Universal Basic Incomes versus Targeted Transfers: Anti-Poverty Programs in Developing Countries

A universal basic income has labor supply distortions because there is still a tax on labor income, and adding it requires adjusting the tax schedule elsewhere.

“3) as emphasized by Saez (2002), it is not the case that a universal basic income has no labor supply distortions, since there is still a tax on labor income, and in particular adding the universal basic income to the tax schedule requires adjusting the tax schedule elsewhere.”

• cites Saez (2002) sourcePartly supported

  The Abstract (p. 2) describes an equity–efficiency trade-off, labor-supply responses to income transfers, and guaranteed income that is taxed away at high rates. This supports the point that transfers can involve labor-supply distortions, but the readable passages do not specifically discuss a universal basic income or say that adding one requires adjusting the tax schedule elsewhere. (LLM, conf=0.77)

  Experimental — accuracy not yet measured

• AssumptionThe authors state this as an assumption: A universal basic income program is fully financed by a progressive income tax schedule.

A universal basic transfer can be recast as a particular feature of a tax schedule, as described by Saez (2002).

“Given that a universal basic income would need to be financed through taxation, one can recast universal basic transfers as a particular feature of a tax schedule, as described by Saez (2002) and illustrated by Figure 1.”

• cites Saez (2002) sourcePartly supported

  The Abstract (p. 2) describes income-transfer programs through optimal tax formulas, including a negative income tax with guaranteed income support; the figures on p. 40 are labeled transfer and tax schedules. This supports treating transfers as part of a tax schedule, but the readable passages do not specifically describe a universal basic transfer or its tax financing. (LLM, conf=0.78)

  Experimental — accuracy not yet measured

In Colombia, as the proxy-means formula became known, reporting bunched substantially just below the eligibility cutoff.

“Using data from Colombia, Camacho and Conover (2011) show that over time as the formula became known, there was substantial bunching of reporting right below the eligibility cutoff.”

• cites Camacho and Conover (2011) sourceSupported

The fundamental tradeoff between inclusion and exclusion error remains even with more sophisticated prediction methods.

“While more sophisticated prediction methods (like more flexible prediction equations, machine learning methods, and so on) can improve on these predictions (McBride and Nichols 2016), the fundamental tradeoff between inclusion and exclusion error that we document here remains.”

• cites McBride and Nichols 2016 sourceCould not verify

  The citation is attached to the assertion that more sophisticated methods can improve predictions; the sentence presents the inclusion–exclusion tradeoff as the citing authors’ own finding. The indexed passages provide repository metadata and an abstract, but no passage from the paper’s full text establishes the prediction-improvement assertion. Searches for machine-learning improvements and inclusion–exclusion error found no decisive full-text passage. (LLM, conf=0.72)

  Experimental — accuracy not yet measured

• AssumptionRests on an assumption: The Indonesian BLT program and the Peruvian Juntos program each target roughly one-third of the population.

In an experimental comparison in Indonesia, citizens vastly preferred community-based targeting to a data-based proxy-means test.

“we found that citizens vastly preferred the community based-approach to a data-based proxy-means test.”

• cites Alatas, Banerjee, Hanna, Olken, and Tobias 2012 sourceSupported

A substantial fraction of households eligible for government benefit programs do not enroll.

“A substantial fraction of households who are eligible for a government benefit program do not enroll in the program.”

• cites Moffitt (2003) sourcePartly supported

  The Summaries of the Chapters (p. 9) report a declining Medicaid participation rate among eligible children, which supports the existence of eligible nonparticipants. Our copy does not establish that a substantial fraction of eligible households fail to enroll across government benefit programs. (LLM, conf=0.74)

  Experimental — accuracy not yet measured
• cites Currie (2006a)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

There was no detectable elite capture in the community-based targeting experiment.

“Furthermore, there was no detectable elite capture (Alatas, Banerjee, Hanna, Olken, and Tobias 2012).”

• cites Alatas, Banerjee, Hanna, Olken, and Tobias 2012 sourceSupported

Communities appear to target based on earning potential rather than actual earnings.

“In Alatas et al. (2012), we showed that communities appear to target based on earning potential, rather than actual earnings.”

• cites Alatas et al. (2012) sourceSupported

Across seven randomized government cash transfer programs, there was no systematic evidence that the programs reduced labor supply.

“We found no systematic evidence that any of these programs reduced labor supply.”

• cites Banerjee, Hanna, Kriendler, and Olken 2017Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

In both Indonesia and Peru, to reach 80 percent of intended beneficiaries, the government would need to tolerate an inclusion error of between 22 and 31 percent.

“In both Indonesia and Peru, to reach 80 percent of the intended beneficiaries (exclusion error of 20 percent), the government would need to tolerate an inclusion error of between 22 and 31 percent.”

• AssumptionRests on an assumption: The Indonesian BLT program and the Peruvian Juntos program each target roughly one-third of the population.

For intermediate cutoffs with inclusion errors around 30-40 percent, horizontal equity is violated about 20 percent of the time in Peru and 30 percent in Indonesia.

“We find that for cutoff levels in an intermediate range, in which inclusion errors were roughly 30-40 percent, it is common to find that horizontal equity is violated about 20 percent of the time in Peru and 30 percent of the time in Indonesia.”

The transfer level to eligible households steeply falls as more people are allowed into the program.

“The transfer level to eligible households steeply falls as you allow more people into the program.”

• AssumptionRests on an assumption: The Indonesian BLT program and the Peruvian Juntos program each target roughly one-third of the population.

Income taxes and import duties make up about 42 percent of tax revenue for low- and middle-income countries, with the rest coming from consumption taxes.

“These taxes combined with import duties (which are also likely to fall on the rich), make up about 42 percent of tax revenue for low-and middle-income countries (according to the most recent World Bank World Development Indicators), with the rest of tax revenue in these countries coming from consumption taxes, which are typically proportional to consumption.”

• cites World Bank World Development IndicatorsNot matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

A substantial universal basic income in most developing economies would need to be financed via domestic taxation because official development assistance is a small fraction of government budgets.

“For most developing economies, a substantial universal basic income would need to be financed via domestic taxation, because official development assistance is a small fraction of government budgets.”

Social welfare will increase discontinuously for the universal basic income because no targeting costs are incurred.

“As before, social welfare will increase discontinuously for the universal basic income because no targeting costs are incurred in this program.”

• AssumptionRests on an assumption: The Indonesian BLT program and the Peruvian Juntos program each target roughly one-third of the population.

Further reductions in exclusion error come at the cost of much higher inclusion error.

“Further reductions in exclusion error come at the cost of much higher inclusion error.”

• AssumptionRests on an assumption: The Indonesian BLT program and the Peruvian Juntos program each target roughly one-third of the population.

In Indonesia, the socially optimal program targets about 19 percent of the population, with 7.4 percent inclusion error and 58.2 percent exclusion error.

“In Indonesia, the socially optimal program calculated in this way targets about 19 percent of the population, with inclusion error of 7.4 percent and exclusion error of 58.2 percent”

Administrative costs translate to an additional 0.8 and 1.7 percent of the overall transfer budget in Indonesia and Peru, respectively.

“Per year, this translates to an additional 0.8 and 1.7 percent of the overall transfer budget in Indonesia and Peru, respectively.”

• AssumptionRests on an assumption: The Indonesian BLT program and the Peruvian Juntos program each target roughly one-third of the population.

Programs that target narrowly and give large per-person benefits to the poorest achieve higher utility than less narrowly targeted programs, including a universal basic income.

“narrowly targeted programs-those focused on distributing large benefits-per-capita to the poorest of the poorappear to achieve much higher utility levels than less narrowly targeted programs, including but not limited to a universal basic income.”

The proxy-means regressions have an R^2 between 0.53 and 0.66.

“The typical fit we found of these regressions (the R 2 ) is between 0.53 and 0.66”

The tradeoff between true positives and false positives in the ROC curves is similar whether the poverty threshold is set at 0.75 times, 1.0 times, or 1.5 times the government's official poverty line.

“In the Appendix, we compare curves setting y * equal to 0.75× the government's official poverty line, which we call the “extreme poor”; 1.0× the government's official poverty line; and 1.5× the government's official poverty line, or the “near poor.”The tradeoff is similar in all three cases.”

People with high education are ranked as wealthier than their actual income alone would imply.

“The same is true for education: those with high education, and hence high earning power, are ranked as wealthier than their actual income alone would imply.”

Widows appear more likely to be ranked poor even conditional on actual income.

“For example, widows-who have lower earning potential than households headed by men-appear to be more likely to be ranked as poor, even conditional on their actual income level.”

Evidence from Indonesia and Peru shows that existing targeting methods, though imperfect, deliver substantial welfare improvements over universal programs because they transfer more per beneficiary to the poor.

“Our evidence from Indonesia and Peru shows that existing targeting methods in developing countries, while imperfect, appear to deliver substantial improvements in welfare compared to universal programs, because they can transfer much more on a per-beneficiary basis to the poor as compared with universal programs.”

Despite small application costs, self-selection substantially improved screening: beneficiaries selected through application were about 20 percent poorer than those selected through automatic enrollment.

“Yet, despite the relatively small costs, we show that selfselection yielded substantially improved screening; the beneficiaries selected by the application-based method were about 20 percent poorer than those selected through automatic enrollment.”

• cites Alatas et al. (2016)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

A universal basic income will not raise everyone's after-tax income by the same amount because financing it requires increasing marginal tax rates for someone.

“However, it will not raise everyone's after-tax income by that same amount, because to finance the universal basic income out of income taxes, marginal tax rates have to increase for someone.”

• AssumptionThe authors state this as an assumption: A universal basic income program is fully financed by a progressive income tax schedule.

A universal basic income acts as an intercept shifter, raising post-tax income at y = 0.

“The universal basic income acts as an intercept shifter, raising post-tax income at y = 0.”

• AssumptionThe authors state this as an assumption: A universal basic income program is fully financed by a progressive income tax schedule.

Future economic growth alone will not eliminate extreme poverty in developing countries.

“While future growth should continue to reduce poverty, it will not solve the problem by itself-both because a substantial amount of poverty remains even in rapidly growing countries, and because sustained economic growth over decades of the type seen in China is the exception, not the norm ( Jones and Olken 2008).”

• cites Jones and Olken 2008 sourceNot checked — no full text of the source

Targeting may introduce distortions.

“Finally, targeting may introduce distortions.”

In a tax-exempt region, everyone receives the same net transfer regardless of income percentile.

“In the taxexempt region, everyone receives the same net transfer regardless of whether they are in the 5th or 70th percentile of income.”

• AssumptionThe authors state this as an assumption: A universal basic income is financed entirely through income taxes (individual or corporate).

Proxy-means tests are inherently nontransparent because eligibility depends on a weighted sum of many variables that ordinary citizens would struggle to understand even if weights were public.

“A proxy-means test has an inherent lack of transparency, because eligibility is determined based on a weighted sum of many different variables, which would be hard for ordinary citizens to understand even if the weights were public information.”

Proxy-means formulas are kept secret because households may misreport or hide assets if they know the formula.

“the proxy-means test formula (essentially the weight that the coefficients in the regression would give to each asset) are kept secret because if they are known, households (perhaps in cooperation with better-informed agents) may strategically misreport or hide assets to make sure they fall under the cutoff.”

Jensen (2016) documents that the share of employment for which people pay income taxes rises substantially with real per-capita national GDP.

“Jensen (2016) documents that the share of employment for which people pay any income taxes (and therefore for which the government observes income) rises substantially with real per-capita national GDP.”

• cites Jensen (2016)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

The tradeoffs of adding a universal basic income to an income tax schedule depend on how the rest of the schedule is adjusted to satisfy the budget constraint, in terms of redistribution and distortionary effects from increased marginal tax rates.

“From this perspective, the tradeoffs involved in adding a universal basic income to an income tax schedule, therefore, depend on how the rest of the income tax schedule is adjusted to satisfy the government budget constraintboth in terms of overall redistribution and potential distortionary effects from the increased marginal tax rates that introducing a universal basic income will necessitate.”

• AssumptionThe authors state this as an assumption: A universal basic income program is fully financed by a progressive income tax schedule.

A universal basic income can make use of income data collected through the tax system.

“1) it can make use of income data collected through the tax system;”

• AssumptionThe authors state this as an assumption: A universal basic income program is fully financed by a progressive income tax schedule.

As inclusion error rises, total funding available for transfers may rise.

“Thus, as inclusion error increases, the overall amount of funding available for transfers may increase.”

Administrative costs are not a large driver of the benefits level compared to the level of inclusion error chosen.

“Thus, administrative costs are not a large driver of the benefits level, especially in comparison to the level of inclusion error one chooses, which is why the discontinuity at the universal basic income point is visually imperceptible in Figure 4B (although you will be able to see it in Figure 5).”

• AssumptionRests on an assumption: The Indonesian BLT program and the Peruvian Juntos program each target roughly one-third of the population.

To finance a universal basic income of a similar size in a country with many people outside the tax net, marginal tax rates would need to increase substantially more for those inside the tax net.

“However, to finance a universal basic income of a size similar to the case in Figure 1, one would need to increase marginal tax rates substantially more for those relatively few people inside the tax net.”

• AssumptionThe authors state this as an assumption: A universal basic income is financed entirely through income taxes (individual or corporate).

Imperfect targeting using proxy measures for income produces both inclusion and exclusion errors.

“Imperfect targeting using various proxy measures for income leads to both inclusion errors (giving the transfer to those who are not poor) and exclusion errors (failing to deliver the transfer to poor individuals who slip through the cracks in the targeting protocol).”

Social welfare calculations with ρ = 1 and ρ = 5 yield qualitatively similar conclusions to ρ = 3.

“(As a robustness check, we also calculated social welfare with ρ = 1 and ρ = 5, and the conclusions are qualitatively similar.)”

Greater transparency of eligible beneficiary lists may be important for effective implementation of proxy-means test-based schemes.

“The evidence suggests that greater transparency of the lists of eligible beneficiaries may be important to effective implementation of a proxy-means test-based scheme.”

Programs that appear somewhat badly targeted can still produce higher welfare than a universal basic income.

“Even programs that appear somewhat badly targeted could still dominate a universal basic income.”

When the income cutoff for benefits is near zero, horizontal equity violations are mechanically close to zero because almost no households receive the program.

“For programs where the income cutoff for providing benefits is near-zero, horizontal equity violations are (mechanically) close to zero, because almost no households receive the program.”

A universal basic income financed through a combination of income and consumption taxes will look closer to a combination of Figures 1 and 2.

“A universal basic income financed through a combination of an income and consumption tax will look closer to a combination of Figures 1 and 2.”

If financed through proportional or progressive taxation, a universal basic income can still result in substantial redistribution to the poor despite giving the same transfer to everyone including the rich.

“Although universal basic income programs distribute the same value of transfer to everyone, including the very rich, if they are financed through proportional or progressive taxation, they can still result in a substantial redistribution to the poor.”

In countries with large numbers of people outside the tax net, targeting resources toward the poorest requires an alternative data collection approach beyond the income tax system.

“In countries with large numbers of people outside the tax net, any attempt to target resources toward the poorest of the poor requires using an alternative data collection approach beyond the income tax system.”

• AssumptionThe authors state this as an assumption: A universal basic income is financed entirely through income taxes (individual or corporate).

When the total transfer budget is more elastic with respect to the number of beneficiaries, universal transfers become more attractive.

“As the elasticity between the total budget and the number of beneficiaries increases, universal transfers become more attractive (Klasen and Lange 2016).”

• cites Klasen and Lange 2016Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Targeting can improve overall social welfare even when targeting errors are high.

“targeting could be an effective mechanism in improving overall social welfare, even with high levels of targeting error.”

Because income predictions are made with error, targeted systems based on proxy-means tests will feature both inclusion and exclusion errors.

“Because predictions of income, of course, happen with error, targeted systems based on a proxy-means test will feature both inclusion and exclusion errors.”

Horizontal equity violations are mechanically zero for a universal basic income because all households receive the transfer and are treated the same.

“Violations of horizontal equity are also mechanically zero for universal basic income: by definition, if all households receive the transfer, then all households are treated the same.”

Political support for redistribution may be higher when everyone receives some transfer.

“political support for redistribution may be much higher if everyone gets a”

• cites Gelbach and Pritchett 2002 sourceNot checked — no full text of the source

Citizens find it very difficult to verify that a proxy-means test scheme is implemented properly.

“Thus, it is very difficult for citizens to verify that a proxy-means test scheme is being implemented properly.”

A more accurate proxy-means formula with fewer errors may induce more labor supply distortions because some households face a steeper effective tax rate over some income range.

“Perhaps ironically, a more accurate proxy-means test formula, with fewer errors, may actually induce more labor supply distortions, since fewer errors in the proxy-means formula will imply that some households face a steeper effective tax rate over some range.”

If the total budget is fixed, the transfer given per person falls as more people become eligible.

“If the total budget for the program is fixed, then the transfer given per-person falls as we move to the right on this curve.”

• AssumptionRests on an assumption: The Indonesian BLT program and the Peruvian Juntos program each target roughly one-third of the population.

The size of relative targeting errors depends on how often governments collect asset data and how much households move in and out of poverty.

“Thus, the size of the relative errors will also depend on how frequently the government collects asset data from households and how much churn in and out of poverty occurs over time.”

• AssumptionRests on an assumption: Household income status changes over time.
• AssumptionRests on an assumption: Governments conduct targeting surveys infrequently.

Whether workfare-style programs are better or worse than a universal basic income depends on whether the utility costs imposed on beneficiaries are outweighed by targeting cost savings.

“Whether these types of programs-universal in nature, but achieving screening through so-called ordeal mechanisms-are better or worse than a universal basic income program depends on whether the substantial utility costs imposed on beneficiaries (like manual labor in the hot sun in order to receive benefits) are outweighed by the cost savings from improved targeting.”

• AssumptionRests on an assumption: People with better job options tend to avoid manual labor at minimum wage in hot conditions.

Health and education conditions can have targeting effects in addition to incentive effects.

“While much attention has been given to the incentive effects of the health and education conditions, they can also have targeting effects.”

Application processes do not always improve targeting.

“However, application processes need not necessarily always improve targeting.”

Costly take-up can benefit targeting.

“However, having some type of costly take-up can also provide a benefit by helping in the targeting of a program.”

The application process substantially reduces inclusion errors in proxy-means testing while holding the official eligibility cutoff fixed.

“The key seems to be that the application process substantially reduces the inclusion errors in the proxy-means test process while holding the official eligibility cutoff c fixed.”

Almost all conditional cash transfer programs are targeted, most often using proxy-means tests.

“While in principle conditional cash transfers could be either universal (anyone who meets the education and health conditions receives the transfer) or targeted using a proxy-means test, almost all are targeted, most often using proxy-means tests (Fiszbein and Schady 2009).”

• cites Fiszbein and Schady 2009Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

A universal basic income may reduce the take-up problem because the government can in theory send checks automatically to everyone.

“A universal basic income may reduce the take-up problem, because the government can in theory send the checks to everyone automatically.”

If communities target earning potential rather than actual income, the system may be less distortionary than both proxy-means testing and a regular income tax.

“To the extent that communities target based on earning potential rather than actual income, such a system has the potential to be less distortionary not only than a proxy-means test, but also than a regular income tax.”

Richer households are generally more likely than poor households to meet health and education conditions.

“One reason for this is that, in general, richer households are more likely to meet the health and education conditions than the poor.”

• AssumptionRests on an assumption: Richer households are generally more likely than poor households to satisfy health and education conditions.

For many developing countries, the paper's simulations suggest that welfare gains from targeting may be substantial.

“Nevertheless, for many developing countries, our simulations suggest the welfare gains from targeting may be substantial.”

If take-up costs or benefit tradeoffs are higher for the rich than the poor, the poor may be more likely to take up the program.

“If the cost/benefit of taking up a program is higher for the rich than for the poor, the poor may be more likely to take up the program.”

The benefits of adding self-selection to a proxy-means test system may be largest when the nontarget population is large relative to the target population.

“This logic suggests that the benefits from adding self-selection to a proxy-means test system may be largest when the nontarget population is large relative to the target population.”

If some very poor households cannot meet conditions and conditions are enforced, some very poor households may be excluded from conditional programs.

“In particular, if some very poor households are unable to meet the conditions, and if the conditions are actually enforced (which they are not always), then some very poor households may be excluded from the programs.”

When incomes are largely unobserved, universal transfers cannot easily be taxed back as income rises.

“In this setting, universal transfers cannot be easily taxed back as one moves up the income distribution.”

• AssumptionRests on an assumption: In developing countries, incomes are unobserved for very large parts of the population.

An ordeal associated with receiving a program can improve screening between rich and poor.

“This idea-that having an”

• cites Nichols and Zeckhauser (1982)Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Kleven and Kopczuk (2011) sourceNot checked — no full text of the source

Randomized trials of conditional cash transfers, including the original Progresa launch, showed substantial improvements on conditioned indicators initially and over the medium term.

“A number of randomized trials of these programs, including the original launch of Progresa, showed that these programs led to substantial improvements on the conditioned indicators, both initially (for example, see Behrman and Todd 1999;Gertler 2004) and over the medium term (Behrman, Parker, and Todd 2011; Barham, Macours, and Maluccio 2017; Kugler and Rojas 2018; Cahyadi, Hanna, Olken, Prima, S”

• cites Behrman and Todd 1999Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

• cites Gertler 2004 sourceNot checked — no full text of the source
• cites Behrman, Parker, and Todd 2011 sourceNot checked — no full text of the source

Community targeting addresses the perception of a horizontal equity problem because villagers can ensure that a broadly agreed poorer person receives the transfer.

“It also addresses the perception of a horizontal equity problem-from the perspective of the villagers, if it is broadly agreed that person A is poorer than person B, they can ensure that person A receives the transfer.”

Without restrictions limiting conditional programs to the poor, a universal version would be regressive rather than progressive because wealthier households more often meet the conditions.

“Thus, in the absence of explicit additional restrictions limiting these programs to the poor, the fact that the conditions are more likely to be fulfilled by wealthier households would actually make a universal version of these programs regressive rather than progressive.”

• AssumptionRests on an assumption: Richer households are generally more likely than poor households to satisfy health and education conditions.

A complicated application form may discourage less literate or bureaucracy-uncomfortable people and worsen targeting.

“For example, a complicated application form may dissuade those who are less literate or comfortable with bureaucracy from filling it out, leading to worse targeting (Gupta 2017).”

• cites Gupta 2017Not matched

  Couldn’t be matched to a paper in Crossref, so it can’t be checked.

Combining self-selection with a proxy-means test may yield many benefits of self-selection with smaller utility costs to recipients.

“However, it may be possible to obtain many of the benefits from self-selection with much smaller utility costs to recipients if one marries self-selection to a proxymeans test.”

Even if enrollment is theoretically automatic, in practice some take-up decisions and enrollment actions will still be needed.

“This suggests that even if enrollment is in theory automatic, in practice some take-up decisions and enrollment actions will need to occur.”

• AssumptionRests on an assumption: Most developing countries lack a complete, unduplicated list of all individuals.
• AssumptionRests on an assumption: In most developing-country settings, most people lack an address and a bank account.

A key result is that one may offer a transfer to all takers but impose a sign-up cost that is differentially costly for the rich.

“A key result is that one may want to offer a transfer program to all takers, but with the catch that signing up for the program entails a cost that is differentially costly for the rich.”

• cites Kleven and Kopczuk (2011) sourceNot checked — no full text of the source

Community targeting avoids some key transparency and horizontal equity pitfalls of proxy-means testing.

“A community targeting approach avoids some key pitfalls of the proxy-means test related to transparency and horizontal equity.”

Community targeting has the potential to be less distortionary.

“Community targeting also has the potential to be less distortionary.”

The primary downside of targeted programs is horizontal equity: because targeting is imperfect, many poor households slip through the cracks and are excluded.

“The primary downside of these programs is horizontal equity-because targeting is imperfect, there will be a substantial number of poor households who slip through cracks and are excluded.”