How to solve a puzzle like productivity

This is a guest blog from Andy Haldane and Gavin Wallis, Industrial Strategy Council

 

There’s a reason that the UK’s weak productivity performance since the financial crisis is still referred to as a “puzzle”. Despite copious amounts of research, including some of our own, we still have a long list of candidate explanations. In part that’s because there is unlikely to be a simple single explanation. But it also reflects us not having good enough data to pin down the causes more precisely. And that makes designing a good policy response, such as an Industrial Strategy, harder – existing research tells us which horses we should back, but not which ones we should bet biggest on.

What is often referred to as “firm dynamics” is persistently found to be an important factor in the productivity puzzle. This basically means the process by which firms are created, grow and close. It is also affected by how the economy, and economic policy, supports this process – for example, by facilitating the reallocation of resources (capital, labour, ideas, innovation) from firms that have closed to those being created. Empirical estimates suggest reduced firm dynamics could account for something like a third of the productivity puzzle.

Getting a good handle on firm dynamics requires you to have good data on business demography – that is to say, a register that tracks firms being created, how they are changing over time and if or when they close.  To understand firms, we need to track their behaviour over the full lifecycle.  Indeed, good and timely data on business behaviour is more important than ever in the current climate, as we try to understand the impact of the COVID crisis on business output and employment in the near-term and productivity and innovation over the medium term.

But the benefits of good business demography data go way beyond regular monitoring of business dynamics. Firms fill in lots of different surveys and often the same survey in successive years.  Combining the insight from these surveys is only possible with good business demography data, enabling richer insights into different facets of firms’ behaviour.

Early UK work in this area focused on combining successive years of a single survey, the Census of Production. That was a lot harder to do than you might imagine, but yielded some important insights. For example, Griffith (1999) showed that foreign-owned plants in the UK car industry have a substantial labour productivity advantage over UK-owned plants.

This has since progressed into linking more than one dataset. For example, Rogers (2006) combined the survey that asks about R&D (Research and Development) spending (the BERD) with the Annual Respondents Database (ARD), which includes a firm-level measure of productivity. That work produced firm-level estimates of the rates of return to R&D for small and medium-size firms (SMEs). Those estimates were very high, suggesting SMEs were constrained in their R&D spending.

These two examples illustrate the usefulness of data-linking when understanding firm behaviour, productivity dynamics and possible policy approaches to raising it. Linking dataset opens up a huge range of analytical possibilities. But it’s fair to say there are often significant barriers to conducting such analysis, including data access and data usability. Good business demography data helps lower those barriers.

That is why we very much welcome the Office for Statistics Regulation review of the business demography statistics and support their recommendations to invest in making improvements to them. Developing a new and improved business register could open-up a wide range of analytical possibilities that could help support frontier research and improved policy making, including around issues of productivity and industrial strategy. This might not solve completely the productivity puzzle, but it would offer an invaluable trail of new clues.