Haiti is today habitually classified as the poorest country in the Western Hemisphere, and has experienced a stagnation or even decline in GDP per capita since at least shortly after the 150th anniversary of its independence in 1954.1 In 2004, the year when Haiti should have celebrated its 200th anniversary of independence, it was briefly the focus of world media attention as President Jean-Bertrand Aristide was chased from the country by armed rebels and civil unrest. After that, Haiti was run by a transition government supported by the UN stabilization force MINUSTAH (Mission des Nations Unies pour la stabilisation en Haïti). The transitional government was replaced by an elected government, following elections for all levels of government in February 2006. One of the many urgent problems facing this government is the widespread and deep poverty that affects the majority of Haitians.
This paper is an abbreviated version of a poverty profile for Haiti.2 We begin by describing the methodology used for the poverty profile. These choices are far from innocent, and we therefore go into some detail on how we have dealt with the technical issues. We thereafter present stylized facts of poverty in Haiti; for the sake of brevity we present a limited number of tables in the paper and give additional information in an annex3. Poverty rates in rural areas are more than twice those in the metropolitan area, and we show that income differences between rural areas and Port-au-Prince are not caused by differences in household characteristics, nor is rural poverty caused by landlessness.4 We find that poverty in Haiti presents the following three specificities compared to other countries in the region: A larger part of the population falls into poverty than elsewhere in the region; the majority of the poor live in the countryside; and finally, landlessness is not a defining feature of rural poverty in Haiti.
II. MEASURING MONETARY POVERTY IN HAITI
The procedure for measuring poverty has generated a large body of literature (see, e.g., Ravallion, 1998 and Deaton, 2001 for an overview of the debate). Even when the theoretical basis is agreed upon - to use a monetary indicator of well-being and to use an absolute definition of poverty - a bewildering number of technical choices must be made when analyzing any given data set. Using data for seventeen Latin American countries, Székely et al. (2000) test the sensitivity of poverty measurements and find that the proportion of poor in Latin America could be said to be 13 percent or 66 percent, depending on the technical choices made. They conclude that "poverty statistics are rather meaningless if the underlying choices and assumptions needed for their estimation are not made explicit." We will therefore outline our own choices here in some detail.
Our starting point is to use a monetary definition of poverty, and apply the US$1 and US$2 per person per day poverty lines that are commonly used for international comparisons. As our data set does not contain consumption expenditures, we are not able to calculate a national poverty line, and the US$1 and US$2 lines provide the best solution.
The reasoning behind monetary measures is that the purchasing power of money gives a certain level of well-being through the consumption of goods and services bought for the money, under standard assumptions of utility-maximizing behavior. Poverty comparisons based on a monetary definition of poverty run into two distinct but related problems: The problems of comparing between countries, and the problems of comparing within countries.
In order to be able to compare between countries, Purchasing Power Parity (PPP) conversion factors are normally used instead of market-based exchange rates, but this approach has been criticized for several shortcomings (Deaton, 2001; Reddy, 2002). There are theoretical problems surrounding the concept of PPP itself; there are problems related to the use of PPP conversion factors for converting the value of the consumption of the poorest part of the population; and there are problems related to the quality of data used for computing the PPP factors. …