Musings on the Science of “Scaling”: Blum Center U.C. Berkeley

No subject brad delong gmail com Gmail

This is not at all the so-called “replication crisis”.

The devices built work as assessed by all engineering yardsticks: cheap, easy to maintain, rugged, and simple to operate. The interventions conducted work as assessed by all social science standards: they pass gold-standard RCT tests with effects that are statistically and substantively significant.

And yet…

“Scaling” is very hard…

My largely uninformed and probably wrong view is that it has everything to do with organizational and systematic robustness. In the engineering lab and in the social science RCT 98% of things go right. But out in the real world societal capabilities vary: while Toyota can hit six nines–99.9999%–at times I think U.C. Berkeley is lucky to hit nine sixes, and Berkeley is, in global context, a relatively functional organization. So: engineering and societal organization institutions designed for robustness, to degrade gracefully when you cannot attain the degree of organizational tautness of a Toyota. How do we do that? That, I think, is the BIG QUESTION here.

(And, of course, if we can attain the degree of organizational tautness of a Toyota, we no longer have a problem of economic development in any sense, do we?)

Blum Center: The Science of Scaling: Building Evidence to Advance Anti-Poverty Innovations:

September 26 @ 8:00 am – 5:00 pm
100 Blum Hall, Haviland Road
U.C. Berkeley
Berkeley, CA 94720 United States

The Development Impact Lab (DIL), headquartered at UC Berkeley and funded by USAID…

…has developed a “Development Engineering (Dev Eng)”… interdisciplinary framework for designing and testing new povertyalleviation and economic growth technologies in the field… encourag[ing] researchers to build scal[ing] into the R&D process, from the beginning. Yet… there are few generalizable mechanisms for scaling evidence-based interventions in emerging markets…. [Thus] DIL[s]… annual State of the Science conference [is] on The Science of Scaling:

The conference will bring together academic researchers, development practitioners, technology developers, and investors to review the evidence on scaling successful anti-poverty innovations…. Are there proven methods for technology transfer from university to government agencies and non-governmental organizations? Why do some products and interventions scale quicker than others? What facilitates the adoption of new technologies by end-users? This event will explore these questions and help articulate a research agenda for the “Science of Scaling”…

http://delong.typepad.com/2016-09-26-08.3057-scanner-pro.pdf

2016 09 26 08 3057 Scanner Pro pdf 1 page

Must-Read: Dietrich Vollrath: The Persistence of “Technology”

Must-Read: If you have not been reading Dietrich Vollrath’s weblog on economic growth, you should. He has been teaching the world a masterclass in understanding the patterns and determinants of economies’ long-run growth trajectories:

The Persistence of Technology Dietrich Vollrath

Dietrich Vollrath: The Persistence of “Technology”: “Diego Comin, Bill Easterly, and Erick Gong… ‘Was the Wealth of Nations Determined in 1000BC?…

…there is a surprising amount of explanatory power in technology measures from 1500AD….

CEG document… that technology levels are incredibly persistent…. CEG… pull out binary measures of technological use for different ethnic/cultural groups. Did your group use wheeled wagons in 1500? Yes? You get a 1. Did you use paper? No? You get a zero. Did you produce steel? No? You get a zero. Average these 0/1 measures across the different measures of technology, and you get an overall score…. Their Figure 2 gives you essentially the whole thrust of the paper. They use the ethnic group technology measures, assign each country a technology level based on the highest scoring ethnic group in their country, and then adjust the country level scores based on the fact that country populations are different in 2002 from in 1500…. They regress 2002 current income per capita on technology levels in past years, they find significant effects. This holds for technology in 1000BC, 0AD, and 1500AD…. Take the 1500AD result in column (3)…. If the technology index goes from 0[.5] ([half] technologies) to 1 ([three fifths of] the original technologies), log income per capita goes up by 2 log points, or by 3…. Also notice the R-squared, which is 50% for the 1500AD results….

CEG… establish that old technology levels have predictive power for current income per capita. They are not looking for explanatory power…. Ultimately, you might want to argue that we want strict causal explanations for why some countries are rich in 2002. But an explanatory paper like this is valuable…. Knowing that anything in 1500AD has strong predictive power for incomes today is informative. It tells us that we have to look back to 1500AD for at least some of those causal forces…. It isn’t the technology in 1500AD per se that matters…. This is an indicator of some kind of variation in culture or institutions (or something else?) that matters… [and] is telling us to something about how powerful those cultural/institutional factors are.

Must-Read: Dietrich Vollrath: Can We Get Rich by “Doing Business” Better?

Must-Read: Dietrich Vollrath: Can We Get Rich by “Doing Business” Better?: “Below I’m going to get to the gory details of why the Doing Business (DB) indicators generally suck…

…But let me start with this note. The DB index [John] Cochrane uses is a ‘distance to the frontier’ index. Meaning you get a number that tells you how close to best practices in business conditions a country gets. If you are at the best practices in all categories, you’d get a 100. Cochrane says, and I quote, ‘If America could improve on the best seen in other countries by 10%, a 110 score would generate $400,000 income per capita…’. Stew on that for a moment. Think about how that DB frontier index is constructed.

Cochrane went there. He said it could go to 11….

[…]

It Gets Worse: In the follow up post, Cochrane appeals to a graph from my textbook with Chad Jones… the relationship of an index of ‘social infrastructure’ and TFP…. I calculated it, graphed it, and stuck it on the slide that Cochrane linked to. I simply scaled and averaged the 6 different components of the World Bank’s governance indicators, much like the DB index. It has all the issues I described above, except worse. This figure tells us very little. Which is why in the book we immediately say that you cannot infer anything causal from it, and then go on to talk about some of the better studies done looking at specific institutions and their effects on economic outcomes…

Can we get rich by Doing Business better Dietrich Vollrath

Must-read: Evan Soltas: “Is Pro-Business Reform Pro-Growth?”

Must-Read: A very interesting and smart early cut at the data on “business climate” reforms and growth from the extremely-sharp young whippersnapper Evan Soltas:

Evan Soltas: Is Pro-Business Reform Pro-Growth?: “I came up a simple trick to identify reforms directly from the World Bank data…

…I calculate the rolling sum of all the increases in the [Business Climate] index that are larger than some threshold kappa, which I define as the 95th percentile of all annual changes…. This is, I think, a reasonable way of doing things: Even if you are distrustful of the index, as am I, if the World Bank says that your country is in the top 5 percent of reformers in some year, there’s probably something to that…. According to this measure, 36 percent of countries had at least one reform in my sample, the average reform entailed an 18-point increase in the index, and there are 109 reform years across countries….

Using data from the Penn World Table on output-side real GDP per capita at chained PPPs…. The latest batch of data covers up to 2012, and the World Bank data begins in 2004, so my analysis is limited to that time period…. Burkina Faso is not many times poorer than the U.S. only because it is many times easier to do business in the U.S. than in Burkina Faso…. To identify the causal effect, we should consider whether a country’s per-capita output rises after a reform relative to similar countries in the same year…. I [find I] can bound the effect of pro-business reforms quite precisely around zero, with a 95-percent confidence interval for the effect of a 10-point reform on the level of per-capita output of -1.4 percent to 3.5 percent. That is far away from the claim that such a reform could double per-capita output….

It’s reasonable to think that the effect of reforms on output lags the reform, but these results are unchanged when I include lags of the reform and drift components…. Another claim one might make is that the effects of reform are being obscured by differences in the path of output across regions or income groups (e.g., low-income countries)…. I can control for it by interacting region or income-group dummy variables with the year dummies…. What these tests show is that my results aren’t unduly sensitive to the manner in which I measure the effect of pro-business reforms….

The issue of pro-business reform is most relevant to developing countries with weak institutions, who face meaningful pressure from international organizations like the World Bank to implement them. Yet the data give little reason to believe that countries can raise per-capita output in any notable way by making it easier for entrepreneurs to do business. Pro-business reforms probably have very limited effects, if any, on economic growth within a reasonable time horizon. There still may be compelling reasons to do them…. More prosaically, there are important limits to my empirical analysis. If the impacts of these reforms only begin to manifest themselves more than five years later, for instance, I won’t be able to see that. Measuring longer-run effects will have to wait for more data….

The lesson here is to beware the TED-talk version of development economics. Shortening the time it takes to incorporate a small business is not a substitute for deeper institutional reforms, such as those that support investment in human and physical capital, remove economic barriers that hold back women and ethnic or religious minorities, or improve transportation, power, and sanitation infrastructure. Easy pro-business reforms should not distract countries from pursuing changes that, while harder to make, we know to be richly rewarding in the long run.

I confess to being surprised: I am enough of a card-carrying neoliberal still to have expected modest positive effects…