n. V2-6.06| THE LIST

So where is the context of IkoCiti? What is the world that this chapter describes? What is the theme? You start without introducing anything. You go back to the poverty measurement theme? At least that what yiu think reading Para 1/2?

01| Before a government can help a poor person it has to decide who counts as poor, and that decision is made by an instrument rather than by a judgement. The instrument has an error rate, the error rate is calculable, and it is almost never published next to the coverage figure that the same programme advertises. For anyone working in a barrio this is not an abstraction: it determines which of your neighbours are on the list and which are not, and it will have decided that before you arrived. Understanding how the list was built explains most of what looks, from inside the barrio, like corruption.

02| The dominant instrument outside rich countries is the proxy means test, and its logic follows from a real constraint. Income cannot be verified where most work is informal and nobody has a payslip, so the state scores households on things that can be observed instead — roof material, floor, assets, household composition — fits those observations against survey consumption data, and draws a line somewhere on the resulting score. It is a statistical prediction of poverty rather than a measurement of it. That distinction sounds academic and is the source of everything that follows.

03| The predictions are weak, and their weakness is not a secret. Most proxy means tests explain only 40 to 60 percent of the variation in consumption between households [Kidd & Wylde 2017, p. ??]. A model carrying that much unexplained variance, used to sort people into included and excluded, will misclassify a large share of them by construction rather than by accident. This is what the instrument does when it is working correctly, which means no amount of better administration will fix it.

04| The measured exclusion errors are correspondingly severe, and they cluster at the poorest end. Indonesia's PKH, targeting the poorest 5 percent, excludes 93 percent of them; Kenya's HSNP, targeting the poorest 26 percent, excludes 62 percent; Mexico's Oportunidades, aimed at the poorest 20 percent, around 70 percent; Cambodia's ID-Poor 56 percent; Ecuador's Bono de Desarrollo Humano, targeting the poorest 8 percent, excludes 60 percent of the extreme poor; Georgia's TSA, aimed at the poorest 15 percent, excludes 50 percent [Kidd & Wylde, Exclusion by Design, 2017 — country table, page numbers to be recovered]. These are not failed programmes. They are the well-known ones, running as designed.

05| Three mechanisms drive those numbers and they compound rather than offset. The first is model error, from the paragraph above, and it is irreducible without exactly the income data whose absence forced the proxy in the first place. The second is data quality: entry errors in the Indonesian survey averaged 14.7 percent across regions and reached 37 percent in some of them [ibid.], because the instrument is only ever as good as a clipboard filled in at a doorstep by someone paid per household. The third is time, and it is the one most often forgotten.

06| Poverty is not a fixed population, and the list assumes that it is. Nearly half of households leave the poorest 20 percent within two to three years, while recertification typically runs on cycles of five to ten [ibid.].

rle// Is poverty a rotating dynamic? How many are still in bottom 20% poverty seen through a five year lens? And how much is the difference between the lowest 20% and the 4th quintile?

A list that was perfectly accurate on the day it was drawn is therefore substantially wrong within thirty-six months, and most lists go far longer than that without being redrawn. What the state is working from is a photograph of a moving crowd, taken years ago, and treated as a register of who is there now.

07| There is a cost beyond the arithmetic, and it lands on the barrio rather than on the treasury. Where neighbours in visibly identical circumstances receive different answers from an instrument nobody can explain, the outcome is read as favouritism — because from inside the barrio it is genuinely indistinguishable from favouritism. Kidd and Wylde record roughly 30 percent of villages in Indonesia's BLT programme experiencing protests, including property damage [ibid.], and the field testimony is plainer than any statistic in this part:

"See this old lady she is totally blind and lives by herself...yet she is not on the list."— Kidd & Wylde, Exclusion by Design, 2017 [p. ??]

rle//Theme? The failure of government to provide more custom-made services?In a world where the private sector seems to know everything about us, gov does not know anything. Big brother is a private company?

08| Targeting therefore does not only misallocate money; it spends social cohesion, and that expenditure appears in no budget line and no evaluation. A barrio that has been through several rounds of an unexplainable list has learned that selection is arbitrary and that somebody is probably being favoured, which is a lesson it will apply to the next organisation that arrives with a selection process of its own. Anyone entering after that history inherits the suspicion whether or not they deserve it. This is a real cost of entry and it is invisible in the data a newcomer would normally consult.

rle// that is what we meet??

09| The alternatives are worth knowing because each trades a different thing. Categorical targeting — by age, disability, or the presence of children — is crude, cheap and verifiable, and in practice it frequently beats statistical targeting on exclusion, because the criterion is observable and can be contested by the applicant who was refused. Community-based targeting uses local knowledge that no proxy captures and can be considerably more accurate at the extremes, at the price of importing local power relations directly into the selection, which is a real cost rather than a hypothetical one. Universal provision removes the problem by removing the instrument, driving exclusion toward zero and cost toward the maximum — an explicit trade, and at least an honest one.

rle// what is the point of Para 09?

10| The strongest argument for the defence has to be admitted, because it is the reason the practice survives. Where a budget is genuinely fixed and small, something has to choose, and a weak predictor may still be better than the alternatives of first-come-first-served, open official discretion, or whoever shouts loudest. The case against proxy means testing is powerful on accuracy and much weaker on alternatives. That asymmetry is precisely why an instrument this inaccurate remains standard practice among competent people who know its error rate.

rle// defence?? you need to be more explicit

11| What this leaves on the ground is a barrio containing two populations that look identical and are treated differently, with no explanation available to either. Some of the people most in need are formally invisible to the programme meant for them, the error runs overwhelmingly toward exclusion rather than inclusion, and the resulting resentment attaches to whoever is nearest. Any project that involves selecting some residents and not others is walking into that history, and will be read through it. Whether a selection can be made in a way the barrio recognises as fair is not a research question, and it is the one this chapter hands forward.

rle//The title is The List. A list of poor people receiving X and other similar people receiving nothing, or something different? How common is this in the world? How are other countries doing it? In Colombia they seem to distinguish barrios in four levels each receiving different things - subsidies?

There is no bridge to the next article and there was no intro how this article connects to the previous? Is thisWho does the proxy means test? The national Gov or the local municipality. How many levels of government are there? National, Stae, regional, city/village??