RDP 2018-10: Wage Growth Puzzles and Technology 1. Introduction: Short-run and Long-run Wage Puzzles

For at least the last six years, and in periods of both rising and falling unemployment, nominal wages growth in Australia has surprised on the downside (Figure 1). A 2015 RBA Bulletin article examined why the decline in wages growth since late 2012 had been so large by historical standards relative to the high and rising unemployment rate: that is, why the Phillips curve appeared to have steepened (Jacobs and Rush (2015); see also Kent (2015)). Two years later, the focus of RBA research was on explaining why, in the face of low and falling rates of unemployment, wages growth was again surprising on the downside: that is, why the Phillips curve appeared to be flattening (e.g. Bishop and Cassidy 2017; Lowe 2017a).

Such observations raise questions about the stability of the Phillips curve and its suitability as a framework for examining what is driving low wage outcomes: a key issue for many central banks. In a late 2017 speech, RBA Governor Philip Lowe stated that understanding why wages growth was so low was a ‘major priority’ for the RBA (Lowe 2017b).

During periods of weak economic growth and rising unemployment, slow growth in nominal wages is to be expected. But the persistence of surprisingly weak nominal wages growth in the context of low and falling levels of unemployment, which has been observed across a number of countries (Figure 2), has been both unexpected and unwelcome. It has led a number of central banks to look for possible structural explanations, with a variety of common causal variables operating across different countries (e.g. Arsov and Evans 2018).^[1]

While this research work has to date raised almost as many questions as it has provided answers, implicit in the behaviour of a number of central banks is the view that longer-lasting structural factors may be at least part of the explanation: that is, the current period of low wages growth may be more than just a cyclical phenomenon. This has been reflected in their willingness to maintain highly accommodative monetary policy settings for longer than would normally be the case, encouraging even lower levels of unemployment. Should this view prove wrong and higher levels of inflation result, they may be costly to reduce. On the other side of the equation, if ongoing structural developments are in fact an important part of the explanation for subdued wages growth, the risk may be one of normalising interest rates too rapidly.

One reason for suspecting that longer-run structural factors may be at play is the observation that wages growth has, in a broader sense, surprised on the downside for decades, across many countries. A standing assumption in the economics profession, at least until relatively recently, was that, over the long run in industrialised economies, nominal wages growth would roughly equal producer price inflation plus labour productivity growth: that is, the wages share of GDP would remain reasonably stable. Indeed, a good deal of modern macroeconomic analysis, along with development economics, was based on this ‘stylised fact’. Keynes himself subscribed to it in 1939, describing it as ‘one of the best-established regularities in all of economic science’.^[2] It was taken a step further by Simon Kuznets in his 1954 Presidential address to the American Economic Association (Kuznets 1955), in which he put forward his famous ‘Kuznets curve’ theory: labour's share of national income, he alleged, at first falls during early stages of industrialisation, then rises, and finally levels out and stabilises as economies move into more advanced stages of industrial development and rising labour productivity is increasingly passed through to the workforce via higher real wages.^[3] This view was seemingly supported by data suggesting that, for most of the twentieth century, the wages share in many western economies did in fact rise as those economies became more industrialised, before levelling out.

However, this cosy view of workers fully benefiting from the fruits of economic growth has been challenged by more recent data. A 2011 International Labour Organization report found that the labour share of national income had fallen significantly in three-quarters of 69 countries for which data exist from the early 1970s to late 2000s (Charpe 2011, pp 55–56). These findings have been confirmed in a number of more recent studies (Figure 3), although the commencement dates for the declining labour share at times differ slightly (e.g. Karabarbounis and Neiman 2013, 2017; Dao et al 2017; IMF 2017b).

A considerable amount of research has been conducted on what has been driving the trend decline in the labour share. This is hardly surprising given its ramifications. The falling labour share is contributing to rising income inequality because, in most if not all western economies, the distribution of income from capital is more concentrated than for income from labour, and hence a shift in factor income shares in favour of profits, other things equal, leads to rising income inequality (e.g. Jacobson and Occhino 2012; Piketty 2014; IMF 2017b). While still somewhat contentious, there is also some evidence of adverse effects of rising income inequality on economic growth (e.g. Mo 2000; OECD 2015; Ostry, Berg and Tsangarides 2014; Brueckner and Lederman 2015).

The importance of understanding what is driving factor share trends, moreover, extends well beyond economics. Downward trends in the wages share of GDP and associated rising income inequality have been widely cited as critical factors in explaining observed disillusionment with mainstream political parties, free trade and globalisation. Indeed, trend increases in inequality have long been seen by some as incompatible with well-functioning democracies.^[4]

While the question of what is driving changes in factor shares has received considerable attention from both academics and bodies such as the OECD and IMF, it has received less attention from central banks. This paper suggests that it warrants closer attention. The factor share puzzle can be seen as either a real wage puzzle – why have real wages been increasing slower than labour productivity – or a long-run nominal wage puzzle – why has nominal wages growth been lower than producer price inflation plus labour productivity growth? Understanding what is driving this may well be important for the conduct of monetary policy. If real producer wages^[5] are increasing less than labour productivity, the difference between the two is being reflected in either higher profit markups or lower output prices, or some combination of the two. Understanding which it is, why and whether the split is changing over time is important for inflation forecasting.

More broadly, both short- and long-run wage puzzles ultimately reflect structural shifts in bargaining power between labour and capital over the distribution of value added. Analysing such structural changes requires models that allow for product and/or labour markets that are not fully competitive and can accommodate changes over time in mark-ups. While to some extent this can be done within a Phillips curve framework, such a reduced form approach is arguably not the best way in which to disentangle and examine slow-moving structural forces that are at play.

Reinforcing the potential importance of central banks understanding what is driving the long-run wage puzzle is its longevity and the fact that it is occurring across so many countries. Both of these features suggest its causes are likely to be structural rather than cyclical. If these causal factors are also relevant to helping explain more recent nominal wage behaviour then, to the extent that they are likely to remain in play or indeed strengthen going forward, central banks can have more confidence in keeping interest rates lower for longer.

A fair amount of work has been done on how the growing take-up of information and communications technology (ICT) and its embodiment in new areas of investment may be affecting nominal wages growth through its impact on bargaining power, job security and product market competition. The RBA has itself done a good deal of research in this area. Most of the empirical results, however, are either inconclusive or help explain only a small part of the slow nominal wage growth ‘surprise’.

Recent academic work that focuses on the broadening but very uneven take-up across firms of the benefits of new technology, in particular ICT (Brynjolfsson and McAffe 2014), may well provide more fertile ground for examining links between new technology and nominal wages growth. A pioneering study by the OECD in 2015 on the diffusion of technology across firms and countries found that, while productivity growth for ‘frontier’ firms remained robust, the productivity gap between frontier and ‘laggard’ firms was increasing across many industries and countries (Andrews, Criscuolo and Gal 2015). Building on these findings, Autor et al (2017) put forward the hypothesis that uneven technology take-up is resulting in technological leaders (‘superstars’) across many industries that have high profit mark-ups and low labour shares of value added, and that this can explain the macro-level decline in labour's share. Empirical tests have confirmed a range of implications that flow from their ‘superstar’ hypothesis, for both the United States and a number of European countries.

This research work linking declining wage shares to the uneven take-up of new technology and different levels of productivity growth across firms may also have important implications for nominal wages growth, on a number of fronts. Firstly, analysis by the Bank of England suggests that the growing gap between productivity frontier and laggard firms across many industries helps explain the slowdown in average productivity growth observed in many countries in recent decades (Haldane 2017a). Low productivity growth has in turn been identified as helping to explain low nominal wages growth across a number of countries (e.g. Brouillette et al 2017; Arsov and Evans 2018). If low productivity growth continues, the capacity for future growth in real wages and living standards will also be increasingly constrained.

Secondly, there is some evidence, both overseas and in Australia, that the larger and more productive firms are increasing their industry market shares as measured in terms of sales or value added, but much less so, if at all, in terms of employment. They are primarily using their market and bargaining power to absorb most of their firm-specific productivity gains into higher profits and lower output prices, rather than into higher wages (e.g. Australian Treasury 2017; Haldane 2017a; Kehrig and Vincent 2017); while at the bottom end of the scale the productivity laggards do not have the capacity to offer anything other than low wage increases in order to remain in business. In short, both the distribution of wage increases and the distribution of employment appear to be much less dispersed than the distribution of productivity growth across firms. The central hypothesis of this paper is that these firm level observations may be key factors behind not just the declining labour share and low productivity growth but also downward pressure on average nominal wages growth.

Such links, moreover, may well be increasing in importance. As the take-up of digital technology spreads – albeit unevenly – across more firms, industries and activities, so will its likely impact on product and factor markets and on bargaining relationships.

Reflecting the above observations, the remainder of this paper is organised as follows. Section 2 looks briefly at the nature of the ICT revolution and associated new globalisation and some of its potential effects on product and labour markets and bargaining power. It then looks at recent RBA research work quantifying the effect of these factors on nominal wages growth, with a more detailed examination in Appendix A. Section 3 evaluates some of the key overseas studies on declining labour shares, in particular recent studies linking it to uneven take-up of new technology and productivity across firms. It also discusses the possible relevance of this framework for Australia.

Section 4 sets out some research proposals designed to further explore these possible links between factor share analysis and the nominal wage puzzle in Australia.

Footnotes

These have included: declining levels of union membership; rising job insecurity; structurally weaker productivity growth; globalisation; automation; hysteresis; more permanently anchored price expectations flowing from the move in the 1980s to more explicit inflation targeting by central banks; and increasing ‘casualisation’ or divisibility of the workforce. For some references see OECD (2017) and Arsov and Evans (2018). [1]

As quoted in Piketty (2014, p 220). [2]

For an interesting and broader examination of the apparent breakdown of some of Kaldor's famous ‘stylised facts’, including constant factor shares, see Eggertsson, Robbins and Getz Wold (2018). [3]

US Supreme Court Justice Louis Brandeis, speaking early in the twentieth century, is alleged to have stated: ‘We may have democracy, or we may have wealth concentrated in the hands of a few, but we can't have both’. [4]

That is, nominal wages deflated by an output price deflator, as against the real consumer wage where the deflator is a consumer price deflator. [5]