The underlying premise is that while a person with one bee sting is highly motivated to get it treated, one with multiple bee stings does not have much incentive to get one sting treated, because the others will still throb. In other words, more of a painful or undesirable thing one has (i.e. the poorer one is), the less likely one is to do anything about any one problem. Or poverty is less a matter of having few goods than having lots of problems.
For example, randomized control trials around the world on improving learning outcomes have found that stand-alone interventions fail to deliver desired results. Researchers from MIT's Poverty Action Lab have found that reduction in class size not complemented with other measures like school committee empowerment and performance tracking of students, fails to yield any significant results.
If this insight is assumed true, then there are formidable limitations on the relative effectiveness of specific policy interventions in achieving their objectives. It may also help partially explain the less than satisfactory response from poor people to individual social policy interventions that encourage people to improve cleanliness and hygiene (use community toilets) or parents to send children to school or vaccinate infants or poor manage their finances optimally.
What is the incentive to overcome the inertia and expend effort to keep toilets clean when the entire neighbourhood is dirty? Where is the motivation to immunize the child when he/she is living amidst unhygenic surroundings and falls sick every week? Where is the motivation to use mosquito bed-nets when if mosquitoes do not catch you, bacterial infections will lay you down? What is the incentive to eat more expensive healthy food, when disease vulnerability is very high and incidence so commonplace?
In other words, a package of policies that incentivize use of mosquito bed-nets, construction of individual toilets, immunization, awareness creation on child nutrition and personal hygiene, and so on may be more effective than piece-meal health and sanitation interventions. Or a bundled financial product that combines savings account with health insurance, education scholarships, and pensions may generate greater savings outcomes than separate products.
Further, if the bee-sting theory reasoning is assumed to hold, there is a strong likelihood that the amount of incentives required to produce a desired behavioral change would be much larger with a single intervention than with a package of interventions. For example, the financial incentive to get children to school or parents to immunize children would be much larger if these interventions were made separately than if made as part of a bouquet. Therefore, an economically efficient approach may be to bundle a number of interventions together and incentivize them with a large enough cash transfer (which would be less than the sum of all the individual incentives).
All this means the likelihood of considerable variations in the relative effectiveness of individual policy interventions when implemented separately and when done as part of a package. Therefore, poverty eradication policies may be more effective if they address the challenge on multiple dimensions instead of piecemeal and one-off affairs. All this point towards the need for a "big push" in poverty eradication, including close co-ordination across departments.
An ideal opportunity to assess this hypothesis was the Millennium Villages Project (MVP) experiment. Unfortunately, as the recent controversy surrounding Michael Clemens and Gabriel Demombynes questioning the effectiveness of MVP in the absence of rigorous impact evaluation indicates, that opportunity may have been lost. They write,
"Comparing trends at the MVP intervention sites in Kenya, Ghana, and Nigeria to trends in the surrounding areas yields much more modest estimates of the project’s effects than the before-versus-after comparisons published thus far by the MVP. Neither approach constitutes a rigorous impact evaluation of the MVP, which is impossible to perform due to weaknesses in the evaluation design of the project’s initial phase. These weaknesses include the subjective choice of intervention sites, the subjective choice of comparison sites, the lack of baseline data on comparison sites, the small sample size, and the short time horizon."
They oppose the wisdom of spending huge amounts and expanding MVP without having access to objective evaluation of the its impact. See Jeff Sachs response here and a summary of the debate here.