Romer and Parallel Developments

by

A couple of months ago, Paul Krugman wrote a review of David Warsh’s new book, Knowledge and the Wealth of Nations. Without having read the Warsh book, I blogged that I thought Krugman was a little harsh in dismissing the impact of Paul Romer’s work. In particular, he wrote that:

Now “Romer 1990″ is a terrific paper — I wish I had written it, which is the highest praise one economist can give to another. Yet I don’t think it can bear the weight Warsh places on it. Nor is it clear that increasing returns really did transform our understanding of economic growth. In fact, Warsh seems to concede as much. “So there is a new economics of knowledge. What has changed as a result? The answer, it seems to me, is not much.”

Now that I have read the book, I think that Krugman actually accurately represents the mainstream view of the Romer Model and that his notion is reinforced by the book. However, I think that there is a sense in which the steady-state view still somewhat limits that importance of the Romer Model and perhaps what it should represent. I will try to explain what I mean, in what I expect to be a rather long post.

David Warsh is a former journalist for the Boston Globe. For reasons that are beyond me, he has taken an extraordinary interest in the workings of academic economics. Presently, he publishes an excellent on-line newsletter, Economic Principals which is a must-read for economics academics. Knowledge and the Wealth of Nations takes all this with a broader brush. In effect, it is about the development of modern academic economics, its politics, how they come to research what they research and what drives them. To anchor the story, he takes the development of a particular article, Romer (1990), and centres much of the book around the career of a particular man, Paul Romer. Suffice it to say, this makes it a gripping read. Enough to put up with a poor overall construction, a lack of a sense of time and too much repetition (so much so that it seems the chapters are a bunch of independent articles strung together). However, I cannot for the life of me imagining it appealing to anyone but academic economists and maybe only a small subset of them too boot.

Nonetheless, I am definitely in the set of people that this book appealed to. Of course, I have broad interests in the economics of technological change and how academic economics operates. But there is a deeper connection here. There is a sense in which I lived through all of these developments but not as an inside observer but as an outsider, in many ways, blissfully unaware of the developments taking place. I say this with a sense of irony as the developments of linking knowledge to growth were all premised on the idea that knowledge would spillover. In my case, they did not; at least not quickly.

The book begins by describing a watershed session on new developments in economic growth presented in the 1996 meetings of the American Economic Association (in San Francisco). As I read about how the world supposedly changed during that session I remembered that I was in fact there. Somehow I missed that the world had changed. Romer presented, Nicholas Crafts (an economic historian) was troubled, Marty Weitzman went off on an obscure angle and Bob Solow commentated. It was a good session — a memorable one, in fact — but hardly a watershed from my perspective. It covered stuff we had known for some time.

Anyhow, why did I attend that session in the first place? Most academics know me as an industrial organisation economist. However, I didn’t start out that way. When I was studying undergraduate economics and thinking about an honors thesis topic I was drawn to the idea of economic growth. This was back in 1988 and I had just had a class at the University of Queensland that had described the basic growth models — Harrod-Domar, Solow-Swan — and the extraordinary incompleteness of these in explaining economic growth. The conclusion was that long-run growth was driven by the exogenous progress of technological change.

At the same time, I had become aware — through a class of Don Lamberton’s — of the economics of information. There were lots of parts to this but the part that captured my attention was the notion that knowledge was costly to produce; something stressed most clearly by Ken Arrow in his informal writings. Now to most people, this seems rather obvious, but it was rare at that time for economists to take that notion seriously.

For that reason I figured that if you were modeling economic growth, you would have to consider the resources devoted to knowledge production. Moreover, I figured that knowledge was different in that the resources required for its production were “independent of its scale of use.” Today, we call this non-rivalry, but the concept is what was important. In particular, not only would knowledge increase productivity in a durable way but also it would be an input into the further production of knowledge and increase knowledge productivity too. Thus, there would be a positive feedback that would mean that growth rates would continue to grow as the resource base of the economy grew; most critically population. That surely would provide a more complete picture of growth than standard growth models.

This is all in my UQ honors thesis which I took it upon myself to re-read this week; a big nostalga trip. What is interesting about it, and a source of endless amusement to Scott Stern whenever he visits my office, is that I wrote it with no particular formal economic skill. That wasn’t the Queensland way. There was a formal model but the language clearly shows the fact that I wasn’t plugged in. A far cry from the skill transformation Stanford would impose on me.

Which brings me back to Warsh’s book. At the same time I was working on my honors thesis, Paul Romer was coming to a similar conclusion about what was needed in an economic growth model. In his PhD thesis, Romer had put increasing returns back into economic growth and solved an important technical problem in the process (Romer, Journal of Political Economy, 1986). However, in that work, knowledge was a pure spillover and did not require resources explicitly devoted to its production. When someone pointed Romer out to me in 1989, it was this article that I read and I dismissed it as not getting the point. According to Warsh, Romer in 1988 had come to the same conclusion and had left this work behind.

In 1988, Romer decided to take a different tack. He modelled the growth of knowledge as a production function with human capital and the stock of knowledge as inputs. The human capital part came from a 1985 series of lectures by Robert Lucas (Romer’s PhD supervisor) who saw human capital exclusively as creating knowledge. Now this was a paper I too was aware of but I, like Romer, was dissatisfied with it just being that. The stock of knowledge surely had a role too.

Here again is where I can return to my honors thesis. Romer wrote the increment to knowledge as a simple function of d.HA.A where d is a productivity parameter, HA is the human capital devoted to knowledge production and A is the current stock of knowledge. Looking at my thesis, I had engaged in more formalism than I remembered as I wrote the increment to knowledge as a function d.R where d was the same as in Romer and R was the resources devoted to research. However, R was itself a function of the total stock of knowledge. Ackward, to be sure, but exactly the same function.

However, from that point on my thesis and the Romer Model depart fundamentally. It is for this reason that I can so readily appreciate precisely what Romer did. For most economists, it is this first insight on the production function for knowledge that they remember when they learn about the Romer Model. For me, that wasn’t the really hard thing to do.

To understand Romer’s contribution, it is perhaps easiest for me to start by recounting where I, with no understanding of how to do economics, took my model. I appear (it is harder to tell from this point) to have assumed that knowledge takes more resources in order to be effectively distributed and become useful in production. Those resources do not have any magic increasing returns effect and so are not scalable. Indeed, some of the resources are wasted productively. In the end, this lead to a balanced growth rate that did not depend on scale; growth in output would depend on the growth of resources (that is, the population). This is something, incidentially that turns out to be borne out by the data, but it also was an implication of the Romer Model as Chad Jones later showed. My thoughts on this were my first publication in 1989 in Information Economics and Policy grandiosly titled “Knowledge of Growth and the Growth of Knowledge.” The article does not cite Romer (1990); I was not aware of it even then.

The other thing I was lacking, which is a problem for an economic model, was a market. It is this that Romer had in spades. He had two in fact. He had a market for intellectual property (which required bringing monopolistic model for the first time into macroeconomics); that itself was derived through demand for new goods. And he had a market for labour — both skilled and unskilled. Using these, he closed the model, actually had consumers who worried about how much to save and how much of the new whiz bang goods to consume, and he ended up with a growth rate related to the level of human capital and not just its growth. All this was beautifully packaged in one of the finest papers ever written in modern economics; Romer (Journal of Political Economy, 1990).

Now the Warsh book, like the steady state view from the profession, downplays precisely this contribution. But it is that that made the Romer Model appealing and something to work with. More critically, however, it is that feature that gives the Romer Model policy teeth. Because it is not about simply pouring more money into science funding that gets you growth. Instead, it is about encouraging people to become scientists and engineers that really matters. Make that labor market work well and be free of market failures and that is how you increase long-run growth rates. It is only in the very last short chapter of the book that Warsh finally gets to this point. From that perspective, it is hardly surprising that the steady state view from the profession misses it too.

For my part, I went to Stanford (mainly because of Ken Arrow) and first was in a room with Paul Romer when Robert Barro presented a seminar on economic growth across countries; another watershed moment according to Warsh that I was blissfully unaware of at the time. Not surprisingly, I aced my macroeconomics comprehensives (a third of which was devoted to the Romer 1990 model that I was able to appreciate in its full glory) and then left growth theory for a while to explore microeconomics and economic history.

But somehow it quickly drew me back when I encountered an article by Murphy, Shleifer and Vishny (1989) on the big push in industrialisation. The development of that article is also in the Warsh book (another touching point for me). It gripped myself and other Stanford students including Antonio Ciccone, Andres Rodriguez and my colleague Catherine de Fontenay. I wrote my PhD thesis on the topic marrying the apparatus of the Romer Model to that of Murphy, Shleifer and Vishny. By that time, there was a flood of growth papers (a fad in fact) and that work never published in good journals.

Returning to Australia, I discovered an Australian program of research into growth led by Xiaokai Yang (at Monash) who came up with a way of modeling increasing returns that was non-standard but with similar conclusions to the Romer model. By that time, however, I was done with growth and so left it behind.

The Warsh book is a reminder of those times but also a history of the subject. Research on knowledge, specialisation and economic growth was described by Arrow as running underground and bubbling to the surface only every 20 years or so. There was a bubble with Alfred Marshall at the turn of the century, another with John von Neumann prior to World War II, a bubble in the 1960s (which is something no one likes to talk about so I have no idea what that was) and then a bubble with Romer in the 1990s. So it is in that sense that Krugman and Warsh are right, and it was a temporary phase. But when the next bubble comes along, I am certain that whomever drives it will begin with Romer (1990) just as that model predicts.

[Postscript: I know Paul Romer fairly well now. I worked on some of the earliest modules for his start-up Aplia (including Supply and Demand if you look at that now), they host my adapted Mankiw text, I write for their blog, and I am on their advisory board. We have talked relatively little about economic growth, however.]
4 Responses to "Romer and Parallel Developments"
%d bloggers like this:
PageLines