## _wp07261

## Source details

**Canonical URL:** [_wp07261](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07261.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07261.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07261.pdf.json)

---

### Summary findings on IMF programs and expectations
- Using a multinomial probit model, IMF loans appear to have a strong effect on agent expectations in the early years.
- Two transmission channels identified:
  - inflow of real money, and
  - signaling effect.
- IMF programs during periods of collapsing growth:
  - reinforce underlying expectations for the future;
  - are associated with positive expectations for those with an optimistic outlook and negative expectations for those with a negative outlook.
- Once recovery is underway and economic uncertainty diminishes, IMF programs cease to have a statistically significant effect on the expectations of economic agents.
- Implication: IMF programs have the biggest impact on expectations during periods of great uncertainty and less impact when countries face minor shocks.

### Conceptual and methodological themes
- Expectations modeling approaches discussed:
  - Background (general framing of expectations),
  - Adaptive Expectations,
  - Rational Expectations,
  - Bounded Rationality.
- Measurement issues:
  - Direct measurement of expectations via survey data emphasized as advantageous compared with indirect measures such as capital flows or yield spreads.
- Econometric approach:
  - Multinomial probit / weighted ordered probit used to estimate effects on agent expectations.
  - Specification includes PERSONAL (personal characteristics), MACRO (growth, inflation, current account/GDP, EBRD Structural Reform index), IMFPRO (IMF program variables), country and time effects, individual effect ui, and time effects tη.
  - Robust clustered standard errors by country used to address intra-cluster correlation.
  - Endogeneity concerns addressed via lag structures and leading tests; macro variables also tested lagged.

### Data and estimation outline
- Data section labeled: A. Data.
- Econometric analysis section labeled: B. Econometric Analysis.
- Main results section labeled: C. Main Results, with subsections:
  - Effect of Personal Characteristics on Expectations,
  - Effect of Macro-Variables on Expectations,
  - Effect of IMF Programs on Expectations,
  - Robustness Test: Comparing 1990-93 with 1994-97.
- Survey data: Central and Eastern Eurobarometer (CEEB) 1990–1997.
  - Original CEEB responses: 125,875 individual responses across 22 countries.
  - After dropping “no answer” and some countries: about 102,000 observations remain for analysis.
  - Regressions use roughly 79,579–79,805 observations depending on specification.
- IMF program data source: IMF “Disbursements and Repayments” data (1990–1997).
  - Stated aggregate: “IMF had programs worth over $25 billion between 1990 and 1997.”
- Country groups: Central and Eastern European and Baltic States (CEEC) vs Commonwealth of Independent States (CIS).
- Dependent variable (survey question) coded:
  - 2 = get a lot better
  - 1 = get a little better
  - 0 = stay the same
  - -1 = get a little worse
  - -2 = get a lot worse

### Main empirical findings — determinants of expectations
- Individual characteristics (consistent effects)
  - Younger respondents, males, higher education, and self-employed respondents have higher probabilities of positive expectations.
  - Unemployed respondents have lower probabilities of positive expectations.
  - Urban residence associated with lower expectations relative to rural.
- Macroeconomic variables (generally small or insignificant effects)
  - GDP growth: small positive and statistically significant effect.
    - Probit marginal effect example: GDP growth coefficient 0.015 (***), marginal effect 0.005 (Table 2).
  - Inflation: minimal or statistically insignificant effect in most specifications.
  - Current account/GDP: small negative effect in simple models; generally not strongly significant.
  - EBRD Structural Adjustment index: negative and statistically significant short-run effect (structural reform variable coefficients e.g. -1.313 (***), marginal effect -0.453 in Table 2).
  - CIS dummy: negative and statistically significant (e.g., coefficient -0.545 (***), marginal effect -0.166 in Table 2), indicating lower expectations in CIS relative to CEEC.
- IMF program variables — measurement nuance and effects
  - IMF Dummy (presence of program) — mixed results:
    - Table 2 (simple probit): IMF Dummy coefficient 0.146 (***), marginal effect 0.050 (positive and significant).
    - Ordered probit (Table 3): IMF Dummy 0.045 (not significant), marginal effect 0.003.
    - Interpretation: IMF programs as a binary indicator do not uniformly raise public expectations across specifications.
  - IMF Inflow/GDP (size of IMF financing) — consistently positive and statistically significant:
    - Table 2: IMF Inflow/GDP coefficient 0.076 (***), marginal effect 0.026.
    - Table 3: IMF Inflow/GDP coefficient 0.034 (***), marginal effect 0.002 (score 2).
    - Interpretation: the amount of IMF financing as a share of GDP matters more than mere program presence.
  - IMF Serial Program (frequency of programs): negative effect on expectations:
    - Table 2 shows IMF Serial Program -0.133 (***), marginal effect -0.047 (probit).
    - Ordered probit Table 3: Dummy Serial Program -0.150 (**) marginal effect -0.012 (score 2).
    - Interpretation: repeated IMF programs reduce credibility and lower expectations.

### Timing and heterogeneity — crisis vs recovery
- Sample split: 1990–1993 (early transition, output collapse) vs 1994–1997 (recovery period).
- 1990–1993 (Table 4A)
  - IMF Dummy (signaling) becomes statistically significant and positive for those with optimistic outlooks in early crisis (IMF Dummy 0.148 (**) marginal effect 0.001).
  - IMF Inflow/GDP more strongly positive and significant (IMF Inflow/GDP 0.089 (***) marginal effect 0.007).
  - IMF Serial Program large negative effect (-0.365 (**) marginal effect -0.039).
  - Conclusion: during initial collapse, both signaling and financing channels operate and raise expectations for optimistic agents; but IMF programs reinforce prior beliefs—positive for optimists, negative for pessimists.
- 1994–1997 (Table 4B)
  - IMF Dummy and IMF Inflow/GDP lose statistical significance in most specifications (e.g., IMF Dummy -0.017 not significant; IMF Inflow/GDP -0.072 not significant in Table 4B).
  - IMF programs cease to have statistically significant effects on expectations once recovery is underway.
- Net interpretation: IMF support matters most in periods of high uncertainty and collapse (1990–1993), primarily through financing and signaling for optimistic agents; during recovery (1994–1997), IMF programs no longer significantly affect stated expectations.

### Quantitative model fit and sample sizes (selected statistics)
- Data and regression counts:
  - Original CEEB responses: 125,875.
  - Post-cleaning survey observations: about 102,000.
  - Regression observations reported: 79,805 and 79,579 in Table 2 and Table 3 respectively.
  - Subsample observations: 1990–1993: 24,314; 1994–1997: 55,491 (Tables 4A and 4B).
- Goodness-of-fit indicators:
  - Probit pseudo R2 in Table 2: 0.038 and 0.039.
  - Ordered probit pseudo R2 in Table 3: 0.0274 and 0.0277.
  - Ordered probit pseudo R2 in Table 4A (1990–1993): 0.02910 and 0.0312.
  - Ordered probit pseudo R2 in Table 4B (1994–1997): 0.03140 and 0.0321.
- Significance notation used throughout tables: (*) at 10% level, (**) at 5% level, and (***) at 1% level.

### Key policy implications and conclusions
- IMF programs can improve public expectations about the economy, but the effect depends on context and measurement:
  - Financing (IMF Inflow/GDP) matters: the size of IMF financing has a statistically significant positive effect on optimistic agents’ expectations, especially during periods of collapsing growth.
  - Signaling (IMF Dummy) matters primarily during acute crisis periods (1990–1993) for optimistic respondents.
  - Frequency undermines credibility: repeated IMF programs (serial programs) have a negative effect on expectations.
- Practical implication: IMF support is most effective at restoring confidence during high uncertainty and disorganization (e.g., immediate post-crisis periods or post-war), and primarily for agents predisposed to optimism; during recoveries, IMF programs are less influential on public expectations.
- Caveats: results are tentative due to potential survey biases, limited observables on individual traits, and the uniqueness of transition economies; findings may not generalize to other regions.

### Appendix 1: data sources and descriptions (selected)
- PERSONAL VARIABLES (Eurobarometer)
  - Confidence: Questionnaire asking people whether they are confident in the economy.
  - Age: 4 age groups are explicitly accounted: 1-20 years old.
  - Gender: Male and Female.
  - Extent of Formal Education: Individuals having Primary, Secondary or Tertiary Education.
  - Individual Employment Status: Employed or Unemployed.
- ECONOMIC VARIABLES
  - Economic Growth: IMF International Financial Statistics.
  - IMF Program: IMF.
  - Current Account: IMF.
  - EBRD Structural Reform Index: EBRD.

*Source: _wp07261 - Appendix 1 (CEEB survey analysis, 1990–1997).*

### 1997. Previous studies, in contrast, have looked at indirect measures, such as capital flows or

### _wp07261 - 1997. Previous studies, in contrast, have looked at indirect measures, such as capital flows or

### Summary findings on IMF programs and expectations
- Using a multinomial probit model, IMF loans appear to have a strong effect on agent expectations in the early years.
- Two transmission channels identified: 
  - inflow of real money, and 
  - signaling effect.
- IMF programs during periods of collapsing growth:
  - reinforce underlying expectations for the future;
  - are associated with positive expectations for those with an optimistic outlook and negative expectations for those with a negative outlook.
- Once recovery is underway and economic uncertainty diminishes, IMF programs cease to have a statistically significant effect on the expectations of economic agents.
- Implication: IMF programs have the biggest impact on expectations during periods of great uncertainty and less impact when countries face minor shocks.

### Conceptual and methodological themes
- Expectations modeling approaches discussed in the chapter include:
  - Background (general framing of expectations),
  - Adaptive Expectations,
  - Rational Expectations,
  - Bounded Rationality.
- Measurement issues:
  - Direct measurement of expectations via survey data emphasized as advantageous compared with indirect measures such as capital flows or yield spreads.
- Econometric approach:
  - Multinomial probit model used to estimate effects on agent expectations.

### Data and estimation outline (as presented)
- Data section labeled: A. Data.
- Econometric analysis section labeled: B. Econometric Analysis.
- Main results section labeled: C. Main Results, with subsections:
  - Effect of Personal Characteristics on Expectations,
  - Effect of Macro-Variables on Expectations,
  - Effect of IMF Programs on Expectations,
  - Robustness Test: Comparing 1990-93 with 1994-97.

### Ancillary information and metadata
- JEL Classification Numbers: (listed but not specified in provided excerpt).
- Keywords: Structural Adjustment, Transition Economies
- Author’s E-Mail Address: pimam@imf.org
- Acknowledgments note: Thanks to IMF Institute seminar participants and named individuals (Ales Bulir, Burkhard Drees, Jemma Dridi, Maher Hassan, Elizabeth Miranda and Juan Zalduendo); usual disclaimer applies.

*Source: _wp07261 - 1997. Previous studies, in contrast, have looked at indirect measures, such as capital flows or (IMF PDF chapter).*

### Appendix 1..............................................................................................................

### _wp07261 - Appendix 1

### I. Introduction and research question
- IMF role per Article I: “give confidence to members by making the general resources of the Fund temporarily available to them under adequate safeguards, thus providing them with opportunity to correct maladjustments in their balance of payments without resorting to measures destructive of national or international prosperity.”
- Research objective: assess the effect of IMF programs on domestic agents’ expectations about the future health of the economy using survey data (Central and Eastern Eurobarometer, 1990–1997), rather than indirect financial proxies.
- Time coverage: 1990–1997; focus on transition economies (CEEC and CIS).
- Key interpretation point: the paper analyzes effects on stated expectations (survey responses), not direct macroeconomic outcomes.

### II. Data and empirical design
- Survey data: Central and Eastern Eurobarometer (CEEB) 1990–1997.
  - Original CEEB responses: 125,875 individual responses across 22 countries.
  - After dropping “no answer” and some countries: about 102,000 observations remain for analysis.
  - Regressions use roughly 79,579–79,805 observations depending on specification (see tables).
- IMF program data source: IMF “Disbursements and Repayments” data (1990–1997).
  - Stated aggregate: “IMF had programs worth over $25 billion between 1990 and 1997.”
- Country groups: Central and Eastern European and Baltic States (CEEC) vs Commonwealth of Independent States (CIS).
- Key survey question used for dependent variable: expected change over next 12 months in the household’s financial situation, coded:
  - 2 = get a lot better
  - 1 = get a little better
  - 0 = stay the same
  - -1 = get a little worse
  - -2 = get a lot worse
- Econometric model:
  - Weighted ordered probit / multinomial-ordered probit with personal characteristics (PERSONAL), macro variables (MACRO: growth, inflation, current account/GDP, EBRD Structural Reform index), IMF program variables (IMFPRO), country and time effects.
  - Individual effect ui and time effects tη included.
  - Robust clustered standard errors by country to address intra-cluster correlation.
  - Addressed endogeneity concerns via lag structures and leading tests; macro variables also tested lagged.

### III. Main empirical findings — determinants of expectations
- Individual characteristics (consistent effects)
  - Younger respondents, males, higher education, and self-employed respondents have higher probabilities of positive expectations.
  - Unemployed respondents have lower probabilities of positive expectations.
  - Urban residence associated with lower expectations relative to rural.
- Macroeconomic variables (generally small or insignificant effects)
  - GDP growth: small positive and statistically significant effect.
    - Probit marginal effect example: GDP growth coefficient 0.015 (***), marginal effect 0.005 (Table 2).
  - Inflation: minimal or statistically insignificant effect in most specifications.
  - Current account/GDP: small negative effect in simple models; generally not strongly significant.
  - EBRD Structural Adjustment index: negative and statistically significant short-run effect (structural reform variable coefficients e.g. -1.313 (***), marginal effect -0.453 in Table 2).
  - CIS dummy: negative and statistically significant (e.g., coefficient -0.545 (***), marginal effect -0.166 in Table 2), indicating lower expectations in CIS relative to CEEC.
- IMF program variables — overall and nuance by measurement
  - IMF Dummy (presence of program) — mixed results:
    - Table 2 (simple probit): IMF Dummy coefficient 0.146 (***), marginal effect 0.050 (positive and significant).
    - Ordered probit (Table 3): IMF Dummy 0.045 (not significant), marginal effect 0.003.
    - Interpretation: IMF programs as a binary indicator do not uniformly raise public expectations across specifications.
  - IMF Inflow/GDP (size of IMF financing) — consistently positive and statistically significant:
    - Table 2: IMF Inflow/GDP coefficient 0.076 (***), marginal effect 0.026.
    - Table 3: IMF Inflow/GDP coefficient 0.034 (***), marginal effect 0.002 (score 2).
    - Interpretation: the amount of IMF financing as a share of GDP matters more than mere program presence.
  - IMF Serial Program (frequency of programs): negative effect on expectations:
    - Table 2 shows IMF Serial Program -0.133 (***), marginal effect -0.047 (probit).
    - Ordered probit Table 3: Dummy Serial Program -0.150 (**) marginal effect -0.012 (score 2).
    - Interpretation: repeated IMF programs reduce credibility and lower expectations.

### IV. Timing and heterogeneity — crisis vs recovery (robustness split)
- Sample split: 1990–1993 (early transition, output collapse) vs 1994–1997 (recovery period).
- 1990–1993 (Table 4A)
  - IMF Dummy (signaling) becomes statistically significant and positive for those with optimistic outlooks in early crisis (IMF Dummy 0.148 (**) marginal effect 0.001).
  - IMF Inflow/GDP more strongly positive and significant (IMF Inflow/GDP 0.089 (***) marginal effect 0.007).
  - IMF Serial Program large negative effect (-0.365 (**) marginal effect -0.039).
  - Conclusion: during initial collapse, both signaling and financing channels operate and raise expectations for optimistic agents; but IMF programs reinforce prior beliefs—positive for optimists, negative for pessimists.
- 1994–1997 (Table 4B)
  - IMF Dummy and IMF Inflow/GDP lose statistical significance in most specifications (e.g., IMF Dummy -0.017 not significant; IMF Inflow/GDP -0.072 not significant in Table 4B).
  - IMF programs cease to have statistically significant effects on expectations once recovery is underway.
- Net interpretation: IMF support matters most in periods of high uncertainty and collapse (1990–1993), primarily through financing and signaling for optimistic agents; during recovery (1994–1997), IMF programs no longer significantly affect stated expectations.

### V. Quantitative model fit and sample sizes (selected statistics)
- Data and regression counts:
  - Original CEEB responses: 125,875.
  - Post-cleaning survey observations: about 102,000.
  - Regression observations reported: 79,805 and 79,579 in Table 2 and Table 3 respectively.
  - Subsample observations: 1990–1993: 24,314; 1994–1997: 55,491 (Tables 4A and 4B).
- Goodness-of-fit indicators:
  - Probit pseudo R2 in Table 2: 0.038 and 0.039.
  - Ordered probit pseudo R2 in Table 3: 0.0274 and 0.0277.
  - Ordered probit pseudo R2 in Table 4A (1990–1993): 0.02910 and 0.0312.
  - Ordered probit pseudo R2 in Table 4B (1994–1997): 0.03140 and 0.0321.
- Significance notation used throughout tables: (*) at 10% level, (**) at 5% level, and (***) at 1% level.

### VI. Key policy implications and conclusions
- IMF programs can improve public expectations about the economy, but the effect depends on context and measurement:
  - Financing (IMF Inflow/GDP) matters: the size of IMF financing has a statistically significant positive effect on optimistic agents’ expectations, especially during periods of collapsing growth.
  - Signaling (IMF Dummy) matters primarily during acute crisis periods (1990–1993) for optimistic respondents.
  - Frequency undermines credibility: repeated IMF programs (serial programs) have a negative effect on expectations.
- Practical implication: IMF support is most effective at restoring confidence during high uncertainty and disorganization (e.g., immediate post-crisis periods or post-war), and primarily for agents predisposed to optimism; during recoveries, IMF programs are less influential on public expectations.
- Caveats: results are tentative due to potential survey biases, limited observables on individual traits, and the uniqueness of transition economies; findings may not generalize to other regions.

*Source: Appendix 1 of the working paper (CEEB survey analysis, 1990–1997).*

### APPENDIX 1

### APPENDIX 1

### APPENDIX 1: DATA SOURCES AND DESCRIPTIONS

### DATADATA DESCRIPTIONSOURCE

### PERSONAL VARIABLES
- Confidence  
  - Questionnaire asking people whether they are confident in the economyEurobarometer
- Age  
  - 4 age groups are explicitly accounted: 1-20 years oldEurobarometer
- Gender  
  - Male and FemaleEurobarometer
- Extent of Formal Education  
  - Indidivuals having Primary, Secondary or Tertiary EducationEurobarometer
- Individual Employment Status  
  - Employed or UnemployedEurobarometer

### ECONOMIC VARIBALES
- Economic Growth  
  - IMF International Financial Statistics
- IMF Program  
  - IMF
- Current Account  
  - IMF
- EBRD Structural Reform Index  
  - EBRD

*APPENDIX 1: DATA SOURCES AND DESCRIPTIONS*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2007/_wp07261.pdf_