availability for this country sample.
- The 6-month U.S. Treasury rate is used because the 1-year U.S. Treasury rate is not available for the full period sample.
- NDFs are used as a proxy for the exchange rate in countries where forward foreign exchange trading has been banned by the government; in most cases they are primarily used for speculation (take exposure) in countries that have capital controls; covered interest rate parity does not necessarily hold in such cases, so the panel specification incorporates both interest rate differential and exchange rate expectation.
- Only two of the 23 MSCI country components are also in the panel, namely Hong Kong SAR and Singapore, minimizing potential endogeneity from adding the MSCI as an explanatory variable.

*Source: _wp09111 - Annex I. Methodological Issues*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09111.pdf_