Who's in the driver's seat?
The elusive relationship between jobs, wages and prices.
If falling dependency ratios and globalisation have eroded workers’ bargaining power in advanced economies, is there still a link between a country’s employment conditions and its rate of inflation?
In economics, we call the relationship between domestic conditions and inflation the Phillips curve (after a pioneering, mid-20th century economist — a very smart but Heath Robinson-esque figure who built one of the first economic ‘models’ not out of 1s and 0s but out of plastic tubing and water). Phillips’ original work established an inverse relationship between a country’s rate of unemployment and its wage growth (the lower the rate of unemployment, the more wages increase), and his peers quickly extrapolated the relationship to unemployment and inflation (today we often differentiate between the ‘wage’ Phillips curve and the ‘price’ Phillips curve (see box “The Phillips curve”).
The relationship works much better if we make the equation a little more complex and add in inflation expectations. In other words, economic conditions only cause actual inflation to deviate around expected inflation; when demand and supply are in balance inflation is anchored by what is expected to be normal.
These days, the relationship between unemployment and price inflation appears less robust than the original conception linking unemployment and wages. The empirical research, especially in the UK and the US, has shown that the nature of the relationship between domestic conditions and inflation — referred to as the ‘slope of the Phillips Curve’ — has varied considerably over time, but with a tendency for inflation to become less responsive — or the slope to ‘flatten’ — over the past few decades (Constancio, 2015; Haldane, 2017). There are several possible explanations:
— globalisation has increased the importance of the global cycle
— inflation has become better ‘anchored’ — the changing nature of work
— the Phillips curve is difficult to specify.
Let’s address each in turn.
While the evidence for a strong relationship between domestic conditions and prices is decidedly mixed, there is much more evidence that the relationship between domestic employment and domestic wages has stayed strong over the past 30 years (Broadbent, 2014) — at least until very recently. One possibility could be that global competition between firms has reduced their ‘pricing power’ — their ability to pass on wage increases into the prices of final products, settling instead for lower mark-ups. At the same time, increasing competition could make firms more willing to cut prices in response to falling demand. This is supported by evidence that inflation is more sensitive to a fall in employment than it is to an increase (Fabiani et al, 2006; Ciccarelli & Osbat, 2017).
Financial globalisation has increased the correlation of borrowing costs and contributed to more synchronous economic cycles between countries. When one region sneezes, another is now more likely to catch a cold. A number of studies have found that the inclusion of a global inflation factor improves inflation forecasting for the majority of advanced economies (cf. Constancio, 2015 for a good bibliography).
Our own analysis confirms that the traditional, domestic measures used in the Phillips curve (unemployment, unemployment relative to trend, or the output gap) no longer explain variation in UK inflation since the mid-1980s — in the language of economics, we say that they have lost statistical significance. The globalisation of production has increased the importance of international prices relative to domestic factors while financial globalisation has increased domestic inflation sensitivities to global shocks. The global business cycle has some influence (statistically significant) over inflation, but it is not particularly strong. A fall from the very top of the average business cycle to the very bottom may cause inflation to fall (temporarily) by just 0.6%, all other things being equal (see appendix).
ii. Inflation has become better ‘anchored’
Better, more credible monetary policy in advanced economies is a highly plausible explanation of the decreasing sensitivity of prices to both domestic and global conditions. If inflation expectations act as an anchor around which the ebb and flow of the economy may cause actual inflation to deviate, a strong anchor means less deviation.
This is also more consistent with the timing of the flattening of the Phillips curve (changes in employment conditions having less impact on inflation), which was largely over before the collapse of the Berlin Wall and the rise of China as the world’s manufacturer (Berganza et al, 2016; Mishkin, 2008). Central banks, with their explicit inflation targets, have successfully established a strong nominal anchor for inflation expectations. Think of it like this: the business cycle will not cause firms and households to push for/accept permanent higher/lower mark-ups or wages because they expect central banks to ensure that the economy will not overheat or deflate for any prolonged period.
Numerous studies confirm that the importance of long-term inflation expectations, which have converged around central bank targets over the past three decades, has steadily increased (Berganza et al, 2016; Yellen, 2015). Many conclude that this was why we did not observe a deflation spiral after the financial crisis despite very large increases in unemployment (Blanchard et al, 2015). The corollary is that the importance of domestic or international economic conditions has decreased and this is manifest in a flatter Phillips curve. There is also evidence that when inflation is low and everyone expects it to stay anchored, it becomes irrelevant for wage negotiations and other price-setting decisions (Akerlof et al, 2000), again resulting in a flat Phillips curve.
Patterns of correlation across national inflation rates — how one country’s rate of inflation moves in relation to another’s — provide further evidence of the diminished role of the business cycle and the predominant role of central bank anchors. In a speech entitled Inflation in a globalised world, BoE Governor Mark Carney demonstrated how, “by the early 2000s, inflation resembled random variations around countries’ targets, and cross-country inflation correlations fell”. While the deflationary shock of the financial crisis and the rise and fall of the commodity super-cycle have increased correlations of headline CPIs, core inflation rates — inflation excluding food and energy prices — have become less synchronous, albeit around a common, stable anchor (Carney, 2015).
Ben Broadbent, a Deputy Governor of the BoE, has observed that the flatter slope of the wage Phillips curve over the past 25 years looks similar to the slope between 1870 and 1913 — the period that Phillips originally identified himself. This evidence revives the theory that a Phillips curve-type relationship is only stable and significant during periods when there is a clear nominal anchor for prices. Between 1870 and 1913 it was the gold standard; today it is a central bank’s inflation target (Broadbent, 2014).
iii. A change in the nature of work
For the past five years, wage growth itself has been persistently weaker than could be explained by under-employment and productivity — the two factors that tend to drive the variation of wage growth around the long-run trend. Even when we account for ‘hidden’ unemployment — workers who are employed but wish to work more hours — wage growth has fallen short. This has led us to wonder if the structural level of wage growth has been driven lower by a new, emerging phenomenon. This might have caused the wage Phillips curve to flatten as well as the price Phillips curve. We can’t rule out that it is still a lingering hangover from the financial crisis, but another idea is that low wage growth is due to changing patterns of work.
There are almost 1 million more self-employed workers in the UK than there were 10 years ago (according to data from the ONS). The total number of people in work has increased by 3 million over the same period. Self-employment didn’t fall during the recession, so it is tempting to say that it was a response to getting laid off during the crisis, but survey evidence suggests that most of the self-employed are perfectly happy with their current status.
Similarly, there are almost 700,000 more workers employed part-time who do not want a part-time job. There are more than 400,000 more part-time workers who do, and the widely reported growth of controversial ‘zero hours’ contracts plays to that (from 150,000 in 2007 to 900,000 in 2017, although the media storm may mean that some of the increase is due to workers realising — and therefore reporting for the first time — that they are employed on such terms).
Flexible forms of employment can bring many benefits and there is some evidence that its growth has enticed previously inactive people back into work. Inactivity rates — the proportion of the population neither in work nor looking for it — have fallen considerably for 25- to 49-year-olds over the past 15 years. More people entering the workforce may dampen wage growth even as demand for workers picks up.
Further, employment is arguably more precarious when you are paid by the task or the hour. Without obvious peers and colleagues against which to benchmark, without human resources departments and structured career progression programmes, and without the value of human capital that comes with knowing one company inside out, the rise of flexible employment (whether voluntary or involuntary) may have reduced workers’ ability to bargain for higher wages.
It has certainly reduced workers’ ability to bargain in groups. Data collected from job listings websites tell us that there is a far greater dispersion of wages around similar self-employed jobs than there is around full-time pay (Haldane, 2017). To put that another way, there is no ‘going rate’ for the ‘flexibly’ employed. As Andy Haldane, the BoE’s Chief Economist, puts it (with what is becoming trademark poetic flourish), “a workforce that is more easily divided than in the past may find itself more easily conquered”.
The UK government is concerned about what these developments may be doing to the economy (and its tax receipts): it has just commissioned an independent review of the situation. Haldane doubts that changes in working patterns are the main culprit in supressing wage growth, “But they have probably been a contributor in the past and, more significantly, are likely to continue to [be] in the future if these trends, as seems likely, perpetuate.”
If this assessment is correct, inflation may become a little less sensitive to the ebb and flow of employment — the Phillips curve may flatten a little more. The BoE’s database tells us that before the industrial revolution, when cottage industry and farming was the norm and workers had next to no bargaining power, the Phillips curve was as flat as a pancake.
iv. The Phillips curve isn’t stable and it’s difficult to specify
Specifying the Phillips curve is tough. Recently, some statistical relationships appear to have shifted that had remained stable for decades. In some countries, the inactivity rate — the proportion of the population neither in work nor looking for it — has shifted higher; in others the proportion of long-term unemployed has risen. In others still, particularly Europe, the relationship between vacancy rates and unemployment has changed: there has been a pronounced shift in the number of jobs available for every unemployed person, possibly indicating that the unemployed no longer possess the right skills. Something called Okun’s law has broken down across the West. This handy rule of thumb stipulated that for every 1% fall in unemployment you would get a 0.5% increase in GDP (both relative to trend). It started to break down in the late 1980s and has fallen apart since the mid-2000s.
All these changes could materially shift the slope of the Phillips curve. As such, it is very difficult to know what measure of domestic conditions is best. The two standards, unemployment or the deviation of actual output from the economy’s potential (known as the output gap), may just be too simplistic.
The Phillips curve
The Phillips curve is a single-equation empirical model, named after William Phillips, describing a historical inverse relationship between rates of unemployment and corresponding rates of inflation that result within an economy. The curve slopes from left to right, highlighting the trade-off policymakers face between controlling inflation or unemployment.