Time to revisit growth regressions?

Accelleration (Source: here)

For as long as I have been alive, everyone has been obsessed with economic growth. Politicians care, because growth is so intimately linked to their political fortunes, which means that you, the voter, also cares. Policy-makers care, because growth affects the size of the resource envelope that determines whether taxes and spending must rise or fall. And economists, whose job it is to understand it, care about growth because we are curious people and some of us even believe in improving the human condition. Some 700 million of us still survive on less than USD 2.15 per day.  

While policymakers, politicians, and academics have different motivations for obsessing about growth, they all share one thing, namely a desire to identify drivers that really push growth rates higher. Such drivers have tended to be inconsistent and elusive, so there is always demand for fresh investigations that promise to unearth them.

I, too, once played that game. Some 26 years ago I completed my doctorate in economics at the Centre for the Study of African Economies in the Department of Economics at Oxford University. And, yes, you guessed it, my thesis was on economic growth! A quarter of a century later, I am having second thoughts about the value of the conventional approach of identifying specific growth drivers and even their importance. I do not really believe in a narrow set of magical switches, which can simply be turned on to deliver rapid growth.

My doubt reflects a life-time of observing the enormous creativity and entrepreneurship of the human race. I have visited 184 countries. I have lived and worked on several continents. If there is one thing I am 100% sure about, it is that the potential for rapid economic growth is latent with all of us, embedded in our big brains, which were produced over hundreds of thousands of years of evolution.

My proposition is that if human capacity to produce rapid growth is present within all of us then maybe we need to focus not on drivers, but on constraints. We should ask, "Why is growth not fast everywhere and all the time?" In this blog post, I will put forward the idea that the absence of rapid growth in many countries and, indeed, no growth at all in others, owes more to binding constraints on growth than a lack of conventional growth drivers per se. If this is correct, then policy-makers should rejoice, because it is easier to use policy to address constraints to growth than it is to use economic policy to boost conventional growth drivers.

_____

Before I develop this argument further, allow me first to establish a few basic facts about economic growth. Economic growth is defined as an increase in the production of goods and services over time. For the economy as whole, growth is measured in gross national product (GNP) or gross domestic product (GDP). Usually, we talk growth in real terms as opposed to nominal terms, because nominal variables conflate growth and inflation.

Economists have made considerable progress in understanding where growth comes from. Basically, there are four conventional drivers. First, economies grow when more capital is added. Each additional unit of capital allows a worker to produce more output than he or she would be able to without the extra unit of capital. Hence, the economy as a whole produces more output with more capital.

Second, the economy grows faster when old capital is replaced with newer, more effective capital. The newer, more effective capital allows each worker to produce more output for that unit of capital. The technical term for this growth driver is technical progress.

Third, an increase in the labour force increases the size of the economy. More workers, given capital, produce more output, all else even. This is why immigration is so good for growth.

Finally, the economy expands when workers produce more output for a given stock of capital. Workers become more productive by accumulating so-called human capital through acquisition of skills, trial and error, or experience. Within the broad category of human capital, economists also tend to distinguish between 'social' and 'institutional' capital, which increase output through better functioning of communities and governments, respectively.

_____

In theory at least, economists therefore know what causes growth. Unfortunately, economists continue to struggle to predict, let alone, deliver growth. Why? Human societies, for various reasons many of which are cultural, place major obstacles in the way of the smooth operation of their economies. Feudal system locked most people into serf-dom, condemning them to stagnation. Religion prevented objective analysis of society and the environment until science became paramount. Gender-based discrimination still bars half the population from gainful employment in some countries. And so on.

In other words, the main problem of weak growth may not be a lack of human capacity for innovation and entrepreneurship, factor accumulation, and the other conventional growth drivers, but rather the obstacles, which prevent human beings from realising their potential to innovate and be entrepreneurial in the first place.

This idea really took hold of me the other day as I listened to this year's John Hicks Lecture at All Souls' College, Oxford. The lecture was given by J. Bradford DeLong, who is an economics professor at the University of California at Berkeley. The focus of his address was economic history with specific reference to long-term growth. As a typical academic, DeLong shied away from offering strong views on the causes growth, but he usefully traced economic growth further back in time than I have ever seen before. He also provided a great deal of context.

Specifically, DeLong presented insights about economic activity as far back in time as pre-Homo sapiens some 700,000 years ago! This is just about as far back as we can go and still see fossil evidence of the division of labour in humanoid populations. Better still, he produced estimates of likely GDP per capita growth from circa 70,000 years ago to today. I have reproduced his figures here:

Long-term economic growth in per capita GDP Source: here

The chart has time in years on the x-axis and annual growth rates of per capita GDP on the left-hand y-axis. The right-hand y-axis shows how many times per capita GDP doubles when the annual growth rate is sustained for a full century. For example, if today's average annual global growth rate of about 2% is sustained for a century then the global economy will be more than 6 times larger a hundred years from now.

To me, what really stands out in this chart is that growth rates appear to be exponential. We would certainly expect GDP (in levels) to describe an exponential path, because the growth process compounds GDP even if growth rates themselves are stationary. What we see in this chart, however, is that growth rates are non-linear over very long periods of time.

Non-linear growth rates are genuinely thought-provoking. Exponentially rising rates of growth imply much, much faster increases in per capita GDP than stationary growth rates. They also raise the question what happens in the future as the exponential increase becomes truly explosive. In DeLong's data, growth rates of GDP per capita hover around USD 1200 for about 67,000 years with a lot of Malthusian support bsistence going on, but then since 3,000 years ago starts to accellerate until it reaches about USD 18,000 today, with the majority of the increase happening within the last 200 years.

The possibility of non-linear growth is intriguing. It may even explain why some countries have been able to grow extraordinarily fast for a very time, such as some Asian economies. The real question the chart poses, however, is how such explosion in per capita GDP comes about.

The conventional answer from growth economists would be that a whole bunch of innovations took place all of a sudden. However, this does not really ring true to me. Indeed, the first thought that struck me when I saw the data was that innovations do not impact economies in this manner. Big innovations are like shocks, whose effects tend to be episodic and then slowly mean-revert due to adoption of the innovation by competitors. However, what we see in DeLong's data is a far smoother albeit accelerating process sustained over very long periods.

Such processes are far more reminiscent of what we typically associate with economic convergence or very gradual and sustained economic liberalisation, not sudden innovations. Could the long smooth non-linear path of global growth caused by incremental gains from a gradual relaxation of constraints to growth over time?

I find the thought appealing. Human evolution is extremely slow. As a species, we have had the ability to innovate far longer than just the last 200 years. Or 3,000 years for that matter. Something prevented us from unleashing our potential earlier and it was not our brain capacity. And whatever finally set us free probably triggered non-linearities, say, through network effects, etc.  

On my peregrinations, I have always admired the enormous creativity and entrepreneurship of individual people, regardless of whether they come from the richest countries or the poorest. These qualities are intrinsic to humans, so high growth rates should be possible everywhere. However, this potential is still largely untapped in huge parts of the world. Much faster growth could potentially gush forth, but only we identify the right constraints and remove them.

That is not to dismiss the importance of key inventions, such as electricity and the steam engine. However, I believe these and many other inventions would have happened regardless of conventional growth policies. But they would clearly have occurred with far greater frequently if policy had aimed specifically at removing the constraints. In fact, I do not even worry too much about innovation as a policy issue, because it is in our make-up as human beings to be creative. Instead, we need to place far more emphasis on removing whatever constrains our innovations and prevent them from turning into opportunity, economic activity, and growth.

But wait, I hear the economists object, isn’t what you are describing simply good old-fashioned growth regressions, where you regress growth on a bunch of hypothesized constraints and let the data tell you which ones are binding? Yes indeed, but with a twist. As I mentioned at the top of this piece, I did a doctorate on economic growth a quarter of a century ago. My thesis was empirical, so I collected a dataset, which was longer and broader than anyone had ever used before and then ran a bunch of regressions to see how commodity price shocks impacted growth. The hypothesis was the commodity price shocks - and possibly other manifestations of volatility - have deleterious effects on growth in poorly-diversified and aid-dependent developing economies.

Countless growth regressions were done at the time I did mine, each one with a new set of explanatory variables to test hypotheses about constraints to growth. Indeed, so many regressions were done that today the term given to this type of analysis is the rather unflattering ‘kitchen sink regression'. For some time now, growth regressions have been out of fashion due to various data issues.

My reading of DeLong's data, however, suggests that maybe we should not be so quick to dismiss growth regressions. The way forward is to look at them in a far longer timeframe and to use brand explanatory variables, which differ sharply from those used in growth regressions of old.

In particular, instead of looking at cyclical explanatory variables, such as fiscal deficits, or exchange rate regimes, or corruption, which undoubtedly impact GDP in the short term, we need to focus on much longer-term variables. I would suggest varables, such as the role of superstition (religion) versus science in society, major epochs of war and instability, transitions from feudal to less feudal systems of government, changes in class structures, changes in economic structure, trends in education and health, women's status in society, etc. The idea is to identify constraints, whose removal unleash the innate capacity of humans to innovate and find smarter ways of doing things, not just over the next four or eight years, but for for decades, centuries, forever even.

My own data set was woefully inadequate. I used quarterly data and only covered some forty years. I assumed growth rates were stationary. If DeLong is right, we need much longer data and we must model growth rates as non-stationary variables in some functional form that allows the non-linearities to play out over very long periods.

The main challenge will undoubtedly be finding suitable long-term growth data as well as data on the explanatory variables. To overcome this obstacle, it could be helpful for students of growth to hook up with economic historians for some tips on how to construct really long data series.

The End