Experiments in industrial policy

Readers of this blog know that I think firms are understudied in development. I’ve said before: the path from $2,000 a head to $10,000 a head in GDP is through industrialization.

But how to spur firm growth? Too many economists are too quick to throw up their hands and say we know nothing or can do nothing. And most of the formal industrial organization research feels completely alien to the issues that concern firms in poor countries.

One of the best sessions I saw at AEA is breaking the trend.

David McKenzie discussed a new experiment that provides temporary wage subsidies to owner-operated microenterprises in Sri Lanka, to hire their first employees and expand. The question: will these temporary subsidies lead to permanent growth and jobs?

Businesses might face fixed costs to hiring their first employee, either because they need time to generate the returns that will make the investment pay off, or because they need to learn how to be a manager. Subsidies might push firms across the cost hump (or reduce the risk of trying), leading to a rise in the level (or even the growth rate) of firm revenues. The results of this experiment are yet to come, but his series of firm experiments with Chris Woodruff and Suresh de Mel is one of my favorite suites of experiments around.

As much as fixed costs matter, the real barrier to firm growth in poor countries could be management. A good manager is partly one with training, experience, and a sense of best practice. A lot of a factory’s success comes down to management as a technology, though. Organizing a production floor, ensuring quality and efficient production: these are less a product of a single manager’s talent and more the product of decades of trial and error in manufacturing plants around the world. The problem: if you are running a firm in India or Mexico or elsewhere, the best practices probably haven’t filtered through.

Two papers looked at bridging the gap. Nick Bloom talked about an experiment (also with coauthors) with large textile firms in India. Some received expert operational consulting for a month. Others just a consulting visit. They found horrendous practices–disorganized factory floors, open garbage, little cleaning, no spare parts, lost tools, and little preventative maintenance—and pushed 38 best practices on their treatment firms. Two thirds of the practices were adopted and the majority seem to have lasted more than the first few months. The firms saw huge improvements in quality and profits.

Antoinette Schoar and coauthors did the same with smaller firms in Mexico (one with 1 to 250 employees). They too received subsidized consulting services, mainly sales, marketing and HR advising for owner managers. Again, sales and profits rose.*

What’s happening here? Why aren’t firms investing in knowledge that provides enormous payback? Several market failures are possible: credit constraints, absence of incentives, behavioral defects (like procrastination), to information that is too messy and inaccessible for non-experts to absorb, even if it’s available.

Even though the mechanism isn’t yet clear, the results suggest a lot of potential for industrial policies that provide short-term subsidies to address start-up costs, technology transfer and human capital. For a serious industrial policy, though, figuring out why the experiments work is probably more important than figuring out how well they work.

All in all, in light of yesterday’s comment on the meaningfulness of development research, all are great examples of the kind of research that matters.

*Sorry–public papers aren’t yet available for most of these presentations.

7 Responses

  1. Oh, and Michael :: completely agreed. You’d be surprised though at some of the very simple things that Indian manufacturing firms do or don’t do :: e.g. piles of trash and old broken tools obstructing the factory floor and creating fire/safety hazards (it’s hard to argue the managers don’t know they should be organized and clean up their factories)

  2. Joe, there will be a public working paper available soon, and if you email me (benn@econ.berkeley.edu) I can send you some slides.

    Michael, exactly the opposite :: frankly almost no economists I’ve ever met would have a clue how to help a large Indian manufacturing firm improve its operations. We had an outstanding team of management consultants from Accenture, with a manager who’d led manufacturing operations in one of the biggest companies in India, lead everything; we just worked with them to design priorities, performance metrics and data collection protocols.

  3. Any other informative, development-related presentations (for students who couldn’t make it out to the AEA conference)?

  4. Why aren’t firms investing in knowledge that provides enormous payback?

    I think this is an example of something much more general, which I’ve noticed particularly in studying the history of science, though also elsewhere. This is that, for any kind of knowledge involving technical expertise,

    a) However obvious something is in hindsight, it’s often extremely difficult to discover in the first place, and

    b) It’s almost impossible to acquire technical knowledge just from textbooks, verbal descriptions, or looking at the final result, even if you know where to look (which can be hard to discover in itself). Almost always, it takes intensive face-to-face interaction, getting explanations, worked examples, exercises, and opportunities to ask questions and get feedback, from someone who has themself worked in an environment where the technical knowledge is put to daily use. This seems to apply to every kind of technical knowledge from advanced mathematics to organising a shop floor, from determine a protein’s chemical structure to running a software project, from painting a crowd scene in oils to designing an advertising campaign. That is why, when you see a new effective technology or technique developing, you can often trace the actual lineages of the workshops by looking at who was trained by whom, going back to a single initial workshop (and when you can’t trace the lineage, you can often see that all the developments occurred in a small geographical region, with lots of opportunities for personal contact). So, for instance, in early industrial Britain, a high proportion of the engineers had actually personally spent years working for Boulton and Watt; and British nineteenth-century expertise in mathematical physics spread from a small number of coaches at Cambridge University.

    Of course, it’s possible to discover things for yourself by setting up a rigorous research program—but running such a research program is itself a technology that needs to be passed on by personal contact! (Also, the techniques for researching one kind of knowledge only partially carry over to other kinds of knowledge.)

  5. Very interesting post! I believe that credit constraints, uncertainty and learning play a decisive role in industrial development. Lack of industrial development may be a result of firms inability to pay for the fixed costs of the best technologies. It can also be a result of entrepeneurs uncertainty about these technologies’ effectiveness in their specific environment (a ponti made by R. Hausmann and D. Rodrik in a paper called “Economic Development as Self-Discovery”). Finally it can be a result of lack of learning on how to best operate the existing technologies (a point made by R. Ossa in a very interesting paper called “A Gold Rush Theory of Economic Development”).

  6. Maybe the experiments are successful because economists are designing consulting strategies instead of consultants? I bet one of these days we’ll get former consultants who became Ph.D. economists and then switched to political science running experiments in electoral campaigns… oh wait….

  7. Chris,
    thanks for this post, very interesting. I am currently working on industrial policy in Rwanda. Similar targeted consultancy has been carried out for major firms here, but I imagine the extent in terms of time, services offered and quality, is doubtfully as high as in the experiments you mention. It would be useful if you could post more detailed descriptions of the consultancy provided when this becomes available if possible. Many thanks.