Tilting at data windmills
A couple of weeks ago, the standard bearer of economic journalism, the Financial Time’s Martin Wolf, wrote a piece on a recently released voxEu.org-article by Castle (et al.) that had caught my attention as well. I was thinking of doing something similar for other economies, and I might still, but just trying to replicate the darn thing turned out to be an interesting exercise in itself.
Do productivity and real wages behave according to the rulebook?
As usual, when something gets me interested, it is often related to productivity growth or (these days), rather, the lack of it. In the article of interest, the authors state that they have solved the 21st century paradox of stagnating productivity and real wages on the one hand, and rising living standards, measured as GDP/capita, on the other hand (at least for the UK). Their answer is unexpectedly strong employment growth.
Let’s see what they have come up with by having a quick look at how productivity and real wages have developed since the turn of the millennia.
Here’s Castle et al.’s chart:
And here’s ours:
While not quite as nice co-movements as in their chart, it rhymes well with the theoretically consistent expectation of productivity driving real wages. However, when looking at developments on a longer time frame, the discrepancies between our data and those in the article become much bigger.
Macrobond moment: There are a number of research data sets in the database, if you have trouble finding the one you’re looking for, try to enter the database via source and release instead.
The productivity – real wage nexus is, arguably, still apparent, as is the obvious flattening of the trends from ca 2007 onwards (which is, admittedly, the main object of study in the Castle et al article). However, the near-identical developments of productivity and real wage seen in the voxEu-article are harder to detect in our chart. Initially, I thought I had done something wrong, but digging deeper into the data used by the authors, they haven’t – as we have here – used the data on real wages readily available in the BoE “A millennium of macroeconomic data” database.
Instead, they have developed a proprietary dataset (I would imagine that the underlying sources are, nonetheless, similar) and made the – in my opinion somewhat unorthodox – choice of deflating nominal wages with the GDP deflator (instead of CPI or PCE-deflator). They do this, they say, to avoid introducing a wedge in their models and continue by suggesting that the effects are negligible. While I can understand the first claim, I feel less than certain on the latter assertion (and the chart above is a case in point). Most countries do demonstrate a structural wedge between growth in value added prices/deflators and consumption deflators, due to their different compositions of goods/services and related productivity.
In essence, using the GDP-deflator instead of a consumption-based price index (like CPI) is to disregard an important relative price that is fundamental in analyzing the relative bargaining power of labor vs capital and, hence, the distribution of real incomes.
In practice, and in the authors’ example, it apparently means aligning wage developments to productivity developments (since our data gives a quite different result). Even under the (very strong) assumption that developments of value-added and consumption prices are similar, it might still affect the results due to the apparent cyclicality in this relative price, even over shorter time frames such as here.
Hence, we should not be surprised that the choice of deflator has a stark effect on our and the authors’ calculations of the wage share (dividing real wages by productivity). Over the period that they look at, from 1995 onwards, the wage share rises between 1995 and 2002, receding on trend thereafter. In our calculations, it rises through 2009 and has recently started to pick up again, suggesting that normal cyclical patterns still hold.
Either way, and especially for developments since 2000, the stagnation in real wages has apparently been sufficient to change the labor market supply/demand balance and push current unemployment rate below even the (then) record-low levels seen before the GFC.
Macrobond moment: In the unlikely event that the built-in analytical capabilities do not suffice, I hope you know that it is extremely easy to push and pull data between your favourite econometrics/statistics software and the Macrobond application?
Analogously, employment has risen dramatically (+25%) and – importantly – considerably faster than the population (+15%) over the last 25 years. Even with stagnating real wages, such a strong influx of employees will of course raise overall GDP levels.
Indeed, while real GDP per employed has only increased by a measly 6% since the trough in 2009, GDP per capita has increased by almost twice as much during the same period, suggesting a quite decent welfare gain also in the wake of the financial crisis.
In today’s blog, we discussed and tried to replicate a recent article aiming to put to rest the paradox of stagnant productivity and real wages yet a still decent gain of welfare in the UK. While our data exploration confirms the findings, due to methodological issues, we also find the paradox to be… – Well, less of a paradox and more of a normal (albeit, perhaps, unusually pronounced), cyclical event. In other words, understanding the low productivity growth, regardless of whether it appeared in 2002 or 2008, should remain the main analytical objective of any do-good economist. Only then can we start thinking about the cure.
 This measure deviates for several reasons, from the measure of wage share readily available in the BoE database.