PAUL M. ROMER

PAUL M. ROMER

Paul Romer was born in 1955 in Denver, Colorado and obtained his BS

(Maths, 1977) and his PhD (1983) from the University of Chicago. His

main past posts have included: Assistant Professor at the University of

Rochester, 1982–8; Professor at the University of Chicago, 1988–90; and

Professor at the University of California, Berkeley, 1990–96. Since 1996 he

has been Professor in the Graduate School of Business at Stanford University.

Professor Romer is best known for his influential contributions to the

field of economic growth, which have led to the renaissance of economic

growth analysis and, in particular, the development of endogenous growth

models which highlight the importance of ideas in driving economic growth.

His most widely read articles include: ‘Increasing Returns and Long-Run

Growth’, Journal of Political Economy [1986]; ‘Growth Based on Increasing

Returns Due to Specialization’, American Economic Review [1987];

‘Endogenous Technological Change’, Journal of Political Economy [1990];

‘Idea Gaps and Object Gaps in Economic Development’, Journal of Monetary

Economics [1993]; ‘The Origins of Endogenous Growth’, Journal of

Economic Perspectives [1994] and ‘Why, Indeed, in America? Theory, History,

and the Origins of Modern Economic Growth’, American Economic

Review [1996].We interviewed Professor Romer in Chicago, in his hotel room, on 5 January

1998 while attending the annual conference of the American Economic Association.

Background Information

Where and when did you first begin to study economics?

I was a math and physics major at the University of Chicago. I took my first

economics course in my senior year because I was planning to go to law

school. I did well in the class and the professor encouraged me to go on to

graduate school to study economics. Economics offered some of the same

intellectual appeal as physics – it uses simple mathematical models to understand

how the world works – and in contrast to physics, it was an area of

academic study where I could actually get a job.

In some ways staying at the University of Chicago was attractive because it

had a very exciting economics department but I had already been there for

four years. Even though I had had very little Chicago economics training I

did not think it was a good idea to spend my whole career as a student in one

place, so I started in the PhD programme at MIT. There I met my wife, who

was just visiting for a year from Queens University in Canada. After two

years at MIT we went back to Queens to finish her final year of training in

medicine. That was when I started working on growth. At the end of that year

I transferred to Chicago, where my wife had a fellowship position, and

completed my PhD. I finished my PhD and entered the job market in 1982 –

my thesis is actually dated 1983 because it took me a year to polish it up.

As a student did you find any of your teachers to be particularly influential or

inspirational?

Well, Sam Peltzman was the professor who encouraged me to switch my

career path from law to economics. I shudder to think what my life would

have been like if he hadn’t asked to talk with me after the mid-term and I had

gone on to law school. It is an episode that I try to keep in mind – that

professors can be very influential, and a little bit of attention to your students

as people can make a big difference in their lives.

Besides having saved me from a life in the law, Sam was also an excellent

teacher. He was the first person to show me that you could take very simple

tools – demand curves or indifference curves – and derive surprising insights

about how the world works. Having mentioned Sam, I should also mention

some other very good teachers that I had. Donald McCloskey, now Deirdre

McCloskey, was the second person I had for economics. Donald, like Sam,

took economics very seriously. Together, they gave me an excellent introduction

to the subject. I should also mention that at Chicago, they did not offer

Paul M. Romer 675

what is known as a ‘principles’ course, the watered-down, mind-numbing

survey course that most universities offer as a first course in economics. At

Chicago, they started right off at the intermediate microeconomics level. So I

had the enormous advantage of starting off with challenging, intellectually

coherent material and first-rate teachers. I was very fortunate.

Later in graduate school, when I was back at Chicago, Bob Lucas and Jose

Scheinkman had a big influence on my style and the way I look at the world.

They set a standard for rigour and discipline – zero tolerance for intellectual

sloppiness – that I have aspired to ever since. But probably the best year of

graduate school was the year I spent at Queens University because I had a lot

of interaction with the faculty there. Normally as a graduate student you do

not really get that much time to sit and talk with members of the faculty as

colleagues. At Queens I had more of that kind of experience. Some of the

people I talked with intensively during that year – Russell Davidson and

James McKinnon – are terrific economists and had a big effect on my career.

Development of Macroeconomics

Are there any particular papers or books that you would identify as having a

major influence on the development of macroeconomics?

For me that’s too broad a question. I could list all the usual suspects, people

like Keynes, and so on. I’d be more comfortable describing the contributions

that have influenced my own work.

Tell us about the influences on your own research interests.

Bob Lucas brought a style to macroeconomics that had a big impact on a

whole generation of people, including me. There are several papers that

exemplify this style. One is his 1972 Journal of Economic Theory paper on

‘Expectations and the Neutrality of Money’. Another would be his 1978

Econometrica paper ‘Asset Prices in an Exchange Economy’. But his 1971

Econometrica paper, ‘Investment under Uncertainty,’ written with Edward

Prescott, is probably the best example because it really brought to the forefront

and crystallized for macroeconomists the connection between what we

did in macroeconomics and what the rest of the profession had been doing in

general equilibrium theory. In that paper Lucas and Prescott used the connection

between solving optimization problems and equilibria that has become

such a powerful tool in modern macroeconomics. That 1971 paper builds on

the work of people like Cass [1965] and Koopmans [1965], who had been

working in growth theory, and this basic approach for characterizing dynamic

equilibria can be traced all the way back to Frank Ramsey’s [1928] paper.

Still, Lucas and Prescott took this approach much further into the core of

macroeconomics. If all you have seen is the theory of investment as devel676

Modern macroeconomics

oped by the macro modellers and presented by the macro textbook writers,

this paper is like a flash of lighting in the night that suddenly shows you

where you are in a much bigger landscape.

You mentioned the influence of Bob Lucas’s work. What do you think has

been the lasting impact of his work, particularly the work he carried out in

the 1970s for which he was awarded the Nobel Prize?

I think the deeper impact of Lucas’s contributions has been on the methodology

of the profession. He took general equilibrium theory and operationalized

it so that macroeconomists could calculate and characterize the behaviour of

the whole economy. Just as Peltzman and McCloskey took intermediate

microeconomics seriously, Lucas took general equilibrium theory seriously.

Many of the people doing general equilibrium theory for a living did not

really seem to believe in what they were doing. They gave the impression that

it was a kind of mathematical game. Economists working in trade and growth

had shown us how we could use general equilibrium models, but they were

not ready to bring dynamics and uncertainty into the analysis. It was economists

working first in finance, then in macroeconomics, who took the theory

seriously and showed economists that fully specified dynamic models with

uncertainty had real implications about the world. A very important result of

that methodological shift was a much greater focus on, and a much deeper

understanding of, the role of expectations. But this is only part of the deeper

methodological innovation. You still wouldn’t know it from reading textbooks,

but to research professionals, it finally is clear that you can’t think

about the aggregate economy using a big supply curve and a big demand

curve.

One of the ironies in this revolution in thinking is that the two people who

did the most to bring it about, Lucas and Robert Solow, ended up at swords’

points about the substantive conclusions that this methodology had for macroeconomic

policy. Solow’s work has also had a huge impact on the profession,

pushing us in the same direction. His work on growth also persuaded economists

to take simple general equilibrium models seriously. Many people

recognize the differences between Lucas and Solow over macro policy questions,

but fail to appreciate the strong complementarity between their work at

the methodological level. If Joan Robinson had won the day and banished the

concept of a production function from professional discourse, Lucas and

Prescott could never have written ‘Investment under Uncertainty’.

During the 1980s, the real business cycle approach to aggregate fluctuations

developed in parallel with new growth theory. How do you view that work, in

particular the way it has sought to integrate the analysis of fluctuations and

growth?

Paul M. Romer 677

A lot of the progress in economics still comes from building new tools that

help us understand very complicated systems. As a formal or mathematical

science, economics is still very young. You might say it is still in early

adolescence. Remember, at the same time that Einstein was working out the

theory of general relativity in physics, economists were still talking to each

other using ambiguous words and crude diagrams.

To see where real business cycle theory fits in, you have to look not just at

its substance and conclusions but also at how it affected the methodological

trajectory I was talking about before. You can think of a hierarchy of general

equilibrium models – that is, models of the whole economy. At the top you

have models of perfect competition, which are Pareto-optimal so that you can

solve a maximization problem and immediately calculate the behaviour of

the economy. Then, at the next level down, you have a variety of models with

some kind of imperfection – external effects, taxes, nominal money, or some

kind of non-convexity. In many cases you can find a way to use some of the

same maximization tools to study those dynamic models even though their

equilibria are not Pareto-optimal. This is what Lucas did in his 1972 paper

‘Expectations and the Neutrality of Money’. Formally it is like an external

effect in that model. It is also what I did in my first paper on growth.

The real business cycle guys went one step further than Lucas or I did in

trying to simplify the analysis of aggregate economies. They said, ‘We can go

all the way with pure perfect competition and pure Pareto optimality. We can

even model business cycles this way. Doing so simplifies the analysis tremendously

and we can learn a lot when we do it.’ My personal view – and

increasingly the view of many of the people like Bob King, who worked in

this area – is that at a substantive level real business cycle theory simplifies

too much. It excludes too many elements that you need to understand business

cycles. This doesn’t mean that the initial work was bad. It just means

that we are now ready to go on to the next stage and bring back in things like

predetermined nominal prices. Methodologically this work helped us refine

our tools so we’ll do a better job of understanding predetermined prices when

we bring them back into the model.

We frequently make progress in economics by seeming to take a step backwards.

We assume away real problems that people have been working on in

vague and confused ways, strip things down to their bare essentials, and get a

better handle on the essentials using some new tools. Then we bring the

complications back in. This is what Solow was doing, and what drove Robinson

to distraction, when he modelled the production structure of an economy using

an aggregate production function. Later we brought back many of those complications

– irreversible investment, limited ex post substitution possibilities,

and so on – back into the model. The real business cycle theorists did the same

kind of thing, and during the simplification phase, they also made people mad.

678 Modern macroeconomics

Economic Growth

In Lucas’s [1988] paper ‘On the Mechanics of Economic Development’ he

comments that once you start to think about growth it is hard to think about

anything else. In the introduction of their textbook Economic Growth Robert

Barro and Xavier Sala-i-Martin [1995] argue that economic growth is the

part of macroeconomics that really matters. In the light of these comments

by very influential macroeconomists, do you think that, on reflection, economists

have in the past spent too much time trying to understand business

cycles?

That is almost right. Remember that we experienced major macroeconomic

calamities in the interwar period. These depressions were sufficient to wipe

out 30 to 40 years’ worth of growth. Economists who grew up during this era

certainly didn’t have any trouble thinking about something else besides

long-run growth. They naturally focused on avoiding those calamities.

So I don’t think that you can make the statement that focusing on growth is

more important in some absolute sense than focusing on stabilization. What I

think is correct is that we now know how to avoid the kind of catastrophic

events that we saw in the UK in the 1920s and in the USA in the 1930s.

Those were both major mistakes in monetary policy and we now know how

to avoid them. We also know how to avoid the disruptive hyperinflations of

the interwar era. Recently, we have even developed better monetary rules for

avoiding the less disruptive but still costly inflation of the 1970s. Once you

have learnt to avoid those kinds of problems, growth stands out as the most

important remaining issue on the agenda.

I do believe that there was a period in the 1960s and 1970s when

macroeconomists were spending too much time looking at business cycles –

the smaller cycles and fluctuations which characterized the post-war period –

and too little time on growth. We should have kept working on stabilization

policy, but we should also have worked on the determinants of long-run

growth. Adjusting the balance is what my career has been all about.

When I teach students I try very hard to get them to get this balance right. I

give them an analogy about a runner who is trying to train for a marathon.

Asking whether growth is more important than stabilization is like asking

whether conditioning is more important than putting on a tourniquet when

the runner starts bleeding. In a sense the training and the technique of running

really are what wins races. But if the runner is bleeding to death, it is

pretty silly to lecture her about getting in better shape.

But now, when we look at the allocation of the profession’s intellectual

resources today, we are in a situation where we can learn more about how to

make minor adjustments in the amplitude of cycles or in the trend rate of

growth. Faced with that trade-off, it is very clear that small improvements in

Paul M. Romer 679

the trend rate of growth can have far greater effects on the quality of life, and

this area has been understudied.

Looking back, one of the reasons why economists avoided questions about

growth was that our tools were not sufficiently well developed. Purely technical

or mathematical issues about the existence of a solution to an infinite

horizon maximization problem, transversality conditions, knife-edge behaviour

and explosive growth deterred economists from asking the right kind of

substantive questions about long-run growth. Now that our tools are better,

we have been able to set those issues aside and make progress on the substantive

questions.

The classical economists were very concerned with long-run issues such as

growth. Did you find any inspiration for your work by going back and looking

at the contributions of the classical economists and other early work on

growth?

I did spend some time thinking about that, reading Adam Smith and Alfred

Marshall. For example, I read the 1928 paper by Allyn Young, which builds

on Marshall’s work. I think it is in the same issue of the Economic Journal as

Ramsey’s paper. So there was a period where I spent a couple of years trying

to sort out the connections between what Young and Smith were saying and

what I was trying to say. I did that for a while and enjoyed it, then I stopped

doing it. I am not sure I would recommend it as a research strategy for a

young person, but it can be interesting and instructive.

When I started working on growth I had read almost none of the previous

literature. I started very much from a clean sheet of paper and only later went

back to try to figure out what other people had said. I think that in a lot of

cases that is the right way to do it. If you devote too much attention to

ancestor worship, you can get trapped and lose the chance to see things from

a new perspective. Of course, in economics, your ancestors are still around,

occupying positions of power in the profession, and they are not always

happy when someone comes along and tries to take a fresh look at things.

During the whole period from the marginalist revolution in the 1870s through

to the mid-1950s economists were mainly concerned with microeconomic

developments and managing the birth of macroeconomics during the Great

Depression. Then the issue of economic growth came back on to the scene

during the 1950s. One of the puzzles is that during the period when growth

theory made great advances, with the contributions of Solow in 1956 and

1957, the field of development economics seemed to evolve as an almost

separate area of interest. Why did that dichotomy happen?

I am probably going to sound like a broken record here, repeating my message

over and over, but the divide was methodological. The growth guys

680 Modern macroeconomics

talked math; the development guys still talked words. They diverged further

and further because they could not understand each other. It was less the

differences in the substantive questions they were asking than the tools they

were selecting to try to address them.

Wasn’t it more the case that development economists actually wanted and

needed to say something about policy issues?

There was an element of that. As I said about the real business cycle theorists,

sometimes you have to take a step back and simplify to make progress

developing new formal tools. This is hard to do when you are in the thick of

the process of trying to offer policy advice.

If you go back and read Smith, Marshall or Young, you have to be struck

by what an incredibly wrenching transformation the economics profession

has gone through, from operating as a purely verbal science to becoming a

purely mathematical one. Remember that Allyn Young’s paper came out at

the same time, even in the same issue, as Frank Ramsey’s. Ramsey was using

tools like the calculus of variations that physicists had been using for decades.

But economists were still having trouble with basic calculus. Jacob

Viner needed help from his draughtsman to get the connection right between

long-run and short-run average cost curves. Nowadays economists use math

that is as sophisticated and as formal as the math that physicists use. So we

went through a very sharp transition in a relatively short period of time. As

we learned how to use mathematics we made some trade-offs. You could

think of a kind of production possibility frontier, where one axis is tools and

the other axis is results. When you shift effort towards the direction of

building tools you are going to produce less in the way of results. So the

development guys would look at Solow and say, ‘What you are producing has

no useful content for policy makers in the development world; you guys are

just off in mathematical space wasting time while we are out here in the real

world making a difference.’ The tool builders should have responded by

explaining the intertemporal trade-off between results and tool building and

that as a result of this work we can give better policy advice in the future.

Overall the right stance for the profession as a whole is one where we

approve of the division of labour, where the people who specialize in those

different activities can each contribute and where we do not try to force the

whole profession into one branch or the other. Ideally we should keep the

lines of communication open between the two branches.

Let us turn to Robert Solow’s contributions. What do you see as being the

main strengths and weaknesses of the Solow growth model? Some economists

like Greg Mankiw [1995] would prefer to modify the Solow model rather than

follow the endogenous growth path.

Paul M. Romer 681

When it was introduced, the Solow model made several very important

contributions to economics and progress in this tool-building direction. It was

a very important demonstration of how you could take general equilibrium

theory and apply it and say things about the real world. As I suggested before,

Solow helped persuade us that there are ways to think about the equilibrium

for the whole economy, using simple functional forms and simplifying assumptions,

and get some important conclusions out of that. It is a very

different style of general equilibrium theory from that of Arrow and Debreu

and their more abstract work that was going on at the same time. Remember

that Solow and Samuelson had to engage in vicious trench warfare about this

time with Cambridge, England, to make the world safe for those of us who

wanted to use the concept of a production function.

At the substantive level – which I think is where your question was directed

– the strength of Solow’s model was that he brought technology

explicitly into the analysis in both his empirical paper and his theoretical

paper. He had an explicit representation for technology, capital and labour.

Those are the three elements that you have to think about if you want to think

about growth. That was the good part. The downside was that because of the

constraints imposed on him by the existing toolkit, the only way for him to

talk about technology was to make it a public good. That is the real weakness

of the Solow model. What endogenous growth theory is all about is that it

took technology and reclassified it, not as a public good, but as a good which

is subject to private control. It has at least some degree of appropriability or

excludability associated with it, so that incentives matter for its production

and use. But endogenous growth theory also retains the notion of non-rivalry

that Solow captured. As he suggested, technology is a very different kind of

good from capital and labour because it can be used over and over again, at

zero marginal cost. The Solow theory was a very important first step. The

natural next step beyond was to break down the public-good characterization

of technology into this richer characterization – a partially excludable nonrival

good. To do that you have to move away from perfect competition and

that is what the recent round of growth theory has done. We needed all of the

tools that were developed between the late 1950s and the 1980s to make that

step.

Let me place the other strand of growth in context, the so-called AK

versions of endogenous growth. In these models, technology is just like any

other good – we might put another label on it and call it human capital or we

can call it generalized capital – but technology is treated as being completely

analogous with physical capital. I think that approach represented a substantive

step backwards compared to the Solow model. The AK models are less

sophisticated than the Solow model because those models do not recognize

that technology is a very different kind of input. As I suggested earlier, I also

682 Modern macroeconomics

disagree with the real business cycle methodology that says ‘Let us do everything

with perfect competition’. Before, you could argue that there was no

alternative, but that’s no longer true. We have perfectly serviceable dynamic

general equilibrium models with monopolistic competition and there is no

reason not to use them if they capture important features of the world.

There is still a group that says ‘Let’s just treat technology as pure private

good and preserve perfect competition’. Then there is another group of economists

who, like Mankiw, say that technology is different, but we can treat it

as a pure public good just as Solow did. I think that both of these positions

are mistaken. There are incredibly important policy issues where the pure

private-good characterization and the pure public good characterization of

technology are just completely off the mark.

Wasn’t your earlier work, as exemplified in your 1986 paper, more concerned

with increasing returns than the determinants of technology change?

You have to look between the lines of that paper at what was going on at the

methodological level, because remember, methodological and formal issues

had been holding everything up. The logical sequence in my 1986 paper was

to say that as soon as you think about growth, you have to think about

technology. As soon as you think about technology, you have to confront the

fact that there is a built-in form of increasing returns – technically, a nonconvexity.

Notice that is all there in Solow’s model. If you look at AF(K, L)

you have got increasing returns in all the relevant inputs A, K and L. So up to

this point, Solow and I are on the same track. You have to think of technology

as a key input and one that is fundamentally different from traditional inputs.

As soon as you think about that, you face increasing returns or non-convexities.

Then you have to decide how to model this from a methodological point

of view. Solow said treat it as a public good. There are two variants of that.

One is that it comes from the sky and is just a function of time. The other is

that the government could publicly provide it. I think Solow had both of

those in mind and it does not really matter which you specify. What I wanted

was a way to have something where there are some increasing returns but

also some private provision. I wanted to capture the fact that private individuals

and firms made intentional investments in the production of new

technologies. So in this sense, the paper was very much about technological

change. To allow for private provision, I used the concept of Marshallian

external increasing returns. This lets you describe an equilibrium with price

taking but still allows you to have non-convexities present in the model. That

was a first provisional step. It was a way to capture the facts: there is some

private control over technology, there are incentives that matter, and there are

increasing returns in the background. What happened between 1986 and

1990 was that I worked hard at the mathematics of this and persuaded myself

Paul M. Romer 683

that the external increasing returns characterization was not right either – just

as the public-good assumption of Solow was not right.

Whenever you write down theories you make approximations, you take

short cuts. You are always trading off the gains from simplicity against the

losses in our ability to describe the world. The public-good approximation

was a reasonable first step, but we needed to keep working and improve on it.

The external increasing returns approximation was something of an improvement

but the later monopolistic competition version [Romer, 1990] was the

one that gets about the right trade-off between simplicity and relevance.

Since Solow’s [1957] paper there has been a huge literature on growth

accounting. What do you think have been the main substantive findings from

this research?

The general progression in that area has been to attribute a smaller fraction of

observed growth to the residual and a higher fraction to the accumulation of

inputs. The way that literature started out was a statement that technology is

extremely important because it explains the bulk of growth. Where we are

now is that technology does not explain, all by itself, the majority of growth.

Initially, we overstated its importance when we claimed that technological

change explained 70 per cent of growth all by itself. But there are some

people who would like to push this further and say there is really no need to

understand technology, because it is such a small part of the contribution to

growth. They argue that we can just ignore it. That is a non sequitur. It does

not follow logically. We know from Solow, and this observation has withstood

the test of time, that even if investment in capital contributes directly to

growth, it is technology that causes the investment in the capital and indirectly

causes all the growth. Without technological change, growth would

come to a stop.

When we spoke to Bob Solow yesterday he explained why he made technology

exogenous in his model. It was simply due to his lack of understanding of

the causes of technological change.

That is a reasonable provisional strategy when you are dealing with a complicated

world.

A great deal of attention during the past decade or so has been focused on

the so-called convergence issue. At the same time as your first important

endogenous growth paper was published in 1986, Moses Abramovitz and

William Baumol also had papers published that drew attention to this catch-up

and convergence debate. This controversy continues to draw research interest,

for example, in a recent issue of the Journal of Economic Perspectives

Lant Pritchett [summer 1997] has a paper entitled ‘Divergence, Big Time’.

684 Modern macroeconomics

When we talked to Edward Prescott two days ago he was reasonably confident

that convergence would eventually occur. Did this important debate

influence your own thinking about growth and what are your views on this

area of research?

It is very important to keep clear what the facts are. The facts are that over the

time horizon that people have looked at the data, say from 1950 to the

present, there is very little evidence of overall convergence. Everybody agrees

about this, even if it is not always stated up front. People who describe this

tendency for countries to converge are saying that if everything else were the

same – if you hold all the right variables constant – then there would be a

tendency for countries to converge. For example, this is one of the key results

in Robert Barro’s work. This is really just a refined statement of the convergence

club interpretation articulated by Baumol. If you look at countries that

have the same values for these variables, then they tend to converge. But it is

also true that in the background, the overall progress towards reduced dispersion

in per capita incomes has been very modest. Pritchett was making a

useful background point. If you go back before 1950, it must be the case that

there was a period where incomes diverged quite a bit – some countries

moved very rapidly ahead as others were left behind. At that time, the overall

distribution of income widened for a period of time. More recently, in the

post-war years, the overall distribution has been roughly constant.

So why do we care about this issue? First you might care about it from a

human welfare point of view, or an income distribution point of view. On those

grounds there is some reason for pessimism – we really have not made that

much progress in the last 30 or 40 years. You might also care about it because

you think it might help you discriminate between different theories of growth –

which ones are right and which are wrong. Many people have asserted that this

process of conditional convergence – everything else equal, incomes converge

– is consistent with a pure Solow style model, that is, one where knowledge is a

public good, all technology is a public good. So they say the evidence is

consistent with the public-good model of technology. That statement is correct

but the evidence is also completely consistent with a model where technology

is not a public good. In this interpretation, the technology gap model, flows of

technology between countries are what drive the convergence process. In this

explanation, the convergence you see is catching up with technology, not just

catching up in the stock of capital per worker. Under the Solow model as

interpreted by Mankiw and others, technology is already the same everywhere

in the world. It is a public good that is in the air like a short-wave radio

broadcast, so under this model there is no room for technological catch-up. It

still mystifies me that people try to justify this model in the face of direct

evidence about the importance of technology flows. But they certainly use the

conditional convergence evidence to back up their position.

Paul M. Romer 685

So I do not think that the convergence controversy has helped us discriminate

between the different models. As a result, I think a great deal of the

attention that the convergence controversy has generated has been misplaced.

Prescott’s assertion is that he does not think that we are going to see continued

divergence. I think he is probably right about that. I personally think that

these flows of technology between countries are very important forces in the

big convergence episodes that we have seen. If you look at a country like

Japan and ask what lies behind its very rapid convergence with the leading

nations of the world, then the transfer of technology was a critical part of the

process. There are grounds for optimism, looking ahead. If we can get the

right institutions in place in these developing countries, the same process of

flows of technologies could be unleashed and we really could see some

narrowing of worldwide income inequality. If you weight it by countries the

situation looks worse than if you weight it by people, at least during the last

ten years. This is because the process of catching up in China will make a

huge difference to the overall picture. And China is a good illustration of

what is wrong with the public-good model. China had a high savings rate

before the reform era. What’s most different now in the sectors of manufacturing

where China has been so successful has been the flow of technology

into China via direct foreign investment – reforms that changed the incentives

that foreign firms faced to bring technology and put it to work in China.

Did you ever look at the work of economists such as Gunnar Myrdal (1957)

and Nicholas Kaldor [1970b], who tended to reject the equilibrating properties

of the neoclassical model in favour of the forces of cumulative causation?

In their models a lack of convergence is no surprise.

It interested me in the same way that Allyn Young interested me. I wanted to

see how much there was in common between what I and what they were

thinking. But it is very hard to tell, quite frankly, when you go back and read

economics that is stated in purely verbal terms. There is always the danger

that you read between the lines and say, oh, they had it exactly right – here is

this mathematical model which shows what they were thinking. But that is

usually based on a charitable reading and one that ignores some of the

ambiguities and confusions. I wrote a paper like that at one point interpreting

Allyn Young’s paper, so one could probably do that for some of the other

economists in this area. For example the big push paper by Murphy, Shleifer

and Vishny [1989a] did this for some of this literature. So the right conclusion

to make is that these were very smart people and they did have some

good ideas, but they were working with very crude tools. I guess I would

describe ancestor worship as a research strategy as probably an unproductive

one [laughter]. But as a consumption activity it is something that can be fun.

686 Modern macroeconomics

Well, we want to keep you on the topic of ancestors for a moment. Given that

your research has concentrated heavily on the influence and determinants of

technological change and the importance of R&D, has the work of Joseph

Schumpeter ever influenced your ideas?

No, I can honestly say that it has not. Schumpeter coined some wonderful

phrases like ‘creative destruction’ but I did not read any of Schumpeter’s

work when I was creating my model. As I said, I really worked that model out

from a clean sheet of paper. To be honest, the times when I have gone to try to

read Schumpeter I have found it tough going. It is really hard to tell what

guys like Schumpeter are talking about [laughter].

Too many words and not enough math?

Yes, and words are often ambiguous.

That problem has also been the source of confusion and the various conflicting

interpretations of Keynes’s General Theory.

Yes, right. Paul Krugman [1994c] has a nice article talking about the big push

idea in development economics. When you state it now in mathematical

terms, the way Murphy, Shleifer and Vishny did, you see how clearly the idea

can be expressed and you wonder why someone had not done it before. I

think that what it shows is that economists now are the beneficiaries of a lot

of development of mathematical modes of thinking and analysis and it seems

very easy to us now because we have those tools to work with. Before these

techniques were available it was really very tough.

Let us go back to the issue of non-rivalry and excludability with reference to

the growth of knowledge and technological change. How do you get the

balance right between encouraging technological change by using incentives

and yet making the new ideas and discoveries available to the rest of society?

There is a trade-off problem here with respect to patent rights.

Sure. What’s interesting about this question is that it is not resolved. If you

take traditional private goods that are excludable and rival, we know what the

best institutional arrangement is: strong property rights and anonymous markets.

That’s all you need. This is a remarkably important insight that economists

must still communicate to the rest of the world. If people understood it, there

would not be so much resistance to pricing roads, pollution or water in

agriculture. Non-economists are still slow to understand how powerful the

price mechanism is for allocating and producing rival goods.

But when you come to non-rival goods, we do not know what the right

institutions are. It is an area that I think is very exciting because there is a lot

of room for institutional innovation. One strategy is to work out a rough

trade-off where you allow patent rights but you make them be narrow and

Paul M. Romer 687

have a finite duration. You would allow partial excludability – less than full

but stronger than zero excludability. We often talk as if that is the general

solution. But in fact, this is not the general solution. You have to break the

question down by type of non-rival good. There are some non-rival goods

like the quadratic formula or pure mathematical algorithms that traditionally

have been given no property rights whatsoever. There are other forms of nonrival

goods like books. You will get a copyright for this book of interviews,

which is a very strong form of protection. The text that you write and my

words – you can take them and put a copyright on them so that nobody else

can reuse them. I can not even reuse my own words without getting permission

from you [laughter]. So that is a very strong form of intellectual property

protection. What we need is a much more careful differentiation of different

types of non-rival goods and an analysis of why different institutional structures

and degrees of property protection are appropriate for different kinds of

goods.

Patent rights or legal property rights are only a part of the story. We create

other mechanisms, like subsidies for R&D. We create whole institutions like

universities which are generally non-profit and government supported, that

are designed to try to encourage the production of ideas. The analysis of

institutions for non-rival goods is more subtle than many people realize.

For example, I have argued that it is very important to distinguish human

capital from ideas – they are very different types of economic goods. Human

capital is just like capital or land. It is an ordinary private good. I agree with

Gary Becker on this. I think a lot of claims about human capital externalities

are wrong. Nevertheless, when people conclude that we should not have any

government subsidies for the production of human capital, I disagree. Why is

that? It is because human capital is the crucial input into producing ideas. If

you want to encourage the production of ideas, one way is to subsidize the

ideas themselves. But another way is to subsidize the inputs that go into the

production of ideas. In a typical form of second-best analysis, you may want

to introduce an additional distortion – subsidies for scientists and engineers –

to offset another – the fact that the social returns from new ideas are higher

than the private returns. You create a much larger pool of scientists and

engineers. This lowers the price of scientists and engineers to anybody who

wants to hire their services to produce new ideas.

So in general, the optimal design of institutions is an unresolved problem.

We have seen a lot of experimentation during the last 100 years. I have made

the claim that the economies that will really do well in the next 100 years will

be the ones that come up with the best institutions for simultaneously achieving

the production of new ideas and their widespread use. I am quite confident

that we will see new societal or institutional mechanisms that will get put in

place for encouraging new ideas.

688 Modern macroeconomics

Research into economic growth has extended into a large number of other

interesting areas in recent years. For example Alberto Alesina and Dani

Rodrik [1994] have explored the relationship between inequality and growth,

Robert Barro [1996], Alberto Alesina and others [1996] have explored the

relationships between democracy, political stability and growth. How do you

view this work? Can we help poor countries more by exporting our economic

systems than our political systems, as Barro has suggested?

Let me back up a little here. One of the disciplines that formal economic

theory forces on you is that you must start with an explicit conceptual

framework. For example, Marshallian analysis makes us think about supply

v. demand when we look at the world. General equilibrium theory forces us

to split the world into preferences and the physical opportunities available to

us. That split is really important and I always try to get my students to think

about it when they approach a question. What do people in your model want?

What are the production possibilities that are available to them?

All of growth theory has been operating under the physical opportunities

question side of the model. We describe the physical opportunities as physical

objects like raw materials and then start to think about ideas as recipes for

rearranging these objects. When you start to think about democracy and

politics, you have to start addressing the other side of the model. What is it

that people want? What drives their behaviour? If you expand the concept of

preferences and say that it is everything that is inside of people’s heads, it

includes all kinds of things that sociologists and psychologists talk about:

tastes, values and norms, and so on. When you start to talk about the connections

between economic growth and democracy you really have to start

enquiring into these issues. Barro’s assertions are based on some empirical

generalizations and they are fine as far as they go, but what is missing there is

any kind of theoretical understanding of the connection between economic

development and political structures. This is not just a problem in economics.

It is also a deep problem in political science. There are many fundamental

issues that have not been addressed in political science. To begin with, why

does anybody bother to vote? The standard theory that political scientists

have is that people go and vote because they have a stake in the outcome and

they want to influence the outcome so it goes their way – fewer taxes and

more transfers, and so on. That theory contradicts itself as soon as you state it

because the probability that any one voter will be decisive in an election is so

trivial that the cost of going to the polls just dwarfs any possible expected

gain that anyone could get from going to the polls.

So I would just assert a cautionary note here. There is a little bit of

empirical evidence that suggests a connection between the level of income

and democracy, but we really face an almost total theoretical vacuum in

studying this question. We are unlikely to make much progress until we have

Paul M. Romer 689

some theoretical foundations that force us to think clearly about the issues

involved.

Another controversial area that has received much attention in the economic

development literature is the relationship between foreign trade and growth.

This is especially topical given the current crisis, which has spread throughout

the ‘Asian Tiger’ economies that are often held up as prime examples of

export-driven growth. As economists we can easily envisage an effect on the

level of GDP coming from trade, but can trade influence the rate of growth?

There are two mechanisms here. From a development point of view the main

thing you want to think about is this process of catching up. The key role for

trade is that it lets developing countries get access to ideas that exist in the

rest of the world. I tell my students that in the advanced countries of the

world, we already know everything that we need to know to provide a very

high standard of living for everybody in the world. It is not that we lack

physical resources; it is not a lack of mass or matter that makes people in

India and China poor. What makes them poor is that they do not have access

to the knowledge and ideas that we have already worked out in North America,

Europe and Japan for doing all the things that we do in the modern economy.

The trick to make them better off is just to get that knowledge flowing into

those countries. Much of it is very basic knowledge – like how to operate a

distribution system so that clothes get from a factory to a store shelf so that

someone can buy a shirt when they want one. How do you make sure that

food does not spoil and is distributed to the right locations at the right times?

How do you implement quality control systems in a manufacturing process?

This is all basic knowledge but it is the stuff that raises living standards. A lot

of that knowledge can be put to work in poor countries if they allow the right

kinds of trade. Direct foreign investment from multinationals, in particular, is

important for getting quick access to these kinds of ideas.

There is also a second issue. If you take the rich economies, OECD

countries for example, the larger the market the bigger the incentives are to

develop new ideas. So free trade in very large market areas creates greater

incentives for innovation and therefore leads to more technological progress.

If you don’t think that this is true, just ask yourself how much innovation

would be taking place in Silicon Valley if products made there had to be sold

just in the USA, or just in California, or just in Santa Clara County? Some, to

be sure, but a lot less than we see right now.

So trade matters for catching up. It also matters for sustaining growth in

the leading countries.

Since growth is so important to the improvement of living standards, it is

inevitable that governments will try to influence the growth rate. What should

690 Modern macroeconomics

the role of government be with respect to growth? In particular, what role do

you see for monetary and fiscal policy here?

On monetary policy it is a bit like the distinction I talked about before –

stopping the bleeding v. getting in shape. There is a certain amount of

emergency medicine that governments have to be prepared to engage in. A lot

of that amounts to an injunction to do no harm. It helps enormously if policy

makers just keep from screwing up the way they did in the interwar period.

But a sensible monetary policy only creates the opportunity for growth to

happen; it does not make it happen. On the fiscal side, a government has to be

able to pay its bills and it must keep from taxing income at such high rates

that it severely distorts incentives.

There are other policies that also matter. Some of those involve creating a

legal framework. What kind of institutions matter if you are in the USA?

Venture capital, fluid capital markets – think of all the things that help a

company like Intel come into existence and grow into a huge force. The

government did not have to do anything very active but it did have to put in

place structures that permitted venture capital, a new-issue stock market and

so forth. Beyond that there are measures related to human capital. There is a

role for government there. The modern university, as it emerged in the USA

in the last century, is one that is very focused on training and practical

problem solving. It is subsidized by the government. As I said before, subsidizing

human capital is a very important way to indirectly subsidize

technological change. So the modern university is an example of the kind of

institution that the government can support.

I should add the caveat that many of the direct roles that people articulate

for the government are not justified. A lot of people see endogenous growth

theory as a blanket seal of approval for all of their favourite government

interventions, many of which are very wrong-headed. For example, much of

the discussion about infrastructure is just wrong. Infrastructure is to a very

large extent a traditional physical good and should be provided in the same

way that we provide other physical goods, with market incentives and strong

property rights. A move towards privatization of infrastructure provision is

exactly the right way to go. The government should be much less involved in

infrastructure provision. So that is one area where I disagree with some of the

wild-eyed interventionists. Another is the notion that the government should

directly subsidize particular research programmes to produce particular kinds

of ideas. If you compare that mechanism with the mechanism of subsidizing

human capital and letting the market mechanism allocate where the human

capital goes and what ideas get developed, the human-capital-based approach

works better. Selecting a few firms and giving them money has obvious

problems. How do bureaucrats get access to all the decentralized information

they need if they are to decide which projects should be supported? How do

Paul M. Romer 691

you keep rent seeking and pork barrel politics from dominating the allocation

process?

A great deal of thought has been given to the design of institutions to avoid

non-trivial rates of inflation. However, the relationship between inflation

performance and growth performance is far from clear, especially at low

rates of inflation. How do you read the evidence on this issue?

Inflation is somewhat damaging and it is probably a non-linear relationship,

so the higher the rate of inflation gets the more damaging it is likely to be.

Is this due to the greater variability of inflation at higher rates?

At least partly. The variability and the higher rates both make the damage

grow more than linearly. There is no trade-off, fundamentally, between growth

and inflation and therefore no reason not to aim at very low levels of inflation

from a growth perspective. The best place to be is at a very low level of

inflation and there is no reason to accept, say, 10 per cent inflation because

we think we can get some benefit in terms of long-run growth. So if you are

trying to do the best job you can on growth, you basically want to aim for

whatever the consensus is on minimal inflation. That will vary between zero

and 2 or 3 per cent at the moment. It may not be too harmful to be up at 6 per

cent instead of 2 or 3, but if it is harmful at all, why accept even that?

During the early 1970s a great deal of interest was stirred up by the book

Limits to Growth [Meadows et al., 1972]. Since then the environmental

movement has become increasingly influential. Do you ever think or worry

about the environmental impact of growth or the possibility of resource

limitations on growth? Can the rest of the world expect to enjoy the same

living standards currently enjoyed in the OECD economies without generating

an environmental catastrophe?

Environmental problems are real problems. They are cases where our current

institutional structures do not put prices on physical objects that should have

prices on them. When you do not have prices on fish in the sea, market

incentives cause fishermen to overfish. We know that we need to institute

either a price mechanism or some regulatory system that has the same effect

as a price mechanism. We will face a big challenge if, for example, human

sources of carbon dioxide prove to be too much for the carrying capacity of

the atmosphere. We are going to have trouble implementing a worldwide

price or a regulatory system to deal with this, but we will need to do it.

However, all this is very different from saying that there are long-run limits

to growth. The way to think about limits is to ask, ‘What does it mean to say

that our standards of living are higher now or that we have more income now

than we had 100 years ago?’ It does not mean that we have more mass, more

692 Modern macroeconomics

pounds or kilos of material. What it means is that we took the finite resources

that are available here on earth and just rearranged them in ways that made

them more valuable. For example, we now take abundant silicon and we

rearrange it into microchips that are much more valuable. So the question is:

how much scope is there for us to take the finite amount of mass here on earth

and rearrange it in ways that people will find more valuable? Here, you can

make a strong case that the potential is virtually unlimited. There is absolutely

no reason why we cannot have persistent growth as far into the future

as you can imagine. If you implement the right institutions, the type of

growth might take a slightly different form from what we anticipated. If

carbon dioxide turns out to be a really big problem and we implement

institutions which raise the price of carbon emissions, then cars will get

smaller. Or we might drive cars somewhat less frequently, or we might rely

on video conferencing, instead of driving automobiles, to meet with family

and so on. We could shift to much greater reliance on renewable biomass or

photovoltaics as a primary source of energy. We have the technology to do

this right now. It’s a more expensive way to generate electricity than burning

oil and coal, but if income per capita is five to ten times higher 100 years

from now, paying a bit more for energy will be a minor issue.

The bottom line is that there are pollution and other environmental problems

that we will need to address. But these problems will not stop microchips

from getting faster, hard disc storage densities from continuing to get higher,

new pharmaceuticals from being introduced, new communications technologies

from emerging, new methods for distributing goods like overnight delivery

and discount retailing from emerging. All those processes will continue in the

rich countries and will spread to the poor countries. In the process, the

standards of living will go up for everyone.

In looking at the post-war economic growth performances of Germany and

Japan compared to the UK, do you think there is anything in Mancur Olson’s

[1982] argument, developed in his book The Rise and Decline of Nations,

that societies which have been stable for a long time such as the UK develop

organizations for collective action which are harmful to economic efficiency

and dynamism?

His conjecture is interesting, but to evaluate it we have to come back to the

discussion we had earlier about production possibilities versus preferences.

What Mancur tried to do was bring back into the discussion some theory

about what is going on inside someone’s head. He wanted to do this so he

could understand the political dynamics that influence policy decisions about

universities, regulations, rent seeking and so on. Those are important questions

both from a development perspective and from a long-run growth

perspective for advanced countries like the UK. These are important issues,

Paul M. Romer 693

but when we think about them it is important to distinguish between assertions

about the physical world and assertions about what goes on inside

someone’s head. Anytime you bring politics into the discussion you are

crossing that divide. At that point it is always important to remind oneself

that we know very little about this area. Mancur is relying on a few empirical

generalizations. He looks at historical episodes where something like a revolution

or a war frees things up and then you see rapid growth. He has also

looked at the general process of the growth slowdown. History is never a

completely reliable guide for these kinds of questions because we do not have

very many observations and the current circumstances are always different

from the past. I always caution someone like Mancur to be honest about the

extent of our ignorance in this area, although I encourage economists to think

about these questions. Just saying that the physical world presents us with

enormous opportunities for growth does not mean that we will necessarily

organize ourselves and take advantage of them as rapidly as we could.

Moses Abramovitz [1986], your colleague at Stanford University, has stressed

the importance of what he calls ‘social capability’ in the catch-up process.

Differences among countries’ productivity levels create a potential for catch-up

providing the follower countries have the appropriate institutions and technical

competence. Can we operationalize a concept like social capability?

Social capability is one of those vague terms like social capital that I think

would benefit from the kind of clarification that you are forced to engage in

when you write down a mathematical model. It could be something that you

understand in this physical opportunity side of the theoretical framework. For

example, you can think of human capital as a key complementary input for

technology. So just as physical capital by itself cannot explain much – neither

land nor labour can themselves produce corn, but the two of them together

can – it could be that human capital is the key complement for ideas or

knowledge just as land is complementary to labour. Just bringing in physical

capital from the rest of the world will not work if you do not have the human

capital there to work with it.

You could also interpret social capability in a broader sense. You could ask

whether a country has a political or social ethic or a set of norms that lets

markets operate, that encourages risk taking, that supports the rule of law as

opposed to either corruption or purely discretionary negotiations. You can

interpret social capability in that broader sense and there are some important

issues there. But when you do this, you have to recognize that you are

theorizing about what goes on in someone’s head.

A great deal of research and effort has been put into investigating the existence,

causes and consequences of the productivity slowdown in the USA and

694 Modern macroeconomics

other advanced industrial countries. What is your personal interpretation of

the findings from this research?

When I talk to students and with people from outside the university, I try to

be honest about our ignorance. It is always very tempting for economists to

claim more than they know. We do not know what happened with the productivity

slowdown in two senses. First, I don’t think we know for sure what the

basic facts are. The quality of the data is such that we cannot speak with

authority and answer the question about what has happened over time to the

rate of growth of productivity. Second, even if there was a slowdown we do

not know the reasons with any confidence. In a recent paper with Kevin

Murphy and Craig Riddell [Murphy et al., 1998] I have started looking at the

labour market evidence which suggests to me that technological change has

proceeded at a pretty rapid but steady pace for the last three or four decades,

neither slowing down nor speeding up. This calls into question some of the

interpretation of the output data that we have, which does suggest that there

has been a big slowdown. But all of the inferences here have to be quite

tentative. You have to be realistic about what you can expect. It could be that

when we get the hard numbers we will conclude that there was a productivity

slowdown and we may never completely understand why it happened. I have

never claimed that endogenous growth theory is necessarily going to be able

to predict or explain precisely all the things that we observe. The economy is

a very complicated beast and the goal for us should not be to predict within a

few tenths of a percentage point the rate of growth, prospectively or retrospectively.

The real test is, does the theory give us some guidance in

constructing institutions that will encourage growth? Does it help us understand

what kinds of things led to difference between the growth performance

of the UK and the USA in the last 100 years? If the theory gives us that kind

of guidance, then it has been successful and can help us design policies to

improve the quality of people’s lives and that is an extremely important

contribution.

Where do you think the direction of research into economic growth is likely to

go next or where should it go next?

I have referred a couple of times to the process of crossing the divide from

thinking only about the physical opportunities to thinking about what goes on

in someone’s head. Once we do that more systematically, we can begin to

understand the choices that individuals and societies make about growth. I

believe that we already know the policies that would speed up growth in a

country like India. What we need to know is why individual and collective

decision procedures in India keep them from impl