Developing Jen

 

I’m taking a class on bridging research and policy with Rohini Pande and Sendhil Mullainathan this semester.  Each week we have to answer discussion questions related to our readings, and last week mine was chosen to share with the class.  I figure that’s a pretty good sign it’s worth sharing with the world at large as well. 

In addition to the technical challenges with creating a global poverty count as outlined in my post below, there are also conceptual issues with poverty counts that I would be remiss to not include.  

First is the simple idea of a head count, as defined by those above or below a defined consumption level.  People's lives are not significantly different just below or just above that line, and yet those above are excluded from our numbers.  Small changes to the line can lead to big changes in the count.  Focusing on poverty counts could create an incentive to move people just above the line, so the numbers imply big improvements when in reality people's lives haven't changed that much.  

Measuring global poverty levels is hard, so hard the actual numbers might be of limited use. According to Angus Deaton, the godfather of poverty measurement, "it seems impossible to make statements about changes in world poverty when the ground underneath one's feet is changing in this way."  I thought it would be helpful to illuminate why.  

Dani Rodrik's teaching forms the basis for how I approach industrial policy - namely that governments should intervene where markets fail, and implement policies that target those failures.  This principle also forms the basis for an issue I have with how he approaches policies to promote economic growth through innovation.

This post is  a follow up to yesterday's about an article in this week's Economist that discusses the work of one of my professors, Ricardo Hausmann.  Though there are important caveats I think people need to understand (see my post), the work has tremendous value in helping us understand how the structure of economies evolve.  The real problem comes when Hausmann partners with Dani Rodrik and Charles Sable to translate his findings into policy prescriptions.  Somehow, something has gone terribly wrong.

The Economist article concludes: "In work with Dani Rodrik of Harvard and Charles Sabel of Columbia University, Mr Hausmann argues that governments should emulate venture funds, backing new enterprises in the hope that one will make the leap into a more densely forested area. They should spread their bets widely, monitor progress closely, and cut losses promptly."

Well first, the Economist gets it wrong in that Rodrik and Hasumann don't advocate a one size fits all, start a public VC approach.  But they've advocated it before, and I think this suggestion is a horrible idea, for two reasons.  One, because  it's probably impossible to operationalize and two, because it violates Rodrik's own first principles for industrial policy, which guide how I think about these types of government-market interactions.

Operationalizing a public venture capital fund is likely to be infeasible for a number of reasons, here I will only note the most obvious.  It would be very difficult to design and enforce an structure that protects against the incentive issues outlined above. Equally critical are the political challenges that arise from failed investments (in their papers Rodrik and Hasumann argue that if you're not failing, you're not doing it right).  Low tolerance for failure would be even greater in developing countries with limited fiscal resources.  Pressure against failure would both exacerbate the incentive issues as well as skew the fund towards less risky investments.   The government is also bound to be much worse at allocating capital than the market due to capability deficiencies.  Staffing a public fund would be extremely challenging, as the skills necessary for venture capital operation is scarce in developing countries.  In the rare instances where it does exist, it will be difficult for the public sector to provide attractive compensation.

More problematic, a public venture fund violates Rodrik's own first principles for industrial policy.  The crux of Rodrik’s approach is that 1) markets fail, 2) governments should intervene where markets fail, and 3) policies should target the failures.  Rodrik and Hausmann never address what the failure is that justifies government intervention in the financial market.  Surely failures do exist in the venture capital market, but it's unclear why the government  should go so far as provide venture capital directly, versus providing incentives to venture capital firms.  Equally central to Rodrik’s approach to industrial policy is the need to define institutional processes that balance the need to interact with the private sector for information to formulate policy with the risk of corruption and rent-seeking.  The partial ownership of firms that results from the provision of venture capital fails in this.  Huge incentive problems are created when the government becomes investor, policy maker, and regulator.

For these reasons I believe public venture funds are infeasible at best and harmful at worse.  A better understanding of failures in the venture capital market is necessary to formulate policy incorporating the insights from Hausmann and HIldago's work. 

There's an article in this week's economist about Ricardo Hausmann's product space.  I'm happy to see it getting mainstream attention.  It's incredibly important work, and it just makes intuitive sense.  But before we start using it for policy decisions, we should be aware of some of the major caveats.

To summarize, what Hausmann and Hildago did was create a mapping of products across the global economy, defining proximity by looking at the probability of a country having a comparative advantage in one product given it has a comparative advantage in the other.  The analysis shows that the product space (aka the forest) is highly heterogeneous, with a dense core and clusters of products  as you move towards the periphery.  When they mapped specific countries and regions in the space, they observe regional patterns of specialization and that developing countries are typically in the sparser areas of the forest. 

The really interesting insights come when they looked at how economies evolve over time across the product space.  The analysis revealed that economies (composed of firms aka monkeys) evolve to produce products that are close to those they already produce (monkeys tend to jump to nearby trees).  This suggests there may be limitations in how far across the forest a monkey can jump. 

We can think about this product mapping as a proxy for capability mapping.  The implication is that economies evolve according to their existing capabilities and can't make jumps across the forest.  Industrial policy then can help facilitate larger jumps than monkeys do on their own, but is fundamentally constrained by the capabilities that exist in the economy.

There are two significant holes that need to be understood.

First, the product space is created using international trade data, which excludes both services and non-tradables.  If the takeaway for all of this is that capabilities matter, then it's important to understand proximity relationships across all economic activity, not just exported goods.  It might be the case that services or non-tradables can help bridge gaps between the exportable goods included in the current mapping.  Based on the current analysis, these are open questions and I think crucial to understand before we base policy decisions on this work. 

Second, Hausmann and Hildago haven't yet explored how the product space evolves over time.  Their analysis of how economies progress through the product space takes it as static based on the data used to create the mapping.  But there's no reason to think that should be the case.  As new technologies emerge the capabilities required to produce products should evolve as well.  They recognize this is a question for further research and that this analysis is only a first approximation, but note that economies evolve across the space at a faster pace than the space evolves itself.

Despite these issues I find this work incredibly intriguing.  The place I really struggle, however, is turning these insights into policy implications.  How to help monkeys jump farther from trees?  Hausmann pulls in Rodrik's work to answer this question.  I will save that for another post. 

Generally, I argue that access to more information via technology is good for development. Access to information (e.g. prices) improves market efficiency.  Communication flows fostered by the Internet strengthen democracy.  With this information, people are empowered to debate the direction they want their societies to move in.  I've read my Amartya Sen, after all.

And while I'm at it, here's a link to Sendhil Mullainathan's TEDIndia talk last November.  In case you don't know who he is, he's a leading academic in the intersection of development and behavioral economics.  I'm lucky enough to be taking a class from him this semester so hopefully I'll have more to share, and often.  

Damon Horowitz, the CTO of Aardvark, is one of the smartest people you'll ever meet.  And one of the most interesting too.  After getting is master's in computer science at the MIT Media Lab and diving deep into artificial intelligence, he took his career for a left turn and got a PhD in philosophy from Stanford.  

A couple weeks ago he gave fascinating and thought provoking talk at the TEDx event in San Francisco about the boundary between machines and people.  He has a lot of interesting thoughts about the negative social implications of technology that hopefully he'll articulate on the TED main stage one day.