The underlying data is incomplete and noisy. The models which underpin the rules, and the assumptions which underpin the models, are deeply flawed, and understood in detail by very few people indeed.
Coming up with a measure of ‘output’ (the numerator in the productivity ratios) is a heroic guess, and the assumptions are so slippery that economists don’t even presume to offer the number with an error bar. GDP (gross domestic product) is a fabrication, and a relatively recent invention. The data used to describe ‘input’ (the denominator in the productivity formula) are even more perplexing, and based around extremely problematic survey evidence requiring extensive manipulation and extrapolation.
This is not to say the economists who prepare these numbers are wasting their time. The figures are useful for some purposes: the origins of national income accounting arise from the desire to manage macro-economic policy to keep aggregate supply and demand in rough alignment to avoid inflation and recession. But, normally, serious economists are circumspect and modest about what relatively small changes in these numbers might mean; in academic circles they are normally careful to qualify their statements.
However, once in front of a microphone or when writing for the public, the uncertainties and ambiguities tend to be downplayed; journalists need a story; politicians need a slogan. The discourse proceeds as if the official figures were a straight-forward thermometer reading. Unlike other kinds of statistics in which the idea of ‘significance’ plays an important role, tiny changes in GDP and productivity figures (and GDP growth and productivity growth, and the growth in the growth) are always (and often wrongly) treated as meaningful.
There are several aspects to the difficulty of accurately assessing the measurement of macro-level productivity that are technically interesting. To start with, the procedures for data collection and calculation are not static; it is quite possible that some of the current ‘puzzle’ of UK productivity is explained by methodological changes introduced since the financial crisis.
Two particular problems are how productivity measurements account for intangible services, and how outputs of differing quality are handled; both of these aspects have been subject to revised methods in recent years. Secondly, changes in society and behaviour – for example, the shift of large numbers of professionals from in-house to freelance roles, the rise of social media, undocumented immigration, the flux of activity between private, public and voluntary sectors, and a shorter product life-cycles – all undermine some of the assumptions of a statistical approach that emerged at a time when large, stable organisations with large stable workforces making easily countable things were more dominant.
Even the terminology used to explain elements of the calculation (for example the prices at the ‘factory gate’) are quaintly archaic. Furthermore, a key issue is that aggregate productivity might go up and down merely by distributional effects; if some low productivity jobs disappear, productivity overall goes up even if everything else stays exactly the same.
It is possible, then, that the most interesting feature of the macro-productivity numbers is not the value of the number itself, but the type of discourse that it provokes. Most of the people deploying the numbers do not really understand how they are calculated, or exactly what they mean, and there is ample scope to concoct a wide variety of stories or explanations. Listen to people ruminating on productivity figures, and you can hear a series of plausible myths. In the next section, I examine some of the major strands that emerge in these stories.
This article is part of a serialisation of the white paper "Productivity: A better way; A look at solving the productivity puzzle".