The fraying thread between pay and productivity

Do workers reap the benefits of productivity growth?  Few questions are more central to the conundrum of faltering living standards. If the 20^th century was a golden era for material wellbeing in Britain, that’s explained by one factor above all others: from 1900 to 2000 UK labour productivity grew roughly fourfold, translating into unprecedented growth in real earned income.

Of all the findings from our recent work at the Resolution Foundation, then, few are more worrying than those that suggest a weakening of this fundamental relationship between economic growth and gains for ordinary workers.  In the last twenty years of the 20^th century, each pound of UK GDP growth was accompanied by around 90 pence of median wage growth. From 2000 to 2007 that figure fell to just 43 pence. This development bears worrying similarities to the experience of the United States where median wages have now been flat for a generation. It is not an exaggeration to say that the pay of Americans in the bottom-half has decoupled from productivity growth.

So is a similar form of decoupling now happening here? And, if so, what’s driving it? The truth is that these are very tough questions to answer, not least because they raise a whole host of definitional issues. How should we measure productivity if we want to compare it to pay? Whose pay are we talking about anyway? What do we even mean by pay (for example, should we include things like employer pension contributions)? And what does economic theory say about the relationship between pay and productivity—should we really expect a perfect link between the two?

This week a new report from the Resolution Foundation by Professor John Van Reenen and João Pessoa of the London School of Economics provides the first definitive answer to these questions for the UK labour market. The paper looks at how the relationship between productivity and compensation for workers has changed over time in the UK and in the US, building on similar work carried out in the US by Jared Bernstein.

So has decoupling taken place in the UK? As ever with difficult questions the answer is that it depends what you mean. The report defines two types of decoupling, each of which tells a very different story. Figure 1 shows Van Reenen’s results for the first type, defined as the relationship between productivity and average hourly compensation. The paper refers to this phenomenon as ‘net decoupling’ and, on this measure, the answer is clear: compensation for ordinary workers has not fallen behind productivity. In fact, aside from short periods in the late 1990s and mid 2000s, the two measures have tracked each other step for step since the early 1970s.

As Van Reenen points out in the paper, this finding is just as economic theory would predict. The only explanation for the two lines diverging would be if labour’s share of national income fell versus the capital share, which in theory should remain broadly constant over time. While some have suggested that just such a decline has taken place, Van Reenen’s results suggest it has not.

Figure 1: No net decoupling?

Average hourly compensation and labour productivity

Source: Pessoa and Van Reenen, Resolution Foundation (2012)

This brings us to the second type of decoupling, which is shown in Figure 2. Again, the measure of productivity used is GDP per hour worked in the UK economy. But this time worker benefits are represented by median hourly earnings. The paper refers to this relationship as ‘gross decoupling’, and using this different measure the picture looks very different. Median pay has diverged markedly from labour productivity in the UK in the last twenty years. In other words, the pay of ordinary working people has not kept pace with the average value of output that workers produce.

Figure 2: Strong gross decoupling

Median hourly wages and labour productivity

Source: Pessoa and Van Reenen, Resolution Foundation (2012)

So what’s the difference between the two types of decoupling—and which one should we care about most?  The answer lies in three differences between the two measures used to represent the benefits that accrue to workers.  First, Figure 1—‘net decoupling’—is based on an average measure.  As a result, it captures the total compensation going to workers in the economy divided by the number of hours worked. By contrast, Figure 2 is based on a median, specifically the hourly wage of the middle worker. The two therefore diverge when the average is pulled up (and the median isn’t) when pay grows very strongly at the top, as it has in the past ten years.

Second, Figure 1 looks at total compensation rather than just wages. This means it includes things like employer pension contributions and employer National Insurance Contributions (NICs), which Figure 2 doesn’t. As such, there’s a reasonable argument that Figure 1 is a better gauge of the complete rewards derived by workers. Again, recent years have seen a widening gap between these two measures. As non-wage aspects of compensation have grown significantly, compensation has grown faster than wages.

The third difference, and the most technical, is that the two measures are calculated using different inflation indices. The first, used for net decoupling, is calculated using the GDP deflator, while the second uses the Retail Prices Index (RPI).  What’s the difference and which one is right?  The answer is that it depends what you’re analysing.  If you want to compare productivity and pay fairly, you should use the same deflator for both—that is, the GDP deflator. But if you want to know how the purchasing power of pay is changing over time, you want to use the RPI. These two measures have moved apart very slightly in recent years. As the paper acknowledges, this trend is hard to interpret.

You might fairly say these are just arguments over definitions without much bearing on policy debates. It’s certainly true that their significance rests very heavily on interpretation. Broadly speaking, I’d argue there are two ways of thinking about this. On the one hand, we could find Figure 1 reassuring. We might note that it confirms the theory that, over the long-term, productivity should track average compensation. In fact, looked at this way, we might even think Figure 2 is simply wrong; once we use the ‘right’ measure of the benefits accruing to workers, as theory dictates, the decoupling that we thought was there turns out to be a phantom.

This argument is not without merit; it’s fair to say, for example, that total compensation is a more complete measure than wages. But what if we look at the question another way, from the viewpoint of a worker in the bottom half of the earnings distribution? Does Figure 1 then make us feel any better?  I’d argue not.  The question that matters for living standards, after all, is not which measure should theoretically track productivity, but which most accurately captures how well off people feel, and whether we’ve seen a change in the relationship between that measure and output, our central indicator of economic health.

In this sense, saying that decoupling is just a statistical quirk of using the median rather than the mean is cold comfort. It doesn’t make it less of a problem that the wages of low paid workers are stagnant and no longer tracking productivity in the way they used to; it just helps to explain why life now feels particularly hard in the bottom half and—in the run up to the crisis—particularly out of step with an ostensibly sunny economic climate.  Likewise, the growing gap between wages and total compensation doesn’t make the stagnation of pay packets any less real (though it is far from obvious that we’d better off if we paid less into our pensions and took more home as wages) it just helps to explain that stagnation. As Jared Bernstein pointed out in a seminar to discuss Van Reenen’s paper late last year: whatever the theory says, if workers weren’t getting better off over time even when our economy was growing and productivity was rising, we have a problem on our hands.

In this sense, the most important takeaway from this new report for the Resolution Foundation is that it helps us understand which factors have contributed to the decoupling of median pay and productivity. The answer is that inequality has played a big role. So has growth in the non-wage aspects of compensation. The next challenge is to understand which of these trends are prices worth paying and, for those that aren’t, the dynamics that lie behind them. Only then will we bring into focus one of the most important aspects of the crisis now facing living standards: the fraying of the golden thread that joins pay and productivity, which for much of the 20^th century helped pull prosperity to unprecedented highs.