Chapter 9. The Digital Revolution and Targeting Public Expenditure for Poverty Reduction
- Sanjeev Gupta, Michael Keen, Alpa Shah, and Genevieve Verdier
- Published Date:
- November 2017
Former Indian Prime Minister Rajiv Gandhi famously commented that only 15 percent of every rupee spent on the public food distribution system reached the poor. Indeed, the widespread view is that anti-poverty programs, and not only those in India, are badly targeted, with massive “leakage” to the nonpoor. Yet the move to cash-transfer programs alongside a burgeoning use of digital technology in recent years is being hailed in India and elsewhere as the solution to this problem of leakage. New digital technologies such as biometrics, goes the reasoning, will improve identification of the poor, and electronic banking will facilitate the transfer of resources to them.
The entrenched problems of so-called leakage notwithstanding, discussions as to the benefits of the digital revolution are prone to hype—perhaps nowhere more so than in policymaking, wherein new information technologies are meant to provide innovative solutions for problems once seen as intractable.
This chapter, taking a contrarian though not a Luddite stance, recognizes the benefits of new technology. But it nonetheless urges caution and a deeper examination of the fundamental trade-offs, as well as the advantages and disadvantages of “fine targeting,” for poverty reduction.1 These trade-offs fall into three major categories: information costs, high marginal tax rates, and political economy. The impact of digitalization on the trade-offs needs to be scrutinized with due reference to the institutions and social norms that structure society.
The chapter begins by setting down the fundamental principles of targeting transfers to minimize poverty, setting the costs of fine targeting against its undoubted benefits.2 It then examines the likely impact of digitalization and new technology on these costs and benefits. And it argues that in some dimensions the trade-offs are quite independent of the use or otherwise of new technology. In other cases the benefits are clear, and in still others the use of technology may worsen the trade-offs.
The Fundamentals of Targeting
We start by specifying the measurement of poverty. Consider a distribution of income, with incomes y. ranging from lowest to highest as i = 1, 2, …n:
Also shown is the poverty line z, with q individuals below the poverty line. The fraction of individuals below the poverty line is thus q/n. The Foster-Greer-Thorbecke class of poverty indices (Foster, Greer, and Thorbecke 1984) Pα is given by
Thus, the proportional poverty gap [z – yi]/z for each person is raised to a power α and summed across the q poor individuals. The parameter α measures the degree of “poverty aversion.” When α = 0, we get the standard head count ratio measure of poverty, also known as the incidence of poverty:
When α = 1 we get the poverty gap measure
When α = 2 the squared poverty gap measure weights the larger shortfalls from the poverty line more severely.
Suppose now that a poverty reduction budget, B, becomes available to the policymaker. Start with a model in which the transfers have no impact on individual incentives to earn income. Further, assume that making transfers has no information or administrative cost. What is then the most effective way of using this budget to reduce poverty? The answer depends on the value of α (Bourguignon and Fields 1990). If α = 0, then the most effective transfer rule is to start with the individual closest to the poverty line, make sufficient transfers to move this individual over the poverty line, then move to the next poorest individual, and so on until the budget is exhausted. If α = 1, then transfer to any individual below the poverty line, so long as the transfer is not so large as to take the individual above the poverty line, is equally effective in reducing poverty P1. But if α = 2, then the effective strategy is the following: start with the poorest individual and make transfers to this individual to bring income up to the next poorest individual, then make transfers to these two individuals until they are brought up to the next highest income, and so on until the budget is exhausted.
The above exercise highlights the key feature of “perfect targeting”—the transfer to each individual is just enough to bring income up to the poverty line, no more and no less. There is no leakage whatsoever to those who are above the poverty line to begin with, and those who were below the poverty line are not given more than is necessary. With this scenario, if the budget available was
then poverty would be eliminated. With imperfect targeting and inadequate budget, poverty would be reduced only partially. For the P1 measure, it can be shown (Fiszbein, Kanbur, and Yemtsov 2014) that the reduction in poverty resulting from the transfers is given by
in which ΔP1 is the reduction in poverty and Tp is the sum of the transfers reaching the poor. The impact of the program on poverty is thus composed of two effects. The first term measures targeting efficiency—the fraction of transfers reaching the poor—while the second term quantifies budgetary adequacy, that is, the ratio of the budget to the poverty gap. The two together produce the poverty reduction we observe. With perfect targeting, the first term is 1. In contrast, an untargeted universal benefit would give an equal amount to everybody in the population, making the first term equal to the incidence of poverty in the population less than one. Fiszbein, Kanbur, and Yemtsov (2014) presents targeting efficiency and budgetary adequacy measures for nearly 50 countries.
It is the scenario of fine targeting, even perfect targeting, which people implicitly have in mind as a benchmark when they criticize a program for being inefficient in reducing poverty. They see individuals above the poverty line, sometimes well above, receiving transfers while payments to the poor are insufficient, and they argue either that poverty could be reduced more with the same budget or that the same poverty reduction could be achieved with a lower budget if targeting could only be finer. Indeed, this is the start of many a discussion between the international financial institutions and ministries of finance on food and fuel subsidy programs, for example.
Fine targeting, if costless, is clearly better than weak targeting or non-targeting if the objective is to minimize poverty with a fixed resource budget. But fine targeting is not costless. As noted, three major categories of issues arise—information costs, high marginal tax rates, and political economy.3
Fine targeting requires “fine” information on individual incomes. In principle, every single income in the economy needs to be assessed and verified, since someone with a high income could still claim to have income below the poverty line. The administrative costs of running such programs, with very detailed participation criteria, have been well discussed in the literature (see Coady, Grosh, and Hoddinott 2004; Grosh and others 2008). This has led to the significant literature on mitigating these informational costs by using easily observable indicators that are correlated with income to undertake contingent targeting. The idea is that income within these categories cannot be distinguished, but statistical properties of the distributions within these groups can be used to design transfer strategies that will be more efficient at poverty reduction than no targeting at all. Thus, for example, in an early exercise, Kanbur (1987) showed that if the objective is to minimize national Pα then transfers should be in proportion to each group’s Pα-1 (for α ≥ 1). An application to food subsidies was provided by Besley and Kanbur (1988) and to land-holding-based targeting by Chao and Ravallion (1989). The literature is now extensive and the basic analytical issues are reasonably well understood.4
One important type of targeting that addresses the information issues is known as “self-targeting.” It does not rely on external assessment and validation to identify the poor and the nonpoor. Rather, it sets up an incentive system such that only those people whose incomes are sufficiently low would come forward to claim the benefit. The best-known example of this is a transfer contingent on employment at a public works site. The employment is guaranteed for all those who present themselves, but clearly only those for whom the opportunity cost is less than the wage at the public works site will show up. If the hourly wage at the public works site is 100, why should someone with an alternative wage rate of 200 elsewhere work at the public site? If the opportunity cost is in turn correlated (negatively) with poverty status, the targeting objective is achieved. An early assessment and validation of this argument is presented in Ravallion (1991). Kanbur (2010) provides a recent application of the argument in the context of the global financial crisis. And analogous reasoning leads to the case for subsidizing coarse rather than fine grains in food subsidies (the nonpoor are more likely to prefer the latter).
Employment guarantee schemes are a type of conditional cash transfer—transfers conditioned on some behavioral response—in this case upon employment at the public works site. However, not all conditional transfers are self-targeting in the same progressive direction (Rodriguez-Castelan 2017). Consider the very popular policy intervention of a cash transfer conditional on keeping children in school. But if education is a normal good, the well-off will keep their children in school in any case—for them the transfer will be a pure inframarginal transfer. For poorer households, who are being incentivized to change their behavior, the value of the transfer will be less than the cash value. Indeed, it may not be worthwhile for the poorest households to participate in the program at all. As Rodriguez-Castelan (2017) shows, the more efficient use of a given budget to reduce poverty may in fact be to provide an unconditional rather than a conditional cash transfer. No doubt the debate on conditional cash transfers will continue, touching also on the issue of cash versus in-kind transfers (Fiszbein and others 2009).
The informational costs of fine targeting, and ways of addressing them, are thus well understood and assessed in the literature. However, a far less-well understood implication of fine targeting exists. Recall once again the description of perfect targeting—every individual gets just enough to rise to the poverty line. Thus, higher-income poor individuals get less transfers. In fact, with perfect targeting the reduction is one for one: for every unit increase in pre-transfer individual income, government transfer declines by one unit (the sum of the pre-transfer income and the transfer is equal to the poverty line). What we have described, in effect, is an effective tax and transfer schedule with a marginal tax rate of 100 percent! In any normal context, the incentive effects of this would be prominently discussed but not, seemingly, in the context of anti-poverty programs.
In any sort of standard labor supply model, a 100 percent marginal tax rate over an extended range could lead to zero labor supply. If this happened and pre-transfer incomes below the poverty line fell to zero, the cost of the transfer would rise significantly to
Thus, in a setting in which incentives for earning income are important, the incentive effects of high marginal tax rates implicit in fine targeting will have to be balanced against the targeting gains. Kanbur and Keen (1989) and Kanbur, Keen, and Tuomala (1994) present early analyses of this issue, showing the extent to which fine targeting will have to be mitigated in the face of these incentive issues. More recently, Banerjee and others 2015 brings up-to-date quantification of the labor supply effects of transfer programs.
This discussion of incentives relates to the earlier discussion of informational costs and the use of easily observable indicators. The tension between fine targeting and high marginal tax rates can be mitigated if we are allowed to use easily observable characteristics to design different transfer schedules for different groups. The extra instrument of “categorical” targeting allows better use of resources, as shown by Immonen and others (1998).
The third dimension of the costs of fine targeting is perhaps most elusive—political economy. Fine targeting means, by definition, confining transfers to the poor. But this means a separation of the interests of the poor from middle-income groups. As the political economy of this plays out, such separation could mean lower overall budgets for poverty reduction transfers. In the words of Gelbach and Pritchett (2002), more for the poor could end up being less for the poor. In an early allusion to these forces, Anand and Kanbur (1991) referred to the reform of the Sri Lankan rice subsidy in the late 1970s, which went from a universal subsidy to one targeted to those below the poverty line. But the real value of the subsidy then fell over subsequent years, with little in the way of political repercussions. Kanbur (2010) also alludes to these forces in discussing the efficacy of employment guarantee schemes as vehicles for rapid response to macro shocks—while their targeting properties are beneficial, political support at the local level may be problematic for this very reason. Gelbach and Pritchett (2002) present a formal model of the forces in play.
Implications of the Digital Revolution
Thus, fine targeting, while obviously a good thing if it can be implemented costlessly, is not in fact costless. As noted, trade-offs exist in information, incentives, and political economy. This section assesses whether the digital revolution can enhance the benefits of targeting while reducing the costs.
Digitalization and Anti-Poverty Programs
What does the digital revolution mean in the specific context of targeting anti-poverty transfer programs, as opposed to the general implications for state capacity? Digitalization is thought to be helpful in at least three ways. The first is ease of payment of cash. Radcliff (2016, 2017) provides a specific example:
The link between payment access and fuel subsidy reform was powerfully demonstrated by Iran’s reform efforts in 2010–11. Then, the Iranian government was spending $70 billion per year on fuel subsidies—a clearly unsustainable subsidy bill. But the government couldn’t raise fuel prices without offering citizens something in return, lest it face a political backlash. So, it decided to replace fuel subsidies with cash transfers, setting aside $30 billion to deliver $40 per month to every citizen. … To make the reform possible, the Iranian Government had to deliver monthly payments to every Iranian household. Today, 67 percent of Iranian adults receive a government payment—higher than any country in the world—and 92 percent of these payments are delivered digitally into an account.
The second is biometric identification, as Gelb and Diofasi (2015, p. 61) discuss for South Africa:
Provincial governments in South Africa have used fingerprint-based biometric ATMs and smartcards since the mid-1990s to deliver pensions and social grants, including in locations with limited connectivity. … Biometric re-registration of over 20 million social grant recipients was completed in 2013 by the South African Social Security Agency in an effort to streamline the recently centralized system. Even though the system had been able to draw on an extensive identity infrastructure initiated during the apartheid period re-registration enabled SASSA [South African Social Security Agency] to remove 650,000 social grants going to non-eligible individuals which resulted in savings of over $65 million annually. … The new system also ensures that payments cease once a beneficiary has died without having to rely on death registration records: all grant recipients must present a “proof of life” once a month by scanning their fingerprints or through voice recognition.
The third is keeping track of payments at the next level up, in the government system itself. Here is how Banerjee and others (2016, 2011) reported the results of a recent study in India:
In collaboration with the Government of Bihar, India, we conducted a large-scale experiment to evaluate whether transparency in fiscal transfer systems can increase accountability and reduce corruption in the implementation of a workfare program. The reforms introduced electronic fund-flow, cut out administrative tiers, and switched the basis of transfer amounts from forecasts to documented expenditures. Treatment reduced leakages along three measures: expenditures and hours claimed dropped while an independent household survey found no impact on actual employment and wages received; a matching exercise reveals a reduction in fake households on payrolls; and local program officials’ self-reported median personal assets fell.
Digital technologies thus hold clear benefits for social programs and, as the costs of digitalization decline, the benefit-to-cost ratio will continue to improve along these dimensions. But notice that the three examples are all somewhat independent of the issue of anti-poverty targeting as presented in the previous section, ensuring that transfers flow only to people living below the poverty line. The Iran example is one where a poorly targeted fuel subsidy program was replaced with a completely untargeted cash transfer program. The problem in the reform was how to make the cash transfer to every household, not how to restrict the transfer only to poor households. The issue in South Africa is how to identify those who meet pension eligibility requirements (basically, age and gender), not how to target flows to the poor. And in the Indian case, targeting to the poor is being taken care of by the self-targeting nature of the public works programs; the issue addressed by digitalization is the standard one of public sector corruption.
So, we return to the fundamentals of targeting, with its issues of (1) the need for information on income or consumption of individuals for fine targeting, (2) the incentive effects of fine targeting, and (3) the political economy of fine targeting. How, if at all, can the digital revolution help ease the trade-offs identified?
Clearly, the most obvious entry point is the potential of the digital revolution to reduce information costs in targeting. Biometrics and identification of individuals is often put forward as the solution to the information problem in targeting. However, what fine targeting needs is not just unique identification of individuals, but detailed information allowing computation of income or consumption and, thus, identification as poor. Further, this computation needs to be updated annually if the program is to continue to be finely targeted. In small, developed, and highly formalized economies, such as Finland’s, such income information is already digitized and linked in to other national databases, and the use of such information is not a problem.5 But in a developing country with a large informal untaxed sector it is not clear how exactly digitalization can help, at least not for many years to come. And it does not seem that informality is declining sharply or at all in many developing countries (Kanbur 2015).
In the absence of detailed income or consumption data for individuals or households, correlates derived from household surveys can be used to fashion a “proxy means test,” as described in the previous section. This provides ground-level implementers of the transfer program with a formula that weights several observable characteristics, and a cut-off value of this weighted sum to identify households to whom the transfer should be made. But this then requires information on these observable characteristics for each household (and in some cases individuals within the household) to be obtained, maintained, and updated at the local level. Clearly, digitalization can help enormously in maintaining and updating these data sets. Yet, verification and validation of some of this information, going beyond births and deaths where digitalization of vital statistics is a complementary input, is not a simple, straightforward technical exercise. Quality of housing is often an element in proxy means tests—whether the house has a tin roof, for example. But this is a subjective assessment—how is a tin roof with holes in it to be counted? Whether the man in a household is employed is another typical criterion, an ambiguous one in rural and agricultural settings, and so on. These issues are not amenable to easy resolution by digitalization.
Marginal Tax Rates
Let us turn now to the basic tension between fine targeting and the high implicit marginal tax rates that fine targeting entails. The tension arises in the attempt to make sure that no poor person gets more than needed to reach the poverty line, which is an important element of fine targeting. The tension can be resolved by giving up on this requirement and going for a universal benefit, but then we are at the other extreme of fine targeting—no targeting at all.
However, with a universal benefit there is leakage to those above the poverty line. This leakage can be reduced by conditioning the transfer on individual characteristics through a proxy means test, as has been discussed. But everyone with the same value of the proxy is treated identically—there is “universalism” among those in the same observable category. So, some of these are getting more than they need to reach the poverty line. There is leakage, but since all incomes within a category receive the same transfer, not conditioned on income, there is not an implicit marginal tax rate. Thus, to the extent that new information technology helps manage differential transfers to several groups differentiated by observable characteristics, universalism within each group avoids the high marginal tax rate issue, while proliferation and optimal use of group information allows better targeting.
Political Economy, Norms, and Institutions
Consider now the political economy dimension. The central issue posed in the previous section is that of the separation of the interests of the poor from the middle-income groups. The resources needed for poverty reduction, even with fine targeting, will have to come from somewhere. The question then is at what income level does the switch between net recipient and net contributor occur? With perfect targeting, the answer is obvious—it occurs at the poverty line. Those above the poverty line must pay for transfers to those below the poverty line. With no targeting at all, everybody gets an equal amount, but this has to be paid for. The switch point depends on the exact nature of the tax schedule and the income distribution. But if the universal transfer is z, enough to make even the lowest income come up to the poverty line, then the switch point will be above the poverty line. Thus, with a high enough universal benefit, the poor and the lower-middle classes have a common cause.
How is the above argument affected by digitalization? On the face of it, not at all. The issue is where the switch point between net gainers and net losers occurs, and it is not clear why digitalization should affect this in general. However, suppose now that over and above the costs of the transfer there are operational costs, and that somehow the costs of the whole transfer operation are lowered by digitalization—the leaky buckets are plugged better, so that fewer resources need to be extracted from the net payers. Then the switch point will rise and more of the middle-income groups will be brought into solidarity with the poor. But such political economy analysis, and the implications of digitalization, is in its infancy and needs to be explored further in a complete model that solves simultaneously for the parameters of the tax-transfer schedule as part of the political economy equilibrium, following on from the work of Gelbach and Pritchett (2002).
Suppose now that we are in the realm of proxy means tests, where transfers are conditioned not on income but on observable characteristics. A political economy framework would now see advantage in a coalition of those with common observable characteristics, combined with agitation by this coalition for increased transfers to those characteristics. The politics of caste coalitions, and the demands for reservations of government posts and state transfers, are of this nature in India.6 A theoretical analysis of community-based targeting is provided in Dasgupta and Kanbur (2005). Thus, while the new information technology makes it easier to develop ever more sophisticated proxy means tests, it may at the same time introduce new and perhaps unintended elements to the political economy of a country by intensifying the logic of group coalitions, fueled now by the prospect of trans-fers to the group from anti-poverty programs. Such political economy models also need further development and exploration.
Finally, another aspect of the political economy dimension of fine targeting is perhaps less well understood in the analytical economics literature (it was not addressed, for example, by Besley and Kanbur 1993), although it is well recognized by those with ground-level implementation experience. This aspect, the use of proxy indicators in targeting, links directly to the informational aspects of targeting and the use of proxy indicators in targeting. A well-developed methodology now exists for deriving these proxies, as weighted sums of observable indicators, and the methodology is being implemented in many parts of the world (see, for example, Nazara and Rahayu 2013). But, by definition, these proxy methods will have inclusion and exclusion errors. These are weighted appropriately in a loss function in deriving the optimal proxy, but outcomes on the ground are a different matter.
The complicated proxies, derived by technocrats, are not easy to explain to ordinary people, who put the non-receipt or receipt of a transfer across households down to political and ethnic connections, thereby undermining solidarity at the local level. Adato (2011) notes in a discussion of conditional cash transfers in Nicaragua how purely quantitative analysis can lead policymakers astray:
The survey found that the program was well targeted, with under-coverage rates of 3 to 10 percent. The qualitative research found, however, that people saw themselves as “all poor” and did not understand why households were selected into or out of the program, resulting in several types of stress and tension in the communities.
Such experiences are found in many qualitative studies. Thus, one female recipient of the Program Keluarga Harapan transfer program in Indonesia is reported in Reality Check Analysis (2015, 29) as follows:
… she was originally selected via a household survey where she was asked her name, house condition, how much land she owned and her employment. However, she said others got bigger allowances, some as much as 1 million [rupiah], “because it was unfairly decided by the last kepala desa—you had to be connected to him.” The last elections she has voted for a family member to ensure that she will benefit in the future.7
Thus, if new technology drives implementation of ever more complicated proxy means tests they may end up worsening tensions at the local level even as they satisfy “better targeting” from a technocratic perspective. More generally, the fineness of the targeting may be unrelated to local conceptions and norms of those who deserve the transfer and those who do not, leading to a disconnect between technocrats believing they are doing well while the local political economy suffers. At the very least, then, qualitative analysis will be needed to identify these ground-level repercussions of fine targeting.
A broader type of institutional issue emerged on a field trip the author made a decade ago to Adilabad district in what was then the state of Andhra Pradesh in India. In discussions on the newly introduced National Rural Employment Guarantee Act, we were told about the virtues of the new electronic system of payment. In “the bad old days,” the muster rolls were used to make physical payment to workers at the public works site. It was alleged, no doubt correctly, that this led to a lot of corruption, with only a part of the payment being handed over, the rest being kept by powerful local interests—their henchmen being the ones who were handing out the payments. But now each worker was required to open an account at a local financial institution and the payment would be made directly into the account electronically, thereby circumventing opportunities for corruption at the public works site. Thus, all seemed well. But further inquiries and private conversations with the workers revealed that, in fact, the henchmen now gathered outside the post office or local bank to collect their take and the counter staff of the financial institutions also took their share before handing the cash to the workers.
This anecdotal evidence is not inconsistent with the rigorous experimental evidence provided by Banerjee and others (2016) on the National Rural Employment Guarantee Act in the Indian state of Bihar. While overt corruption, like skimming off the payroll before it is handed out, may be reduced by various forms of e-governance, there is nothing to stop the skimming from happening outside the gaze of the electronic eye. It is encouraging that Banerjee and others (2016) find that “local program officials’ self-reported median personal assets fell.” But the general point remains, that digitalization can only do so much to address deep-seated norms and practices that reflect longstanding power relations in society.
Sounding a note of caution, this chapter has examined whether the digital revolution can help mitigate the significant trade-offs in aiming for fine targeting of anti-poverty transfers to the poor and only to the poor. Many studies have identified the trade-offs and ground-level experience in the implementation of several programs, as well as conventional empirical evidence, confirm and highlights them.
Clearly, new information technology can help mitigate some of the informational and administrative costs of targeting—one of three trade-offs scrutinized here—for example by facilitating the maintenance and updating of local databases on individual characteristics on which targeting relies. But the system is only as good as the information put into it. Income information is problematic in countries with a significant informal sector, and the issues of ground-level assessment and validation of the proxies used in proxy means tests are not overcome by new technology, which can only record and process the information once it has been generated.
To the extent that digitalization can allow better use of observable and unalterable individual characteristics to segment the population into groups, across which there is a variation of transfer but within each of which there is the universalism of an identical transfer, this can help to mitigate the tension between fine targeting and high implicit marginal tax rates, the second of the chapter’s main issues.
The political economy of targeting, meanwhile, relates to the relationship between the nature of the targeting (fine or not) and the resources that the political system will generate for poverty reduction. It is not clear that new technology will affect this trade-off greatly. However, proxy means targeting and group-based targeting can create new tensions on the ground and new forms of political economy based on the groups used for targeting. These unintended consequences will have to be taken on board in any assessment of the consequences of easier group-based targeting because of the digital revolution.
Thus, the hype around the digital revolution needs to be duly mitigated by the lessons that the conventional literature presents on targeting anti-poverty transfers.
Adato, Michelle. 2011. “Combining Quantitative and Qualitative Methods for Program Monitoring and Evaluation: Why Are Mixed-Method Designs Best?” World Bank Other Operational Studies 11063, World Bank, Washington, DC.
Anand, Sudhir, and RaviKanbur. 1991. “Public Policy and Basic Needs Provision in Sri Lanka.” In The Political Economy of Hunger, Volume III: Endemic Hunger, edited by J.Dreze and A.Sen. Oxford: Clarendon Press.
Banerjee, Abhijit, EstherDuflo, ClementImbert, SanthoshMathew, and RohiniPande. 2016. “E-Governance, Accountability, and Leakage in Public Programs: Experimental Evidence from a Financial Management Reform in India.” NBER Working Paper 22803, National Bureau of Economic Research, Cambridge, MA.
Banerjee, Abhijit, RemaHanna, GabrielKreindler, and Benjamin A.Olken. 2015. “Debunking the Stereotype of the Lazy Welfare Recipient: Evidence from Cash Transfer Programs.” MIT Working Paper, Massachusetts Institute of Technology, Cambridge, MA.
Besley, Timothy, and RaviKanbur. 1988. “Food Subsidies and Poverty Alleviation.” Economic Journal 98 (392): 701–19.
Besley, Timothy, and RaviKanbur. 1993. “The Principles of Targeting” In Including the Poor, edited by M.Lipton and J.Van der Gaag. Washington, DC: World Bank.
Bourguignon, Francois, and GaryFields. 1990. “Poverty Measures and Anti-Poverty Policy.” Recherches Economique de Louvain56 (3/4): 409–27.
Chao, Kalvin, and MartinRavallion. 1989. “Targeted Policies for Poverty Alleviation under Imperfect Information: Algorithms and Applications.” Journal of Policy Modeling11: 213–24.
Coady, David, MargaretGrosh, and JohnHoddinott. 2004. Targeting of Transfers in Developing Countries: Review of Lessons and Experience. Washington, DC: World Bank and International Food Policy Research Institute.
Daley, Suzanne. 2015. “Speeding in Finland Can Cost a Fortune, If You Already Have One.” New York Times, April 25.
Dasgupta, Indraneel, and RaviKanbur. 2005. “Community and Anti-Poverty Targeting.” Journal of Economic Inequality3 (3): 281–302.
Fiszbein, Ariel, NorbertSchady, FranciscoH.G.Ferreira, MargaretGrosh, NiallKeleher, PedroOlinto, and EmmanuelSkoufias. 2009. “Conditional Cash Transfers: Reducing Present and Future Poverty.” World Bank Policy Research Report, World Bank, Washington, DC.
Fiszbein, Ariel, RaviKanbur, and RuslanYemtsov. 2014. “Social Protection, Poverty and the Post-2015 Agenda.” World Development61: 167–77.
Foster, James, JoelGreer, and ErikThorbecke. 1984. “A Class of Decomposable Poverty Measures.” Econometrica3 (52): 761–66.
Gelb, Alan, and AnnaDiofasi. 2015. “Preliminary Discussion Paper on the Future of Identification and Development.” Discussion Paper, Center for Global Development, Washington, DC.
Gelbach, Jonah, and LantPritchett. 2002. “Is More for the Poor Less for the Poor? The Politics of Means-Tested Targeting.” The B.E. Journal of Economic Analysis & Policy2 (1): 1–28.
Government of India. 2017. Economic Survey 2016–17. New Delhi: Ministry of Finance, Department of Economic Affairs.
Grosh, Margaret, C.Del Ninno, E.Tesliuc, and A.Ouerghi. 2008. For Protection and Promotion: The Design and Implementation of Effective Safety Nets. Washington, DC: World Bank.
Immonen, Ritva, RaviKanbur, MichaelKeen, and MattiTuomala. 1998. “Tagging and Taxing: The Optimal Use of Categorical and Income Information in Designing Tax/Transfer Schemes.” Economica65 (258): 179–92.
Kanbur, Ravi. 1987. “Measurement and Alleviation of Poverty: With an Application to the Impact of Macroeconomic Adjustment.” IMF Staff Papers, International Monetary Fund, Washington, DC.
Kanbur, Ravi. 2010. “Macro Crises and Targeting Transfers to the Poor.” In Globalization and Growth: Implications for a Post-Crisis World, edited by MichaelSpence and DannyLeipziger. Washington, DC: World Bank.
Kanbur, Ravi. 2015. “Informality: Causes, Consequences and Policy Responses.” CEPR Discussion Paper 10509, Centre for Economic Policy Research, London.
Kanbur, Ravi, and MichaelKeen. 1989. “Poverty, Incentives and Linear Income Taxation.” In The Economics of Social Security Reform, edited by A.Dilnot and I.Walker. Oxford: Oxford University Press.
Kanbur, Ravi, MichaelKeen, and MattiTuomala. 1994. “Labor Supply and Targeting in Poverty Alleviation Programs.” World Bank Economic Review8 (2): 191–211.
Kanbur, Ravi, MichaelKeen, and MattiTuomala. 2016. “Groupings and the Gains from Targeting.” Research in Economics70: 53–63.
Nazara, Suahasil, and SriKusumastutiRahayu. 2013. “Program Keluarga Harapan (PKH): Indonesian Conditional Cash Transfer Programme.” Policy Research Brief 42, International Policy Centre for Inclusive Growth, Brasilia.
Radcliff, Dan. 2016. “Digital Payments as a Platform for Improving State Capacity.” Background Paper, Center for Global Development, Washington, DC.
Ravallion, Martin. 1989. “Land-Contingent Policies for Rural Poverty Alleviation.” World Development17 (August): 1223–33.
Ravallion, Martin. 1991. “Reaching the Rural Poor through Public Employment: Arguments, Evidence, and Lessons from South Asia.” World Bank Research Observer6 (2): 153–76.
Reality Check Approach + Project Team. 2015. People’s Views and Experience of the National Social Assistance Programmes. Jakarta: Effective Development Group.
Fine targeting refers to the targeting of benefits to the poor and only to the poor.
While the chapter focuses on poverty reduction as the objective, it should be clear that the issues raised apply as much to more general objectives as well.
These categorizations were formulated in Besley and Kanbur (1993).
An amusing account of this phenomenon is to be found in this report of a speeding fine in Finland: “The fines are calculated based on half an offender’s daily net income, with some consideration for the number of children under his or her roof and a deduction deemed to be enough to cover basic living expenses, currently 255 euros per month. … Then, that figure is multiplied by the number of days of income the offender should lose, according to the severity of the offense. … In today’s digital age, however, a few seconds is all it takes for the police, using mobile devices, to get information directly from the Finnish tax office” (Daley 2015).
The literature on this topic is vast. For an article targeted toward the generalist see The Economist. “Affirmative Action: Indian Reservations.” June 29, 2013.
Kepala desa is village head. IDR 1.0 million is about $75.0.