Why Growth Dwarfs Everything

Robert Lucas wrote that once you start thinking about the consequences of economic growth for human welfare, it is hard to think about anything else. The sentence is often quoted as rhetoric. It is closer to arithmetic.

An economy growing at one per cent a year doubles in about seventy years. At three per cent, in about twenty-four. Two countries starting level, differing by two percentage points, diverge by roughly a factor of three within a single working life, and by a factor of ten in a century. Nothing else in economics has leverage like that. Redistributing an existing pie perfectly is a one-off gain of a few per cent; two extra points of growth, sustained, is a different country.

So: where does it come from?

Robert Solow's answer in the 1950s set up the question everyone still argues inside. Output grows because you add labour, or because you add capital, or because of something else. The first two are measurable. Solow measured them, subtracted them from the observed growth of the American economy, and found that the majority of twentieth-century growth was in the something else.

That residual is called total factor productivity, and it is glossed as technology and better organisation, knowing how to get more out of the same inputs. It is also, as Moses Abramovitz put it with some exasperation, a measure of our ignorance: it absorbs technology, institutions, management quality, measurement error and everything unaccounted for. The most important term in the equation is defined by what it is not.

The model also predicted convergence. Capital should earn higher returns where it is scarce, so investment should flow to poor countries; technology can be copied more cheaply than invented, so latecomers should climb faster. Poor countries should therefore grow faster than rich ones and catch up.

For most of the postwar period the data said otherwise. Some countries converged spectacularly (South Korea, Taiwan, later China) and a great many did not, and some fell further behind. What fits better is conditional convergence: countries converge toward their own ceiling, set by their institutions, education and infrastructure, rather than toward the global frontier. That is a considerably less cheerful proposition, because it says the gap is not self-closing.

Which turned attention to the ceiling, and to institutions: the rules constraining power and enforcing agreements. Courts that work. Property that cannot be arbitrarily seized. Contracts that are enforced against the powerful as well as by them. The literature associated with Douglass North and later Daron Acemoglu and James Robinson argues these explain more than geography or culture, and its favourite evidence is the natural experiment: Korea partitioned in 1953, same people, same climate, same history up to the line, diverging to a factor of more than twenty in income per head.

The honest difficulty is that nobody knows where good institutions come from. "Adopt inclusive institutions" is advice without a mechanism, and the countries that need it are the ones whose existing power-holders benefit from the arrangement as it stands. The literature is much stronger on diagnosis than on prescription.

All of which is measured in GDP, so it is worth being clear about what that measure is.

Simon Kuznets built the national accounts in the 1930s to answer a specific question: how much can this economy produce? In his 1934 report to Congress he warned, in as many words, that the welfare of a nation can scarcely be inferred from a measure of national income. The warning was ignored more or less immediately.

What GDP counts is production. So a traffic jam's petrol counts, an oil spill's clean-up counts, and a longer commute counts. What it does not count is unpaid work (childcare, cooking, the enormous volume of care done inside households) which means a country can raise its GDP by paying people to do for each other what they used to do for themselves, with no change in what actually gets done. It says nothing about who receives the output. And it treats the depletion of natural capital as income, in the sense that cutting down a forest appears as production while the loss of the forest appears nowhere.

The cleanest demonstration is a hurricane. Destruction is followed by reconstruction, and reconstruction is activity, so GDP rises. The destroyed houses never appear as a loss, because the headline accounts have no balance sheet. The resources spent rebuilding would otherwise have built something additional, so the country ends up with the same houses and without whatever else it might have had: the broken window fallacy, running at national scale, in the official statistics.

None of this makes GDP a bad number. It makes it an answer to the question Kuznets asked, which is a useful question, and a poor answer to a question he explicitly said it could not answer. The alternatives (median income, life expectancy, the Human Development Index, various green accounts) each capture something GDP misses and each miss things GDP catches, and the sensible position is that no single number is going to do it.

What survives all the caveats is the compounding. Whatever the measure's faults, the difference between a society that grows and one that does not is the difference between your grandchildren living as you do and living as your grandparents did, and no amount of measurement error is going to make a factor of ten disappear.