According to a recent study in the analysis of productivity trends, Byrne et al. (2016) have argued that the declining growth in TFP and labour productivity cannot be traced to mismeasurement problems in (IT-)service and -goods sectors, like many studies have suggested before (i.e. Brynjolfsson and McAfee (2014); Mokyr (2014)). Byrne et al. (2016) do not find a significantly (growing) relationship between IT (intensity) and economic performance in the US, an indicator for a potential source for mismeasurement. Even in contrast, US-production of IT-products has shifted to places abroad, leading to an even smaller impact of IT-production in US national accounts over the last years. Also, the authors find similar patterns in IT even prior to the productivity slowdown - a weighty argument against the theory of mismeasurement (Byrne et al. (2016, p. 2)).
Besides, Byrne et al. (2016) also reject the argument that modern technologies create additional value for consumers, which does not appear in the statistics (see also Syverson (2016) for this aspect). This and other general measurement problems cannot account for the large gap, the mismeasurement hypothesis has to explain. Their effect is simply too small (Byrne et al. (2016, p. 48); the “too small”- argument is also shared by Syverson (2016) and will be presented for the US and Germany in this study).
According to Mokyr (2014) mismeasurement in the service and information sectors is a rather natural fact. Productivity numbers and the ’traditional’ measurement in statistics (i.e. provided by the Bureau of Labour Market Statistics (BLS) or the Statistisches Bundesamt)
“are designed for a steel- and wheat-economy, not one in which information and data are the most dynamic sector (sic!)” (Mokyr (2014, p. 88)).
It could indicate that there is mismeasurement but related to a poor and old-fashioned measurement framework. If there is an “interplay between science and technology [which, O.Z.] creates a self- reinforcing or ‘auto-catalytic’ process that seems unbounded.” Mokyr (2014, p. 87) but “economists are trained to look at aggregate statistics like GDP per capita and its derivatives such as factor productivity.” (Mokyr (2014, p. 88)), there is no real problem with poor productivity statistics in the classical sense. As already stated in the previous chapter though, an indirect problem exists, when policy-actions are based on wrong data.
However and even before any of these measurement problems were emphasized (they must have existed earlier, as for example Griliches (1988) correctly states; the decade of the energy price problems of the 1970s offers the vast bulk of research in this field), declining trends in productivity can be identified, which were not able to be explained. As serious analyses of productivity slowdowns can be traced back (at least) to the 1970s, the current era in economic history might work as a starting point (or re-starting) for a discussion on measurement errors.
Falling numbers in productivity trends caused by measurement errors can be explained, at least partly, by emphasizing the effect of better technologies on working and non-working activities, which do not show up in the statistics. A phenomenon Robert Solow (1987) has already questioned many years before. He once famously commented in a review article of Cohen and Zysman (1987) in The New York Times in 1987: “You can see the computer age everywhere but in the productivity statistics” (Solow (1987, p. 36)). A try for explanation has been met by Sichel (1997), emphasizing the small impact factor83 of computers as part of the capital stock on economic growth and development - an
argument not tenable nowadays. Besides the “too-small” argument by Sichel (1997), David (1990) argues on the base of a “diffusion-lag” explanation, so that there is a lag between the actual invention and the time when the effects become visible (see chapter 9.2.1 for further details on this discussion). Non-working activities’ utility, more precisely non-market activities’ utility, is derived from a com-
83More precisely, Sichel (1997) argues that the share of computer hardware in relation to the US economy’s capital
bination of market products, sold at a market price, and the ’investment’ of the consumer’s time (Byrne et al. (2016) relating to Becker (1965)). These activities are not limited to the provision of a higher degree of consumer benefits (i.e. gaining more benefit or utility from the same amount of leisure-time). Availability of information almost everywhere and every time (i.e. due to modern information and communication technology) make consumers use their living time more efficiently in general. This also includes less waiting time for public transport, less time spending searching due to availability of modern GPS-based services and many more.
When discussing welfare effects in the context of productivity numbers, the missing link towards people’s utility has to be established and explored. National accounts and any GDP-related indicators, however, do not provide information on the final utility of a person.
Cardarelli and Lusinyan (2015) argue in the same way in a study based on US-state level. Tremen- dous effects of superior technology have improved efficiency of consumers, using their non-market time to produce services they value. Modern communication via smartphones, to provide an example, has shifted efficiency in data availability by years, more likely by decades. Information nowadays is avail- able ’24/7’ and can be shared with others immediately. Business communication via email can be executed whilst walking across the streets or riding public transport; saving up time which was spent as “non-productive” prior to the modern communication age.
The authors also reject an ICT-service-productivity-pattern (declining numbers in productivity- growth correlated to ICT-producing and ICT-using sectors). Not only do certain states in the US or only specific sectors show poor TFP-developments84; it is also shown, that such effects are widespread
over the entire country. If the source of measurement errors were found in the ICT-industry, a statis- tically robust relationship between labour productivity developments and the so-called ’ICT-intensity’ is required.
Syverson (2016) and the subsequent analysis on mismeasurement in this study tackles this propo- sition and finally rejects it. No stable (and statistically significant) relationship between labour pro- ductivity and the ICT-sector of an economy is identified. Even though there is the possibility of measurement error patterns arising in many countries simultaneously (independently from their spe-
84Cardarelli and Lusinyan (2015) use multifactor-productivity (MFP/TFP) as a relevant measure for productivity.
cific ICT-intensity), this makes the validity of the mismeasurement hypothesis less likely, however. Cardarelli and Lusinyan (2015) argue that TFP-decline in the US indeed reflects a true loss of efficiency and/or market dynamism over the last two decades. If ICT-impact were to account for the missing portion in national accounts, then this pattern should appear in more countries. Their conclusion draws a rather dark picture of the (US) economy, as declining efficiency is ultimately linked to international competitiveness and finally less welfare for the US economy.
However, data and a bulk of studies show that there is no significant relationship between a coun- tries’ ICT-sector and the decrease in productivity trends (see Syverson (2016), Cardarelli and Lusinyan (2015), Byrne et al. (2016), Mas and Stehrer (2012), Connolly and Gustafsson (2013), Pessoa and Van Reenen (2014) for example, who provide studies on the ICT- and productivity relationship in other countries).
Syverson (2016) instead uses labour productivity as the relevant measure for productivity but also dismisses mismeasurement as an explanation for the slowdown. Addressing to the mismeasurement- hypothesis, the author provides four reasons (patterns) not to believe in mismeasurement causing decreasing productivity in the statistics. As a side note on his study, he also shows that there is no significant relationship between the size of an economy’s ICT-sector and productivity trends.