IMMPA:Integrated Macroeconomic Model

for Poverty Analysis

Pierre-Richard Agénor, Alejandro Izquierdo

Hippolyte Fofack, and Derek Chen

The World Bank

Background. Why build a (relatively) complex model? IMMPA: main features. Calibration and solution of the IMMPA prototype for

low-income countries. Policy exercise: debt relief, public expenditure

allocation, and poverty reduction. Modifications and extensions IMMPA prototype for middle-income countries: the

case of Brazil (re-specification of financial system; bond financing).

Background

Sept. 1999: World Bank and IMF adopt Poverty Reduction Strategy Papers (PRSPs) as basis for concessional lending to low-income countries, including resources under the enhanced initiative for debt relief for Heavily Indebted Poor Countries (HIPC).

PRSP: in principle, formulated and written by national authorities in consultation with civil society; aims to describe the poverty conditions in the country and to set out a medium-term action plan to reduce the incidence of poverty and strengthen economic growth.

Design of PRSPs: involves considering the nature of policy tradeoffs that policymakers may face in allocating (scarce) public resources.

Understanding the nature of these trade-offs requires to

take into account the key linkages between micro and macro factors in designing a poverty reduction strategy;

use appropriate policy tools to prepare and compare consistent, quantitative assessments of alternative strategies.

More generally, poverty reduction has become an overriding objective of economic policy in low- and middle-income countries alike.

Poverty is as much an issue in Chile, Brazil, or Morocco, as it is in Burkina Faso.

Eradication of poverty is the first among the Millenium Development Goals (MDGs), adopted by the UN in Sept. 2000.

Monitoring poverty trends and providing advice on poverty reduction strategies: key responsibilities of the World Bank.

MMiilllleennnniiuumm DDeevveellooppmmeenntt GGooaallss ((MMDDGGss))

Goals and Targets Indicators Goal 1: Eradicate extreme poverty and hunger

Target 1: Halve, between 1990 and 2015, the proportion of people whose income is less than one dollar a day

Proportion of population below $1 per day Poverty gap ratio [incidence x depth of poverty] Share of poorest quintile in national consumption

Target 2: Halve, between 1990 and 2015, the proportion of people who suffer from hunger

Prevalence of underweight children (under-five years of age) Proportion of population below minimum level of dietary energy

consumption

Goal 2: Achieve universal primary education

Target 3: Ensure that, by 2015, children everywhere, boys and girls alike, will be able to complete a full course of primary schooling

Net enrolment ratio in primary education Proportion of pupils starting grade 1 who reach grade 5 Literacy rate of 15-24 year olds

Goal 3: Promote gender equality and empower women

Target 4: Eliminate gender disparity in primary and secondary education preferably by 2005 and to all levels of education no later than 2015

Ratio of girls to boys in primary, secondary and tertiary education

Ratio of literate females to males of 15-24 year olds Share of women in wage employment in the non-agricultural

sector Proportion of seats held by women in national parliament

Goal 4: Reduce child mortality

Target 5: Reduce by two-thirds, between 1990 and 2015, the under-five mortality rate

Under-five mortality rate Infant mortality rate Proportion of 1 year old children immunised against measles

Goal 5: Improve maternal health

Target 6: Reduce by three-quarters, between 1990 and 2015, the maternal mortality ratio

Maternal mortality ratio Proportion of births attended by skilled health personnel

http://www.worldbank.org/research/povmonitor

Issues for macroeconomists:Macroeconomic and financial stability is not an

end in itself; putting poverty explicitly in the “welfare” function requires renewed effort to better understand analytically the micro-macro linkages that matter when designing poverty reduction strategies.

Existing policy tools must be amended, and new ones must be developed, in order account quantitatively for those channels that are deemed important in determining how adjustment policies affect the poor.

IMMPA: an integrated quantitative macro-economic framework developed at the World Bank for analyzing the impact of policy and external shocks on income distribution, employment and poverty in both low-income, highly-indebted countries and middle-income developing economies.

Dwells on the analytical and applied research

conducted in academic and policy circles over the past two decades on macro- economic and structural adjustment issues in developing economies.

Why builda (relatively) complex model?

Issue is complex; it serves no good purpose to pretend that it is not.

Existing, commonly-used policy tools do not come anywhere close to capturing some of the most important channels through which exogenous and policy shocks are transmitted to the poor.

Models are issue-specific; trying to “force” a model to answer questions that it is not designed to address hampers our ability to address relevant policy questions.

Models are not built only to produce numbers but also to provide qualitative insights (general equilibrium effects).

Before looking for “shortcuts”, one needs a conceptual roadmap to understand the costs and benefits of simplification.

Lack of skills is indeed a constraint in many cases; but many middle-income countries, and some low-income countries, have the capacity to implement the model.

Lack of adequate data: also a problem. But do you wait until the data have improved sufficiently, or do you start with existing data, no matter how imperfect, and improve the database gradually?

IMMPA: Main Features

Treatment of the production structure and the labor market (wage formation and sources of segmentation, skills acquisition, rural-urban migration).

The financial system and the credit market (portfolio structure--stock decisions--and the treatment of credit market imperfections).

Adverse effect of external debt on domestic private investment (foreign exchange constraint, confiscation risk, or other channels).

Allocation of public expenditure (to current transfers, infrastructure, education, health).

Systematic link with a household income and expenditure survey; allows a more accurate assessment of poverty effects of shocks.

Dynamic structure allows the analysis of dynamic tradeoffs that poverty-reduction strategies may entail regarding the sequencing of policy reforms—particularly between short-term stabilization policies and structural measures.

http://www.worldbank.org/immpa

Structure of IMMPA

Production Structure Rural Production. Two agricultural goods, traded

and non-traded.

Urban Production Informal sector: produces non-traded goods.Formal sector:

Private traded and nontraded goods. Public non-traded good.

Urban production

Production ProcessAggregate production

Rural production

Nontraded good

Traded good

Formal Informal

Private PublicTraded good

Nontraded good

Nontraded good

Nontraded good

The Labor Market Rural Unskilled Workers: employed in production

of traded or non-traded agricultural goods.

Urban Unskilled Workers: employed in private sector, public sector or in informal economy.

Skilled Workers: employed in formal urban sector (public or private), or unemployed.

WagesFlexible in informal sector and nontraded good

sector in agriculture.Fixed in real terms in traded good sector in

agriculture and for urban unskilled (minimum wage).

Efficiency wage for urban skilled workers (wage-productivity link).

Both wages and employment of skilled workers are determined by firms in the urban private sector.

Other assumptions are (of course) possible.

Rural to urban migrationDepends on expected relative wages.Harris-Todaro tradition.

Human capital accumulationDecision to acquire skills by unskilled workers is

endogenous.Depends on a) the ratio of the cost of acquiring

education relative to the initial wealth of unskilled workers (decision is subject to credit constraints);

b) relative wages, and c) the stock of public capital in education.

Population growth rate: inversely related to the composition of the labor force (ratio of skilled to unskilled labor); fertility effect.

Implications:Modeling the informal economy and labor market

segmentation is essential for understanding the short-run dynamics of employment and urban poverty.

Modeling the decisions to migrate and acquire skills is essential for understanding the medium- and long-run dynamics of poverty and the role of public policy.

Rural unskilled

Unemployment

Employment and Skills

Labor force

migration Urban unskilled

skills acquisition

Skilled

FormalInformalNon-traded

Traded

Rural productionUrban production

Labor demand

Financial Assets

Household savings are allocated to financial wealth accumulation.

Portfolio allocationCurrency. Depends on income, interest rates and

inflation.Bank deposits. Allocation between domestic and

foreign-currency deposits (held abroad) depends on relative rate of return on these assets.

No real assets.

Monetary base

Household savings

World interest rateand deposit rate

Real currency balances

Financial Assetsand the Money Market

Domesticdeposits

Foreigndeposits

Balance of payments

Credit to government

Financial wealth

Endogenous

Exogenous

The Credit Market Households: deposits in banks (domestic source

of funding).

Firms borrow for both physical capital accumulation

and financing of working capital needs.

Given retained profits and foreign loans, borrow from domestic banks.

BanksCollect domestic deposits. Reserve requirements

determine funds available for lending to domestic firms.

Lending to government is exogenous. If domestic funds are not sufficient to satisfy

demand, they borrow on world markets.Alternatively: with exogenous foreign borrowing,

excess liquid reserves adjust to clear the credit market.

Lending rate is set as a markup over the cost of funds (domestic and external).

Markup rate (premium): Two possible under-lying explanations.

A. Markup depends inversely on the ratio of the stock of private capital to total loans. Collateral (or net worth) effect.

B. Markup is a function of monitoring and enforcement costs of loan contracts. Costs are a positive function of the amount of loans (scaled by total bank assets) due to congestion in courts.

Both are consistent with recent models of credit market imperfections.

Dependence of the cost of funds on net worth: a critical aspect of the model.

For instance, a nominal exchange rate devaluation will reduce firms’ net worth and may dampen private investment by increasing the cost of capital.

Worsening of the real burden of debtors,and increase in lending rates, may offset the expansionary effect due to capital gain for net creditors in foreign currency (households).

Devaluation may be contractionary; longer-run effect also due to impact on capital formation.

Foreign borrowing by commercial banks

Working capital and Investment

Firms’ borrowing(net of retained earnings)

Foreign borrowing by firms

Loan rate

Domestic funds for bank lending

Domesticdeposits

World interest rateand deposit rate

The Credit MarketReserve

requirement

Lending to government

Demandfor credit

Endogenous

Exogenous

Premium

Net worth

Public Sector Central Bank

Prototype: CB sets both the exchange rate and the bank deposit rate.

Nominal supply of currency: changes in the monetary base.

Counterparts to the monetary base: official foreign reserves and (exogenous) credit to government.

Alternatively, model can be solved with a flexible exchange rate.

GovernmentSets tax rates on domestic sales, imports, income

and profits.Allocates public expenditure to current

consumption and investment (infrastructure, health, education).

Public spending has three types of effects.Conventional aggregate demand effects

(impact on consumption);

Effect on private investment (complementarity between public investment in infrastructure and private capital formation);

Supply-side effect. Private production (agriculture and urban formal sectors) depends on the stock of government capital in infrastructure and health.

Deficit financing Various options are possible. Prototype: financing is given from “below the

line.”

Flow budget constraint determines lump-sum transfers.

Implications: focus on the income effects of induced changes in the budget;

no changes in distortionary taxes and incentives.

Aggregate demand Private

disposable income

Credit fromcentral bank

Commercial bank loans

to government

Financeable Government deficit

Lump-sum transfers

Foreign borrowing

by government

Interest paymentson debt

Non-interestexpenditure

Distortionary tax

revenues

Endogenous

Exogenous

PublicEnterprise

Net Revenues

Public Sector

Balance of Payments Current account

Imports, exports and debt service are endogenous.

Capital flows and the capital accountNet borrowing by firms and government is

exogenous.Borrowing by domestic banks and deposits

abroad by households are endogenous. Movements in official reserves determine the

monetary base.

Balance of payments(official reserves)

Monetarybase

Currentaccount

Foreign deposits

by householdsImports

Balance of Payments

Total foreign

borrowing

Capitalaccount

Exports

Debt service

Firms Banks Government

Endogenous

Exogenous

Standard CGE specifications for production functions (e.g. intermediate inputs: Leontief

specification); consumption (linear expenditure system); composite goods and prices; consumer price index is an average of composite

prices.

Savings rate. Depends positively on the real deposit rate and thus inversely on inflation.

Other Features

Private investment and external debt. Inverse and possibly nonlinear relation (inclusion

of a quadratic term for external debt service to tax revenues).

May capture various factors: foreign exchange constraint, crowding-out effect of high indebtedness, confiscation risk.

Factor income

Rural sector

Urban sector

Nontraded goodTraded goodInformal sector Formal sector

Private sector firmsPublic sector

Traded good

Nontraded good

Nontraded good

Nontraded good

Production

Labor force

Unskilled labor supply formal sector

Skilled labor supply

Skills acquisition

Skilled unemployment

Migration

Private consumption

Savings

Unskilled labor demand formal sector

Skilled labor demand

Skilled wageUNSKILLED MINIMUM WAGE

Financial wealth

Foreign depositsDomestic deposits

Government spending

Money balancesMonetary

baseBalance of payments

Financeablefiscal deficit

Domestic funds for bank lending

RESERVE REQUIREMENTS

WORLD INTEREST RATE Lending rate

Premium

Private Working capital needs Physical investment

Firms' borrowing(net of retained earnings)

FIRMS' FOREIGN BORROWING

Firms' demand for credit

Private capital stock

Unskilled labor supply informal sector

Informalsector wage

Unskilled labor demand informal sector

Domestic demand

Composite equilibrium prices

FOREIGN PRICES OF IMPORTS

FOREIGN PRICES OF EXPORTABLES

CREDIT TO GOVERNMENT CENTRAL BANK

AND COMMERCIAL BANKS

Imports

Current account

Lump-sumtransfers

Disposable income

PUBLIC FOREIGN BORROWING

Interest payments on debt

Tax revenue

BANK FOREIGN BORROWING

Total Foreign Borrowing

Total debt service

Figure -. IMMPA: Analytical Structure

Labor demand

Nontraded agricultural sector wage

TRADED AGRICULTURALSECTOR WAGE

Skilled reservation wage

Rural labor supply

Urban labor supply

Domestic supply

Interest incomeOverall level of prices

Exports

Public capital stock

Note: Exogenous variables are in capital letters.

DOMESTIC DEPOSIT RATE

Public enterprisenet revenue

NON INTEREST EXPENDITURE

Official Reserves

Private

Poverty and Income Distribution Analysis

A. Measures of income distribution:

Gini coefficient and Theil index.

Based on six households categories: workers in the rural traded sector, rural non-traded sector, urban unskilled informal economy, urban unskilled formal sector, urban skilled formal sector; and capitalists.

Other measures can be added.

B. Link with household surveys:

IMMPA simulation results can be linked to survey data on income and expenditure…

...to estimate the impact of shocks on income within each group as well as average income variations among groups.

This allows us to calculate measures of poverty and changes in income distribution across groups.

Approach:

Step 1. Use the information provided in the household survey to classify the available sample into IMMPA’s six categories of households, so as to establish an interface between the model’s simulation results and actual household income and expenses.

Step 2. Following a shock to the model, calculate real growth rates in per capita consumption and disposable income for the six categories of households.

Step 3. Apply these growth rates separately to each individual per capita (disposable) income and consumption expenditure observation in each of the six groups of households in the survey…

…this gives absolute income and consumption levels for each household, in each group, following the shock.

Step 4. Given rural and urban poverty lines (expressed in monetary units and rising at the rural and urban unskilled CPI growth rates), and using the new absolute levels of income and consumption in each group, calculate

post-shock poverty indicators (headcount index and poverty gap index);

income distribution indicators (Gini coefficient and Theil inequality index).

Step 5. Compare post-shock indicators with baseline values to assess impact of the shock on poverty and income distribution.

For all indicators, IMMPA generates three measures: short-term measure (first two periods following a shock); medium-term measure (between 3 and five periods); and long-term measure (6 and 10 periods).

Choice of intervals is somewhat arbitrary and can be changed.

These measures allow the analyst to identify and discuss possible Dynamic tradeoffs in the analysis of policy choices, by contrasting their short- and longer-run effects on the poor and income distribution.

Note I: limitations of Headcount index

1. Does not indicate how poor the poor really are (it remains unchanged even if all people with incomes below the poverty line were to experience e.g. a 50% drop in income).

Put differently: when a poor person become poorer, the index will not increase.

2. Implies that income distribution among the poor is homogeneous (no distinction between a poor person who earns one monetary unit less than the poverty line and one who earns 100 monetary units less than the poverty line).

Link IMMPA-Household SurveySurvey Shock to IMMPA

Aggregate datain 6 IMMPA categories

Growth rates of income and consumption for 6 categories

Apply to each individualin 6 categories

(new absolute levels)Rural

poverty lineUrban

poverty line

Poverty indicators(short, medium and long term)

Comparison with baseline scenario

Urbanprice index

Ruralprice index

Poverty Sheet - Consumption Based

Note II:

Procedure above: assumes that the user matches households as defined in the macro component of IMMPA and a household survey using information on the main source of income of household heads.

Alternative treatment: possible if the household survey provides sufficient detail regarding the composition of income among individual members of each household.

“Light surveys” tend to concentrate on the household head, whereas more in-depth surveys provide richer information.

If the information is detailed enough, and if each member of a household can be “allocated” to one of the six IMMPA income groups, model-generated growth rates of income and consumption can be applied separately to each individual income-earner (as in Step 3 above).

Poverty and income indicators can be generated using either “individual” income earners or “composite” households.

However, whether accounting for heteroge-neity in the sources of income among individual household members makes a difference or not is generally case specific; it depends on

the characteristics of the intra-household distribution of income (which depends on risk diversification strategies);

the extent to which growth rates of income and consumption generated by IMMPA following a shock differ among the various income groups.

If, for instance, the intra-household distribution as given in the survey is such that most of the income of each composite unit is generated by the household head…

…treating the household as a homogeneous unit and applying the same growth rate of income to each member should not result in significant errors.

Note III:

Approach remains subject to the assumption of a stable within-group distribution (relative income and consumption levels--and positions--within each group do not change because same growth rate is applied to each individual).

Alternative approach: include individual data directly in the model and use micro simulation techniques to exploit intra-group information.

Benefit: allows the analyst to distinguish, in the evolution of poverty indicators, the specific contribution of three factors: changes in the poverty line (when it is treated as endogenous), across-group variations, and changes in intra-group distribution.

However, complex and costly to implement.

Existing studies: not obvious that changes in intra-group distribution are large compared to inter-group distribution (results appear to be shock-specific).

Solution and Calibration

Prototype: Initial values (and shares) selected to represent a low-income country.

Parameter values (elasticities) are consistent with existing CGE models.

Computer programs used for solution: Excel and Eviews; emphasis on ease of use.

Model is solved iteratively using either a Gauss-Seidel algorithm or the Newton-Raphson technique.

The equilibrium condition of the money market is omitted due to Walras Law, but used for consistency check.

Existing operational manual (see Chen et al. (2001)):Procedures for building the financial SAM that

underlies IMMPA;

How to estimate econometrically some key behavioral functions (private investment, skills acquisition, migration function, asset demand equations);

How to solve IMMPA and simulate policy and exogenous shocks.

Endogenous Variable File

Exogenous Variable File

Eviews Simulation Program

Output File

Stage 1

Stage 2

Stage 3Stage 4

Stage 5

Graphs(Excel File)

Plots of current-base run values for endogenous variables are generated

Household Survey Data File

Stage 4

IMMPA Simulation Package Setup

Exogenous Variables

Terms-of-Trade Shock

Domestic Credit Shock

Simulation Exercise:Debt Relief, Public

Expenditure Allocation,and Poverty Reduction

Savings from debt relief. Reallocation of government expenditures (at a

given level of the overall deficit) toTransfers; Investment in infrastructure; Investment in education.

Simulation captures possible effect of lower external debt on private investment.

Reminder: public spending has three types of effects:

Conventional aggregate demand effect (impact on consumption);

Effect on private investment,as a result of the complementarity between public investment in infrastructure and private capital formation;

Supply-side effect. Private production (agriculture and urban formal sectors) depends on the stock of government capital in infrastructure and health.

Investment in Education Effect on incentives to acquire skills:

Unskilled urban workers invest in acquiring skills depending on relative wages, the stock of public capital in education, and initial wealth.

Increase in supply of skilled workers must be matched with increase in labor demand to avoid higher unemployment.

Effect on labor market and wages:

As unskilled urban workers become skilled, unskilled labor supply in the formal urban economy is replenished by workers previously employed in the informal economy.

Given a fixed minimum wage for unskilled labor in the formal urban economy, informal wages increase due to reduction in supply of labor in the informal economy.

Effect on migration:

Increase in average urban wage paid to unskilled labor is an incentive to migrate from the rural sector.

Increase in the supply of urban unskilled workers.

Puts downward pressure on informal sector wages and counteracts initial effects.

Increase in wages in the nontraded agricultural sector.

Investment in Infrastructure

Complementarity effect on private investment.

Infrastructure improvements have a positive aggregate effect on the economy:

they improve labor productivity by facilitating economic activity.

This increases labor demand in all sectors, and consequently reduces poverty (“crude” headcount index).

Better infrastructure also raises private investment, as it improves the productivity of private capital and returns on private capital. This further stimulates economic activity.

Combining public investment alternatives: Infrastructure and private investment affect

labor demand in all sectors, but especially the demand for skilled workers.

This is because private capital and skilled workers have a high degree of complementarity (or low net substitutability).

Thus, considering alternative public investment strategies is important.

Investment in education increases the supply of skilled workers, whereas investing in infrastructure stimulates private investment, which raises the demand for skilled labor.

Investment in Health Investment in education primarily affects the urban

economy, whereas investment in health benefits all workers and therefore has aggregate productivity effects.

Investment in health has similar effects as investment in infrastructure; acts as a complement to spending in education and infrastructure.

Poverty Outcomes “Crude” poverty measure: rural non-traded and

urban informal workers.

Evidence of poverty concentration in these sectors (65-90% in rural areas for sub-Saharan Africa).

Associated with concept of poverty reduction as access to higher quality jobs.

Policy Rankings

Total PovertyHeadcount

Infrastructure Education Transfers

Urban PovertyHeadcount

Education Infrastructure Transfers

Rural PovertyHeadcount

Infrastructure Education Transfers

- Infrastructure is PPF enhancing- Education increases the supply of skilled labor but there may be unemployment- Has direct effects in urban areas- Policy mix? Better than Education only

Why?

Modifications and Extensions

IMMPA: a flexible tool that can be amended or extended in various ways.

The credit market If banks cannot borrow on world capital markets, excess

liquid reserves may be the equilibrating variable. Alternatively, credit rationing can be introduced.

Informal credit market Important in some countries. But it adds another layer of complexity.

Household wealthReal assets (livestock, land) can be added.May be important for rural households.Some assets are held in “unproductive” form.But even if household expenditure is independent

of total wealth: reallocation of assets would affect spending (through interest income on deposits).

However: serious measurement problems.

Cost of foreign borrowing: can be made endogenous (risk premium).

Savings rate Depends negatively on inflation due to the rate-of-return

effect. But it may also depend positively on inflation

(precautionary motive).

Productivity growth TFP growth: accounts for a large share of output growth

in many countries (see Agénor (2000)). Can be endogenized by using government policy

variables that affect incentives to adopt new technologies or to use resources efficiently.

Government deficitAlternative financing rules are easily

implemented.Examples:

Endogenous domestic borrowing (from commercial banks or the central bank) or foreign borrowing.

Endogenous current expenditure.Endogenous tax rates.

Exchange rate regimeFlexible exchange rate regime can be easily

implemented.Possibly large effect on bank balance sheets and

household wealth (valuation effects).

Adverse effect of external debtHigh public external debt may also lead to private

capital flight.Debt reduction may not only stimulate private

investment but also private capital inflows (treated as exogenous in the prototype model).

Another source of positive externalities.

Unskilled unemploymentModel: urban informal labor market is

characterized by free entry and high degree of wage flexibility.

Adverse shocks to the formal economy translate into lower (average) productivity in the informal sector.

However, despite the absence of entry restrictions, urban open unemployment is often high for unskilled workers.

Some evidence suggests that labor mobility between the formal and the informal sectors, although very high, is not perfect.

Analysis could be extended to account for unskilled unemployment

Argument: informational frictions may force unskilled workers to remain unemployed while they are searching for a job in the formal sector.

Harris-Todaro type mechanism.See IMMPA applications for Brazil (Agénor,

Fernandes and Haddad (2002)) and Morocco (Agénor and El Aynaoui (2002))

References

Agénor, Pierre-Richard, The Economics of Adjustment and Growth, Academic Press (San Diego, Cal.: 2000).

-----, “Business Cycles, Economic Crises, and the Poor: Testing for Asymmetric Effects,” unpublished, the World Bank (October 2001).

-----, “Macroeconomic Adjustment and the Poor: Analytical Issues and Cross-Country Evidence,” unpublished, the World Bank (January 2002a).

-----, “The Analytics of Micro-Macro Linkages for the Design of Growth and Poverty Reduction Strategies,” paper in progress, the World Bank (February 2002b).

Agénor, Pierre-Richard, Reynaldo Fernandes and Eduardo Haddad, “Analyzing the Impact of Adjustment Policies on the Poor: An IMMPA Framework for Brazil,” work in progress, the World Bank and University of Sao Paulo (January 2002).

Agénor, Pierre-Richard, Alejandro Izquierdo, and Hippolyte Fofack, “IMMPA: A Quantitative Macroeconomic Framework for the Analysis of Poverty Reduction Strategies,” unpublished, the World Bank (December 2001).

Agénor, Pierre-Richard, Karim J. P. El Aynaoui, “Labor Market Reforms, Unemployment and Poverty in Morocco: A Quantitative Analysis,” work in progress, the World Bank (January 2002).

Chen, Derek, Hippolyte Fofack, Henning Jensen, Alejandro Izquierdo, and Daouda Sembene, “ IMMPA: Operational Manual,'' unpublished, the World Bank (November 2001).