## _wp1631

## Source details

**Canonical URL:** [_wp1631](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1631.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1631.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1631.pdf.json)

---

### I. Introduction — context and framing
- “Western Balkans States” (WBS) refers to Albania, Bosnia and Herzegovina, Croatia, FYR of Macedonia, Montenegro, and Serbia; Kosovo analyzed separately where data allow.
- Emerging Europe experienced a ‘U-shaped’ transition path; the Western Balkans followed a more uneven ‘W-shape’ with recovery, reversal, and recovery again (Sanfey and Cviić, 2010).
- Key observations:
  - Even in 2006, over half of the population in the Western Balkans thought they were worse off than in 1989; 11 percent thought they were better off.
  - Real GDP per capita changes between 1989 and 2006:
    - roughly doubled in Croatia and Kosovo;
    - increased less than 1.5-fold (if at all) in Albania, Bosnia and Herzegovina, Macedonia and Montenegro;
    - was still below its 1989 level in Serbia in 2006.
- Post-1990s reforms: opening to trade, greater export orientation, expanded private sector role, deregulation, institution building — yielding robust growth and improved macroeconomic stability, but structural transformation stalled in the mid-2000s (Murgasova et al., 2015).
- Research objective: analyze evolution and determinants of inequality and poverty in the Western Balkans, complementing macro-level measures with household-level analysis of subjective perceptions of poverty.

### II. Literature review — subjective versus objective poverty measurement
- Limitations of standard objective measures:
  - Aggregation shortcomings of the Gini coefficient and headcount measures.
  - Expenditure/income measures require normalization for cost-of-living and household size; small changes in assumptions can have large effects (e.g., equivalence scales, PPP difficulties).
  - Non-market goods (access to public services) are often excluded; composite indices face ad hoc weighting across dimensions.
- Subjective measures:
  - Respondents’ self-assessments can complement objective measures.
  - Empirical findings highlighted:
    - Subjective poverty lines typically yield (much) higher headcounts than objective measures; results sensitive to survey question framing and whose responses are aggregated.
    - Determinants of subjective poverty perceptions include income/expenditure plus household size, employment status, farm income, assets (home ownership, durable goods), age, and health status.
    - Prior Western Balkans subjective assessment: Carletto and Zezza (2004) for Albania — employment status, satisfaction at work, household wealth, relative wealth and vulnerability affected perceived welfare even controlling for income; evidence of economies of scale effects.

### III. Trends in inequality and poverty — stylized facts and recent dynamics
- 1990s: transition across Emerging Europe with marked heterogeneity; Western Balkans lagged Central and South Eastern Europe on several transition indicators.
- Early 2000s: steady income increases, sharp declines in poverty (headcount and poverty gap), but mean consumption remained far below Central and Eastern Europe.
- Inequality dynamics:
  - Rapid growth in the early 2000s brought large increases in inequality measured by the Gini index, driven by increases at the top.
  - Inequality continued to increase until 2005 across the region.
  - After 2005, the share of the top decile declined in Albania, Bosnia and Herzegovina and Serbia and to a lesser extent in Montenegro; the share continued to increase in Croatia.
- Measurement note:
  - The headcount is defined as the percent of the population living in households with consumption per person below the poverty line.
  - The poverty line used in this section is USD 38 per month, corresponding to the World Bank USD 38 per month threshold.

### IV. Analytical approach and scope
- Methodology overview:
  - Combines macro-level time series analysis of country determinants of poverty and inequality (government, taxation, social safety nets, education policies, privatization) with micro-level cross-sectional household survey analysis to study subjective perceptions.
  - Micro-level analysis examines whether objectively poor households feel poor; assesses how non-income characteristics influence poverty perceptions; compares poverty rates under objective and subjective measures.
- Data sources:
  - World Bank Povcal database (household consumption surveys converted to 2005 PPPs), EBRD Transition Indicators, World Economic Outlook, World Development Indicators, Life in Transition Surveys (LITS) 2006 and 2010.
- Empirical strategy:
  - Macro-level: pooled panel regressions and fixed effects estimators for a joint sample of the Western Balkans countries and the New Member States (16 countries, data not available separately for Kosovo), covering years 2002-2011. Dummy variables for Western Balkans included.
  - Micro-level: household-level analysis using LITS (target ~1,000 households per country; 2006 round used primarily because the ‘minimum income’ question was asked then).

### V. Macro-level findings (inequality, poverty and drivers)
- General patterns:
  - Poverty fell across the Western Balkans up to the financial crisis; the crisis brought an increase in poverty with substantial cross-country variation.
  - Inequality fell as a result of the crisis, mostly driven by a decline in incomes at the top.
- Key empirical results (Western Balkans and New Member States, 2002-2011):
  - Lagged level of income: higher income associated with lower inequality and lower poverty; however, once government spending is controlled for, the level of income is no longer significant.
  - Lagged growth: increases the share of the top quintile but reduces poverty.
  - Privatizations:
    - Large-scale privatizations initially appear to increase inequality and poverty; this effect operates in part through increases in unemployment.
    - Small-scale privatizations, once unemployment is controlled for, appear pro-poor.
  - Unemployment:
    - Strong positive correlation between unemployment rate and inequality.
    - Unemployment increases inequality and the share of the top, and reduces the share of the bottom.
  - Government spending composition:
    - Goods and services expenditure is significantly pro-poor: reduces inequality, reduces the share of the top quintile and reduces poverty.
    - Social security expenditure is not found to have a significant effect on inequality or poverty in the cross-country regressions.
    - Capital expenditure appears to increase inequality and the share of the top in contemporaneous regressions.
  - Flat income tax rate: mixed effects—reduces the share of the top but also reduces the share of the lowest decile and increases poverty.
  - Urbanization: larger urban population increases inequality, increases the share of the top, and reduces the share of the bottom.

### VI. Micro-level findings (subjective poverty, perceptions and vulnerability)
- Poverty benchmarks and headcounts:
  - Main poverty benchmark used: 1.25 USD per day (in 2005 PPP).
  - National poverty rates are “considerably higher, in the order of magnitude of 15-30 percent.”
- Comparison of subjective vs objective poverty:
  - Among households objectively measured as poor (bottom two deciles by income), 38 percent also self-assess as poor.
  - Among the 62 percent of objectively poor households that do not self-assess as poor, 96 percent have household income below their reported ‘minimum income’.
  - Reported ‘minimum income’ is higher for richer deciles; subjective poverty lines are higher than official poverty lines and much higher than the World Bank’s extreme 1.25 USD per day line.
  - Headcounts using subjective ‘minimum income’ lines range by country from 27 percent in Croatia to 92 percent in Albania.
- Determinants of subjective poverty (regression evidence):
  - Income (household expenditure) significantly reduces the probability of feeling poor and increases self-assessed relative standing; it also increases reported ‘minimum income’.
  - Other significant factors (holding income constant): household size, age, health status, education, employment status, assets (owning a house), farm income, feelings of uncertainty and expectations about future welfare (chance of reemployment).
  - Specific patterns:
    - Households owning a house or with farm income are less likely to feel poor.
    - Better health, higher education, and employment reduce subjective poverty and are associated with higher reported minimum income.
    - Unemployed respondents feel poorer and report higher ‘minimum income’; among the unemployed, higher perceived chance of reemployment reduces the likelihood of feeling poor.
  - Urban-rural differences: urban residents report needing higher income; in Central and South Eastern Europe urban residents are less likely to report feeling poor, while in the Western Balkans this protective effect of urban residence is much smaller and not statistically significant.
- Explaining variance:
  - Regressions leave over 80 percent of variation unexplained for poverty perceptions and over 50 percent unexplained for ‘minimum income’.
  - Unexplained variation consistent with measurement error, idiosyncratic responses, moods, personality, and omitted factors.
- Crisis impacts and coping:
  - Households in the Western Balkans were hit harder by the crisis than those in Central and South Eastern Europe—lower/delayed wages, lower remittances, unemployment.
  - Poorer households were more likely to cut staple consumption, health and utility spending, and withdraw from education.
  - Vulnerability and inability to cope with shocks significantly affect subjective poverty perceptions.

### VII. Key statistics and country variation (selected)
- Poverty benchmark: 1.25 USD per day (2005 PPP).
- National poverty lines typically “in the order of magnitude of 15-30 percent.”
- Macro panel: 16 countries, years 2002-2011.
- Among objectively poor households (bottom two deciles), 38 percent self-assess as poor; 96 percent of objectively poor-but-not-self-assessed-poor have income below their reported minimum income.
- Subjective headcounts (using reported minimum income) range across countries from 27 percent to 92 percent (country examples: 27 percent in Croatia; 92 percent in Albania).
- Regression explanatory power: over 80 percent of variation unexplained for poverty perceptions; over 50 percent unexplained for minimum income.

### VIII. Policy implications and recommendations
- Social safety nets:
  - Improve targeting and extend coverage of the poor; shifting from categorical to means-tested benefits could often improve targeting at no additional fiscal cost.
  - Reform targeting and access to address low coverage rates and regressive targeting.
- Labor market policies:
  - Enhance labor market institutions to raise employment and participation rates; unemployment is a key driver of rising inequality and household vulnerability.
  - Consider sequencing and communication of reforms carefully, as expectations and uncertainty shape subjective perceptions and political economy dynamics.
- Public spending composition:
  - Prioritize goods and services expenditure that is significantly pro-poor.
  - Recognize that short-run regressions may underestimate longer-term pro-poor benefits of capital expenditure (infrastructure).
- Privatization and transition policy:
  - Mitigate short-term adverse distributional effects of large-scale privatizations through measures addressing unemployment.
  - Support small-scale privatization and private sector development conditional on labor market absorption.
- Regional and demographic focus:
  - Address urban-rural divides and regional pockets of poverty; pay attention to demographic pressures and the regional prominence of poverty.

### IX. Annex highlights — household correlates, regressions, safety nets, and coping
- Household characteristics and regressions (LITS 2006 and 2010):
  - Log household expenditure: consistently highly statistically significant negative coefficients in subjective-poverty regressions (examples: -0.54***, -0.68***, -0.58***, -0.50***, -0.70***, -0.51***).
  - Self-reported health: large negative and significant coefficients (examples: -0.27***, -0.25**, -0.27***; -0.31***, -0.37***, -0.24***).
  - Highest degree (1-5): negative and significant (examples: -0.13***, -0.10, -0.14***; -0.14***, -0.19*, -0.12***).
  - Number of children often positively associated with subjective poverty (examples: 0.12**, 0.17*, 0.13**; 0.13**, 0.18, 0.11).
  - Country indicators show marked differences (examples from Table A.3: Croatia 0.75***, Macedonia 0.49***, Serbia 0.89***).
  - Number of observations examples: 2,210; 4,630; 1,840; 2,195; 4,600; 1,833. Pseudo/Adjusted R-squared examples: 0.169, 0.248, 0.177, 0.071, 0.090, 0.074.
- Objective vs subjective poverty lines and minimum income perceptions:
  - Figures compare household expenditure, reported minimum income needed to make ends meet, official poverty line, and 1.25/day in PPP across Western Balkans countries.
- Crisis impacts and coping mechanisms:
  - Shocks reported include reduced wages, lower remittances, delayed wages, reduced hours, job loss.
  - Coping mechanisms include increased hours, second job/additional work, finding new job.
  - Consumption- and expenditure-related coping strategies include reduced consumption of staples, health-related saving (postponed/skipped doctor visits, stopped buying medications), postponed/withdrew from education, delayed/defaulted on loan installment, sold an asset.
- Social safety nets (Annex 2):
  - Two reform approaches: evolutionary reforms (Bosnia and Herzegovina, Macedonia, Montenegro, Serbia) and full overhaul (Albania, Kosovo).
  - Non-contributory programs: last-resort social assistance, family and child protection benefits, disability benefits for those without disability-insurance, region-specific benefits for war veterans and their families.
  - Coverage and targeting: public spending comparable to New Member States, but a growing share allocated to categorical programs; coverage of the poorest quintile is low.
  - Benefit generosity: generally between 7 and 20 percent of the minimum wage in most countries; exception: Montenegro where the social assistance transfer is about 75 percent of the minimum wage.
  - Pensions: significant in reducing poverty despite being very low and close to absolute national poverty lines.

*IMF staff summary based on the provided content unit.*

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

### _wp1631 - References .............................................................................................................

### I. Introduction — context and framing
- Emerging Europe experienced a ‘U-shaped’ transition path; the Western Balkans followed a more uneven ‘W-shape’ with recovery, reversal, and recovery again (Sanfey and Cviić, 2010).
- “Western Balkans States” (WBS) refers to Albania, Bosnia and Herzegovina, Croatia, FYR of Macedonia, Montenegro, and Serbia; Kosovo analyzed separately where data allow.
- Key observations:
  - Even in 2006, over half of the population in the Western Balkans thought they were worse off than in 1989; 11 percent thought they were better off (compared with around 35 percent and about a third, respectively, in the New Member States).
  - Real GDP per capita changes between 1989 and 2006:
    - roughly doubled in Croatia and Kosovo;
    - increased less than 1.5-fold (if at all) in Albania, Bosnia and Herzegovina, Macedonia and Montenegro;
    - was still below its 1989 level in Serbia in 2006.
- Post-1990s reforms: opening to trade, greater export orientation, expanded private sector role, deregulation, institution building — yielding robust growth and improved macroeconomic stability, but structural transformation stalled in the mid-2000s (Murgasova et al., 2015).
- Research objective: analyze evolution and determinants of inequality and poverty in the Western Balkans, complementing macro-level measures with household-level analysis of subjective perceptions of poverty.

### II. Literature review — subjective versus objective poverty measurement
- Limitations of standard objective measures:
  - Aggregation shortcomings of the Gini coefficient and headcount measures.
  - Expenditure/income measures require normalization for cost-of-living and household size; small changes in assumptions can have large effects (e.g., equivalence scales, PPP difficulties).
  - Non-market goods (access to public services) are often excluded; composite indices face ad hoc weighting across dimensions (examples: Morris’s Physical Quality of Life Index, HDI, MPI).
- Subjective measures:
  - Originated in psychology and economics; respondents’ self-assessments can complement objective measures.
  - Empirical findings from previous studies:
    - Subjective poverty lines typically yield (much) higher headcounts than objective measures; results sensitive to survey question framing and whose responses are aggregated.
    - Determinants of subjective poverty perceptions include income/expenditure plus household size, employment status, farm income, assets (home ownership, durable goods), age, and health status.
    - Carletto and Zezza (2004) is noted as the only prior subjective poverty assessment in the Western Balkans (Albania): employment status, satisfaction at work, household wealth, relative wealth and vulnerability each affected perceived welfare, even controlling for income; evidence of economies of scale effects (single-person households: highest subjective poverty incidence but lowest objective poverty incidence).
- Two caveats stated:
  - Objective and subjective measures capture different concepts and are complements, not substitutes.
  - Data constraints limit the analysis of the direct effects of wars in the 1990s.

### III. Trends in inequality and poverty — stylized facts and recent dynamics
- 1990s: transition across Emerging Europe with marked heterogeneity; Western Balkans lagged Central and South Eastern Europe on several transition indicators.
- Early 2000s: steady income increases, sharp declines in poverty (headcount and poverty gap), but mean consumption remained far below Central and Eastern Europe.
- Inequality dynamics:
  - Rapid growth in the early 2000s brought large increases in inequality measured by the Gini index, driven by increases at the top.
  - Inequality continued to increase until 2005 across the region.
  - After 2005, the share of the top decile declined in Albania, Bosnia and Herzegovina and Serbia and to a lesser extent in Montenegro; the share continued to increase in Croatia.
- Measurement note:
  - The headcount is defined as the percent of the population living in households with consumption per person below the poverty line.
  - The poverty line used in this section is USD 38 per month, corresponding to the World Bank USD 38 per month threshold.

### IV. Analytical approach and scope
- Methodology overview:
  - Combines macro-level time series analysis of country determinants of poverty and inequality (government, taxation, social safety nets, education policies, privatization) with micro-level cross-sectional household survey analysis to study subjective perceptions.
  - Micro-level analysis examines whether objectively poor households feel poor; assesses how non-income characteristics influence poverty perceptions; compares poverty rates under objective and subjective measures.
- Purpose:
  - Provide insights for policymaking on targeting scarce resources and understanding political economy constraints associated with persistent dissatisfaction and perceptions of under-delivered reforms in the Western Balkans.

*Source: _wp1631 - References .............................................................................................................*

### 1.25  per  day  extreme  poverty  line  (in  2005  PPP,  as  proposed  by  Ravallion,  Chen  and  Sangraula,  2009).  Th

### _wp1631 - 1.25 per day extreme poverty line (in 2005 PPP, as proposed by Ravallion, Chen and Sangraula, 2009). Th

### Poverty measures, data and methodology
- Main poverty benchmark used: 1.25 USD per day (in 2005 PPP).
- National poverty rates are “considerably higher, in the order of magnitude of 15-30 percent.”
- Data sources: World Bank Povcal database (household consumption surveys converted to 2005 PPPs), EBRD Transition Indicators, World Economic Outlook, World Development Indicators, and the Life in Transition Surveys (LITS) 2006 and 2010.
- Empirical strategy:
  - Macro-level: pooled panel regressions and fixed effects estimators for a joint sample of the Western Balkans countries and the New Member States (16 countries, data not available separately for Kosovo), covering years 2002-2011. Dummy variables for Western Balkans included.
  - Micro-level: household-level analysis using LITS (target ~1,000 households per country; 2006 round used primarily because the ‘minimum income’ question was asked then).

### Macro-level findings (inequality, poverty and drivers)
- General patterns:
  - Poverty fell across the Western Balkans up to the financial crisis; the crisis brought an increase in poverty with substantial cross-country variation.
  - Inequality fell as a result of the crisis, mostly driven by a decline in incomes at the top.
- Key empirical results from regressions (Western Balkans and New Member States, 2002-2011):
  - Lagged level of income: higher income associated with lower inequality and lower poverty; however, once government spending is controlled for, the level of income is no longer significant.
  - Lagged growth: increases the share of the top quintile but reduces poverty.
  - Privatizations:
    - Large-scale privatizations initially appear to increase inequality and poverty; this effect operates in part through increases in unemployment.
    - Small-scale privatizations, once unemployment is controlled for, appear pro-poor (possibly reflecting entrepreneurship and private sector development).
  - Unemployment:
    - Strong positive correlation between unemployment rate and inequality (Figure 4).
    - Unemployment increases inequality and the share of the top, and reduces the share of the bottom—suggesting the poor are more likely to become unemployed and/or that social safety nets provide limited insulation.
  - Government spending composition:
    - Goods and services expenditure is significantly pro-poor: reduces inequality, reduces the share of the top quintile and reduces poverty.
    - Social security expenditure is not found to have a significant effect on inequality or poverty in the cross-country regressions (possible explanations: higher spending where poverty/unemployment is higher, low coverage rates, regressive targeting).
    - Capital expenditure appears to increase inequality and the share of the top in contemporaneous regressions (benefits for the poor may be underestimated due to lags and confounding by infrastructure gaps).
  - Flat income tax rate: mixed effects—reduces the share of the top but also reduces the share of the lowest decile and increases poverty (consistent with a potentially regressive impact).
  - Urbanization: larger urban population increases inequality, increases the share of the top, and reduces the share of the bottom.

### Micro-level findings (subjective poverty, perceptions and vulnerability)
- Subjective measures used:
  - Ten-step ‘economic ladder’ question (self-placement 1-10).
  - ‘Minimum income’ question: “what would be the minimum amount of money that this household would need to make ends meet at the end of each month?”
- Comparison of subjective vs objective poverty:
  - Among households objectively measured as poor (bottom two deciles by income), 38 percent also self-assess as poor.
  - Among the 62 percent of objectively poor households that do not self-assess as poor, 96 percent have household income below their reported ‘minimum income’.
  - Reported ‘minimum income’ is higher for richer deciles; subjective poverty lines are higher than official poverty lines and much higher than the World Bank’s extreme 1.25 USD per day line.
  - Headcounts using subjective ‘minimum income’ lines vary widely by country (range cited: 27 percent in Croatia to 92 percent in Albania).
- Determinants of subjective poverty (regression evidence):
  - Income (household expenditure) significantly reduces the probability of feeling poor and increases self-assessed relative standing; it also increases reported ‘minimum income’.
  - Other significant factors (holding income constant): household size, age, health status, education, employment status, assets (owning a house), farm income, feelings of uncertainty and expectations about future welfare (chance of reemployment).
  - Specific patterns:
    - Households owning a house or with farm income are less likely to feel poor (may reflect imputed value of home production or access to land as a buffer).
    - Better health, higher education, and employment reduce subjective poverty (and are associated with higher reported minimum income).
    - Unemployed respondents feel poorer and report higher ‘minimum income’; among the unemployed, higher perceived chance of reemployment reduces the likelihood of feeling poor.
  - Urban-rural differences: urban residents report needing higher income; in Central and South Eastern Europe urban residents are less likely to report feeling poor, while in the Western Balkans this protective effect of urban residence is much smaller and not statistically significant.
- Explaining variance:
  - Regressions leave over 80 percent of variation unexplained for poverty perceptions and over 50 percent unexplained for ‘minimum income’—consistent with measurement error, idiosyncratic responses, moods, personality, and omitted factors. Subjective measures should complement, not substitute for, objective measures.
- Crisis impacts and coping:
  - Households in the Western Balkans were hit harder by the crisis than those in Central and South Eastern Europe—lower/delayed wages, lower remittances, unemployment.
  - Poorer households were more likely to cut staple consumption, health and utility spending, and withdraw from education.
  - Vulnerability and inability to cope with shocks significantly affect subjective poverty perceptions.

### Key statistics and country variation (selected)
- Poverty benchmark: 1.25 USD per day (2005 PPP).
- National poverty lines typically “in the order of magnitude of 15-30 percent.”
- Macro panel: 16 countries, years 2002-2011.
- Among objectively poor households (bottom two deciles), 38 percent self-assess as poor; 96 percent of objectively poor-but-not-self-assessed-poor have income below their reported minimum income.
- Subjective headcounts (using reported minimum income) range across countries from 27 percent to 92 percent (country examples: 27 percent in Croatia; 92 percent in Albania).
- Regression explanatory power: over 80 percent of variation unexplained for poverty perceptions; over 50 percent unexplained for minimum income.

### Policy implications and recommendations
- Social safety nets:
  - Improve targeting and extend coverage of the poor; poorly targeted social assistance could often be improved at no additional fiscal cost by shifting from categorical to means-tested benefits.
  - Low coverage rates and regressive targeting may explain the weak measured effectiveness of social security expenditure—reforming targeting and access is important.
- Labor market policies:
  - Enhance labor market institutions to raise employment and participation rates; unemployment is a key driver of rising inequality and household vulnerability.
  - Consider sequencing and communication of reforms carefully, as expectations and uncertainty significantly shape subjective perceptions and political economy dynamics.
- Public spending composition:
  - Prioritize goods and services expenditure that acts as stimulus and is significantly pro-poor.
  - Recognize that the short-run regressions may underestimate longer-term pro-poor benefits of capital expenditure (infrastructure).
- Privatization and transition policy:
  - Acknowledge distributional effects of large-scale privatizations through unemployment; support measures to mitigate adverse short-term impacts on the poor.
  - Small-scale privatization and private sector development can be pro-poor, conditional on labor market absorption.
- Regional and demographic focus:
  - Address urban-rural divides and regional pockets of poverty; pay attention to demographic pressures and the regional prominence of poverty.

*Italic: IMF staff summary based on the provided content unit.*

### ANNEX 1

### _wp1631 - ANNEX 1

### Poverty perceptions and household characteristics (Western Balkans, 2006–2010)
- Table A.1 and Table A.2 report mean household characteristics and comparisons across groups who are objectively poor (household expenditure in bottom two deciles) and/or subjectively feel poor.
- Key household-level correlates reported (2006 and 2010 regressions and means):
  - Household expenditure reported as median (brackets denote interquartile/adjacent values in table).
  - Perceived relative standing in bottom 2 deciles reported as fractions (examples: 0.17, 0.18, 0.15, 0.18, 0.18, 0.22).
  - Ownership of a house: values like 0.94, 0.90, 0.89, 0.94, 0.84, 0.87.
  - Farming as important source of income: values like 0.14, 0.09, 0.06, 0.12, 0.06, 0.12.
  - No. of adults in household and children, age, self-assessed health, highest degree (1-5), employment in last year, urban residence, chance of finding job, attitudes such as "Should reduce inequality" and "Prefers market economy" are tabulated as means.
- Table A.3 and Table A.4 present multivariate determinants of subjective poverty perceptions:
  - Log household expenditure: consistently highly statistically significant negative coefficients (examples: -0.54***, -0.68***, -0.58***, -0.50***, -0.70***, -0.51***).
  - Self-reported health: large negative and significant coefficients for feeling poor (examples: -0.27***, -0.25**, -0.27***; -0.31***, -0.37***, -0.24***).
  - Highest degree (1-5): negative and significant (examples: -0.13***, -0.10, -0.14***; -0.14***, -0.19*, -0.12***).
  - Number of adults and children in household: mixed effects; number of children often positively associated with subjective poverty (examples: 0.12**, 0.17*, 0.13**; 0.13**, 0.18, 0.11).
  - Urban residence, employment in last year, marital status categories, and country fixed effects show statistically significant variation in subjective-poverty outcomes.
  - Country indicators show marked differences (examples from Table A.3: Croatia 0.75***, Macedonia 0.49***, Serbia 0.89***; in the subjective-feeling columns these country indicators often show large negative coefficients for "feeling poor" regressions).
  - Number of observations and model fit reported: Number of obs. examples 2,210; 4,630; 1,840; 2,195; 4,600; 1,833 and Pseudo/Adjusted R-squared values reported (examples: 0.169, 0.248, 0.177, 0.071, 0.090, 0.074).

### Objective vs subjective poverty lines and minimum income perceptions (2006)
- Figure A.2 compares for each Western Balkans country (Albania, Bosnia and Herzegovina, Croatia, Macedonia, Montenegro, Serbia) distributions of:
  - household expenditure,
  - "minimum income needed to make ends meet",
  - "minimum income needed to make ends meet, conditional",
  - official poverty line,
  - 1.25/day in PPP.
- Respondents' reported "minimum income required to make ends meet" in 2006 is tabulated (example aggregate numbers shown in tables and figures of the annex).

### Crisis impacts and coping mechanisms (Western Balkans and Central and South Eastern Europe)
- Figure A.3 displays incidence (by decile) of shocks and coping mechanisms:
  - Shocks reported include: wages reduced, lower remittances, wages delayed/suspended, working hours reduced, household head lost job, increased hours.
  - Coping mechanisms include: increased hours, second job/additional work, found new job.
  - Consumption- and expenditure-related coping strategies reported include: reduced consumption of luxury goods, reduced consumption of alcohol/tobacco, reduced consumption of staples, reduced vacations, utilities-related saving (including delaying payments or having utilities cut), reduced use of own car, health-related saving (postponed/skipped doctor visits, stopped buying medications, cancelled health services), cut TV/phone/internet, postponed/withdrew from education, delayed/defaulted on loan installment, sold an asset.
- Comparative lines show responses for Western Balkans and Central and South Eastern Europe; figures present relative frequencies by decile and region.

### Social safety nets in the Western Balkans (Annex 2)
- Two broad types of safety-net reform approaches in the region:
  - Evolutionary reforms (Bosnia and Herzegovina, Macedonia, Montenegro, Serbia): retained and reformed pre-transition systems; systems more generous in scope of risks but fragmented with multiple categorical benefits that undermine overall efficiency.
  - Full overhaul (Albania, Kosovo): introduced entirely new benefits with fewer types and less fragmentation, but lower levels of protection, especially for families with children and the unemployed.
- Non-contributory social assistance program categories (four broad categories):
  - Last-resort social assistance programs: mostly targeted to the chronic poor, less capacity to identify and protect the transient poor.
  - Family and child protection benefits: multiple objectives (alleviate poverty, increase fertility); Albania and Kosovo have no stand-alone child benefits; other countries restrict child allowances via rigorous income tests (Bosnia and Herzegovina, Serbia) or a means test (Montenegro).
  - Disability benefits for those without disability-insurance: Western Balkans spend more on disability allowances than the New Member States, with Albania having the highest spending.
  - Region-specific benefits for war veterans and their families: highly regressive; particularly large share in Bosnia and Herzegovina, sizeable in Kosovo, Montenegro, and Serbia, and increasing across the region.
- Coverage and targeting:
  - Public spending on social assistance in the Western Balkans is comparable to the New Member States.
  - A growing share of allocations is going to categorical programs rather than means-tested benefits.
  - Coverage of the poorest quintile with all types of social benefits is low in all Western Balkan countries.
- Generosity:
  - Benefit generosity varies significantly; generally low compared to minimum wages:
    - Ranges between 7 and 20 percent of the minimum wage level in most countries.
    - Exception: Montenegro where the social assistance transfer is about 75 percent of the minimum wage.
  - Household-level perceptions: respondents generally perceive financial support as too small to live off and insufficient relative to shocks like job loss. For some programs (e.g., energy benefits) many poor households did not apply because application costs were high relative to potential rewards.
  - Role of safety nets in helping poor households handle shocks appeared relatively minor based on qualitative descriptions (Turk, 2013).
- Pensions:
  - Pensions have a significant impact in reducing poverty and have retained universal coverage roles.
  - Pensions are generally very low, close to absolute national poverty lines.
  - Example: in Serbia during 2008-2011 minimum pensions helped reduce poverty among regular pensioners; poverty among pensioners was somewhat lower than nationwide poverty, while those over 65 not entitled to pension benefits were much more vulnerable (Nestorovic, 2013).

### Additional methodological and data notes in the annex
- Source datasets used include LITS 2006, LITS 2010 and IMF staff calculations.
- Regression tables report robust standard errors in parentheses; significance denoted by * (5% level), ** (1% level), *** (0.1% level).
- Figures and tables disaggregate objective and subjective poverty measures and report respondents' self-assessed minimum income thresholds and perceptions of relative standing.

*Source: LITS 2006 and LITS 2010 data and IMF staff calculations, as presented in ANNEX 1 of the source PDF.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2016/_wp1631.pdf_