Syndicate content

David McKenzie's blog

Can predicting successful entrepreneurship go beyond “choose smart guys in their 30s”? Comparing machine learning and expert judge predictions

David McKenzie's picture

Business plan competitions have increasingly become one policy option used to identify and support high-growth potential businesses. For example, the World Bank has helped design and support these programs in a number of sub-Saharan African countries, including Côte d’Ivoire, Gabon, Guinea-Bissau, Kenya, Nigeria, Rwanda, Senegal, Somalia, South Sudan, Tanzania, and Uganda. These competitions often attract large numbers of applications, raising the question of how do you identify which business owners are most likely to succeed?

In a recent working paper, Dario Sansone and I compare three different approaches to answering this question, in the context of Nigeria’s YouWiN! program. Nigerians aged 18 to 40 could apply with either a new or existing business. The first year of this program attracted almost 24,000 applications, and the third year over 100,000 applications. After a preliminary screening and scoring, the top 6,000 were invited to a 4-day business plan training workshop, and then could submit business plans, with 1,200 winners each chosen to receive an average of US$50,000 each. We use data from the first year of this program, together with follow-up surveys over three years, to determine how well different approaches would do in predicting which entrants will have the most successful businesses.

Weekly links Jan 19: soft skills and maybe a robot can’t take your job after all, the Starbucks indicator of Indian middle class growth, high fees are deterring citizenship, and more...

David McKenzie's picture
  • On VoxEU, using Yelp data to track the local economy.
  • Ted Miguel on plans for long-term follow-ups of child health and cash transfer programs.
  • Priced out of citizenship? From Stanford News, with the cost of U.S. naturalization now $725, an experiment gave vouchers to cover these costs to low-income immigrants in NYC and found naturalization application rates rose 41%.
  • David Deming in the NBER reporter on the value of soft skills in the labor market: “the very term soft skills reveals our lack of understanding of what these skills are, how to measure them, and whether and how they can be developed... Social interaction is perhaps the most necessary workplace task for which there is currently no good machine substitute... Researchers ought to stop relying on convenient, off-the-shelf measures of soft skills and start creating metrics that are theoretically sound and suitable for the task at hand”

Weekly links Jan 12: Big Thinkers brought down to size, can you beat the World Bank at predicting poverty? Chinese minimum wage rises all get spent, three job openings, and more…

David McKenzie's picture
  • Duncan Green summarizes Stefan Dercon’s view of 10 top thinkers in development. E.g. on Acemoglu and Robinson “their policy advice is just ‘buy yourself a better history/don’t start from here’. Not very useful for aid”. Alice Evans responds to the lack of women on Stefan’s list with five big problems in development and female scholars to learn from on these.
  • How did Chinese consumption respond to changes in the minimum wage? Dautovic and co-authors on VoxEU report that “For the period 2002-2009, we identify more than 13,874 changes in the local minimum wage across China's 2,183 counties and 285 cities…many counties experienced substantial nominal increases in their minimum wage above 20%...we show that low-income households spend their entire additional income from a higher minimum wage…for poorer households, 40% of the additional minimum wage income is spend on health care and educational expenditure”
  • Looking to try out machine learning for poverty prediction? The World Bank has launched a competition (with prize money) to see how well you can predict poverty.

Six Questions with Mark Rosenzweig

David McKenzie's picture
Mark Rosenzweig is Frank Altschul Professor of International Economics at Yale University, and was one of the original leaders in bringing theory and micro-level data to addressing development questions. We caught up with him after a recent symposium, which honored his achievements, and celebrated him turning 70 and continuing to produce important new work.

Statistical Power and the Funnel of Attribution

David McKenzie's picture

Often there are many steps or stages between the starting point of some intervention and its ultimate goal, and at each step, people can drop out. The result can be extremely low power to measure impacts on this end outcome, even though we might be able to detect impacts on the intermediate steps. This post illustrates this point, with the goal of making clear the importance of trying to measure intermediate outcomes, and concludes with suggestions of ways to partially overcome this problem.

Weekly links Jan 5: papers you should have read last year, how to measure early childhood development 147 ways, move people to where the jobs are, and more…

David McKenzie's picture

Weekly links December 15: non-frivolous frivolous expenses, Indian internal borders, aspirations, and much more…

David McKenzie's picture

Weekly links November 17: What’s new in trade research, fungibility is too painful to think about, an employment program that worked, and more…

David McKenzie's picture

Weekly links November 10: how to properly pre-register, trade and inequality, surprising findings and more…

David McKenzie's picture
  • Data Colada on how to properly pre-register a study: “it may be helpful to imagine a skeptical reader of your paper. Let’s call him Leif. Imagine that Leif is worried that p-hacking might creep into the analyses of even the best-intentioned researchers. The job of your preregistration is to set Leif’s mind at ease. This means identifying all of the ways you could have p-hacked – choosing a different sample size, or a different exclusion rule, or a different dependent variable, or a different set of controls/covariates, or a different set of conditions to compare, or a different data transformation – and including all of the information that lets Leif know that these decisions were set in stone in advance”…but on the other hand “it should contain only the information that is essential for the task at hand… We have seen many preregistrations that are just too long… you don’t need to say in the preregistration everything that you will say in the paper. A hard-to-read preregistration makes preregistration less effective…” – comes with a nice example table of what bad specifications and good specifications look like.

Weekly links November 3: posters against stunting, but are RCTs bad for kids? Publishing lab experiments and replications, and more…

David McKenzie's picture

Pages