The problem the researchers set out to solve is deceptively simple to state. Since 2010, a globally comparable Multidimensional Poverty Index has been published for more than 100 countries by OPHI and the United Nations Development Programme, and more than 55 countries now maintain official national MPIs. At the same time, the international commitment to “leave no one behind” has generated legitimate demand for disaggregated data and for group-specific measures capturing the distinctive deprivations of children, persons with disabilities, ethnic minorities, farmers and other populations. The result, in many countries, is a pile-up of bespoke indices: one multidimensional index for poverty, another for women, another for youth, another for people living with disabilities, and still others combining health and housing indicators with climate hazards, violence or digital literacy. Even when these indices share a common counting-based methodology, their differing indicators and weights mean their policy implications must be studied one by one.

Alkire and her colleagues argue that this proliferation collides with a bottleneck that statisticians often overlook: the finite cognitive and time resources of policymakers. Drawing on classic findings from organizational research, including a well-known observation that in information-saturated environments “attention, rather than information, is the scarce resource,” the authors note that officials expected to absorb multiple indices may simply sideline the ones they cannot master. Advocacy groups promoting group-specific measures may end up competing with each other and with national poverty strategies, diluting their collective impact even when their underlying priorities—education, water and sanitation, housing, decent work—overlap almost completely. A 2026 assessment by the Inter-Agency and Expert Group on Sustainable Development Goal indicators reached a strikingly similar conclusion, warning that the current SDG indicator framework “has proven overly complex for many countries to implement” and urging greater simplicity and focus.

The technical heart of the new paper is the Alkire-Foster counting method, the framework that underlies both the global MPI and national MPIs reported as SDG Indicator 1.2.2. In this approach, each person is first assessed against a deprivation cutoff for every indicator—whether a child attends school, whether a household has clean cooking fuel—producing a deprivation score that is the weighted sum of their deprivations. A second, cross-dimensional poverty cutoff then identifies who counts as poor: anyone whose weighted deprivation score meets or exceeds that threshold is classified as multidimensionally poor. From the resulting censored deprivation matrix, analysts compute the headcount ratio of poverty, its intensity (the average deprivation score among the poor), and the adjusted headcount ratio known as the MPI, which can be broken down by every component indicator and disaggregated across regions, age groups and other characteristics.

The innovation of the Linked MPI is a procedure the authors call the “drawer approach.” Starting from an official National MPI, analysts restrict the population to a focal group—say, all children aged 0 to 17—and then open an additional drawer of group-specific, individual-level indicators. To keep the two measures coherent, the original indicator weights are proportionally scaled down so that they sum to (1 − γ), while the new indicators receive the remaining weight γ, and the poverty cutoff is proportionally adjusted: the linked cutoff equals the original cutoff multiplied by (1 − γ). This proportional rescaling produces a mathematically guaranteed property: every person identified as poor by the National MPI is necessarily identified as poor by the Linked MPI. The converse does not hold—people who live in households that escape national poverty can still be newly identified as poor if they experience enough group-specific deprivations—which is precisely where the added value lies. The two measures also send congruent policy messages on every shared indicator, because those indicators and their relative weights remain the same.

To demonstrate the approach, the team constructed a Linked Child MPI for Nepal using the 2014 Nepal Multiple Indicator Cluster Survey. The original Nepali National MPI spans three equally weighted dimensions; the Child MPI adds a fourth dimension of individual child indicators covering the lifecycle of childhood, so each dimension carries one-quarter of the total weight. For children under five, the new indicators capture early childhood development and nurturing conditions—nutrition, exclusive breastfeeding for infants under six months, immunization for those aged 6 to 23 months, availability of toys and adequate care for toddlers, and time spent with adults reading, telling stories and singing for three- and four-year-olds. For children aged 5 to 13, the indicators are school attendance and child labour; for adolescents aged 14 to 17, they are schooling or working and child labour. Because the national poverty cutoff of one-third becomes one-quarter in the four-dimensional structure, the arithmetic link is preserved exactly.

The empirical results reveal how much a household-based national measure can miss. According to the National MPI disaggregated for children, 33.8 percent of Nepali children are multidimensionally poor. The Child MPI, however, identifies 40.9 percent—a full 7.2 percentage points more—because these additional children live in households that are not nationally poor yet personally experience child-specific deprivations such as child labour or missed schooling. The breakdown is illuminating: 19.1 percent of all children are poor by the National MPI but deprived in no individual child indicator, while 14.6 percent are nationally poor and also experience at least one child-specific deprivation. Only 3.5 percent of children are deprived in a child indicator without being nationally poor. In other words, the child indicators primarily deepen understanding of poverty among children already counted as poor, rather than radically redefining who counts. Disaggregation by gender found slightly higher poverty estimates for girls than boys, though the difference was not statistically significant.

At the provincial level, the two measures told a convergent and complementary story. Both rank Karnali as the poorest province, followed by Madhes, but the child measure reveals sharper priorities: in Karnali the child poverty headcount of 63.7 percent exceeds the national figure of 54.4 percent by 9.7 percentage points, and in Lumbini the gap is 8.6 points. Indicator contributions differ dramatically across provinces. The child development indicator covering nutrition and school attendance contributes more to poverty in Madhes, while early cognitive development and child labour weigh more heavily in Karnali. Living-standard deprivations, particularly the lack of electricity, are most acute in Karnali, whereas educational deprivations dominate in Madhes, and child undernutrition contributes most in Gandaki province. Because the same shared indicators drive both indices, provincial planners can allocate budgets and design integrated programs from a single, internally consistent information platform.

The authors argue that the implications extend well beyond Nepal. The same linking strategy can be applied to women, elderly people, ethnic groups, persons with disabilities or farmers, and can also incorporate entirely new dimensions—environmental hazards, social connectedness, digital literacy, conflict or even monetary poverty—without abandoning the familiar national index. Standard robustness analyses that accompany national MPIs can be applied simultaneously to linked measures, and because the methodology and core indicators are shared, the cognitive load required to grasp a linked measure’s value-added is far lower than for a disjoint alternative. The approach is already feasible in official statistics: Nigeria, Afghanistan and Sri Lanka have experimented with linked national and child MPIs in practice.

The researchers are candid about the limits. Data for both national and group-specific deprivations may not exist; some group-specific deprivations are difficult to measure parsimoniously; and the Nepal illustration, constrained by its survey, does not claim that every age-cohort indicator captures deprivations of equal seriousness—something better achieved with expert and participatory input from children themselves. Intensities and MPI values of the two measures cannot be compared directly because of differing weight vectors, though a simple rescaling restores comparability. Future research, the authors suggest, should explore how multidimensional metrics can evolve into genuine management tools that empower impoverished actors and reward successful evidence-based interventions.

The deeper message resonates with an old “paradox of information supply”: the less a supplier offers, the more it may be used—yet an overly narrow supplier risks overlooking the very characteristics, such as age, disability, minority status and migration, that anti-poverty policy cannot afford to ignore. Linked MPIs, the authors conclude, offer a new balance: they heighten focus on the most vulnerable without losing the information richness that effective policy demands, consolidating rather than dispersing the scarce attention of the people who must act.