Silvia Miranda-Agrippino, Sinem Hacioglu Hoke and Kristina Bluwstein
Can shifts in beliefs about the future alter the macroeconomic present? This post summarizes our recent working paper where we have combined data on patent applications and survey forecasts to isolate news of potential future technological progress, and studied how macroeconomic aggregates respond to them. We have found news-induced changes in beliefs to be powerful enough to enable economic expansions even if different economic agents process these types of news in very different ways. A change in expectations about future improvements in technology can account for about 20% of the variation in current unemployment and aggregate consumption.
Beliefs-driven business cycles
Over the years, different generations of economists have entertained the idea that beliefs, and beliefs alone, could be powerful enough to set the economy on a path to expansions and recessions solely determined by expectations of rosy times being eventually fulfilled, or otherwise. This fascinating view, often referred to as the news-driven business cycles hypothesis, allows economic booms to arise just because economic agents collectively expect some of the fundamental drivers of growth, such as technology, to improve in the future, having received some signal – or news – about it. This wave of optimism sustains more consumption, more investment, and more employment, eventually improving economic conditions in advance of the anticipated technological progress. If, however, technology turns out ex post not to have improved at all, the disillusion, and consequential realization that the boom was in fact disconnected from fundamentals, plants the seed for a subsequent economic contraction. But is this mechanism at all plausible?
Measuring technology news
One of the reasons why the answer to this question has remained elusive, despite the many contributions in the field (see e.g. Beaudry and Portier, 2006, Barsky and Sims, 2011 for alternative views), is that these types of news are spectacularly hard to measure in practice. This has led many to rely heavily on assumptions, sometimes strong, motivated by economic theory, and mostly statistical in nature, in an attempt to isolate the fluctuations in macroeconomic aggregates that could be ascribed to such news. And, as is often the case, different assumptions have led to different conclusions. In our paper, we approach the question from a different and novel perspective. We are able to dispense with the assumptions used in earlier studies by relying instead on a direct measure of technological news that we construct starting from a new dataset of patent applications filed at the US Patents and Trademark Office since 1981. Proceeding in this way grants us two critical advantages. First, it enables us to remain open about the consequences of such news without imposing any patterns on their transmission. Second, it allows us to evaluate to what extent the restrictions imposed in earlier studies find any empirical support.
The reasoning behind our approach is quite intuitive. By their very nature, patent applications constitute a potential promise of future technological improvements. This has long been recognized in the literature, see e.g. Shea (1998). What is new is that we further acknowledge that today’s patents can themselves be the result of past news, or of other concurrent economic factors, and that they measure technological news only with some error. To account for all these aspects, we construct our direct measure of technological news – or an external instrument for the identification of news shocks in econometric parlance – by removing from patent applications the component that correlates (a) with expectations about the macroeconomic outlook that were formed prior to the application filings, as well as (b) with other contemporaneous monetary and tax policy changes that may influence current economic conditions and thus, in turn, individuals’ and firms’ decisions. The resulting instrumental variable is associated with large increases in indices that measure the expected economic importance of technological innovations, and in turn correlate strongly and positively with forward citation counts, a measure of their scientific value.
How does news affect productivity…
Armed with our instrument, we set out to study how the aggregate economy reacts to news about technological improvements that may occur sometime in the future. Technological news induces a slow but steady increase in aggregate productivity (Figure 1). The shape, timing and significance of the response are fully consistent with the well-documented fact that technology diffuses slowly through the economy, that it does so following an S-curve, and that while there seem to be some initial positive spillovers, news does not essentially alter productivity for the first few years. This result can be used to ‘test’ the plausibility of the common assumption made in earlier studies that current productivity does not react to news, which constrains the first point in the figure to equal zero.
Figure 1: Response of aggregate Total Factor Productivity to a technology news shock
Note: TFP is the Fernald (2014) series adjusted for inputs utilization. Impulse response function to a technology news shock identified using our external instrument. Maximum response normalized to 1 percentage point. Results are based on a Bayesian Vector Autoregression estimated on quarterly US data and with 4 lags over the sample 1971:1-2016:12.
…and the rest of the economy
Despite the long time that it takes for news to translate into actual meaningful technological improvements, the expectation that it may happen pushes up consumption, investment, and, consequentially, output as soon as after a few quarters (Figure 2). While it is clearly the case that the news that we are measuring turns out to be true on average (i.e. TFP eventually rises, see Figure 1), the large a-synchronicity in the timing of the peak responses in Figures 1 and 2 suggests that it is beliefs, rather than the actual future improvement in TFP, that play a crucial role in driving the business cycle expansion at short horizons. Our estimates show that they account for close to a fifth of the variation in consumption and employment over periods that go from 2 to 8 years, the normal definition of a business cycle. In this sense, our results support the view that changes in beliefs can generate business cycle type of fluctuations.
Figure 2: Response of output and components to a technology news shock
Note: The figure reports the impulse response functions of real GDP, real consumption and real investment to a technology news shock identified using our external instrument. Results are based on a Bayesian Vector Autoregression estimated on quarterly US data with 4 lags over the sample 1971:1-2016:12.
Not everyone in the economy, however, processes these types of news in the same way (Figures 3 and 4). News can be thought of as signals about the future, surrounded by noise. The extent to which economic agents update their forecasts about the future after having received such signals depends on their information processing capacity. The more they are equipped (or indeed willing) to filter out the noise, the more their behavior – an expression of their updated forecasts – will reflect the revealed signal rather than be determined by current economic conditions (see e.g. Coibion and Gorodnichenko, 2015).
The stock market is quick to price in the expected innovations, and welcomes the news with buoyant attitude. More generous expected valuations, bigger expected dividends, and higher expected profits are all presumably part of what matters in the response of the stock market.
The central bank, presumably instead primarily concerned with the gradual fall in aggregate prices, counteracts the expected deflation by shifting towards an accommodative stance that leads to a decline in nominal short-term interest rates. And, one could argue, despite the fact that future expected productivity gains are typically associated with a higher natural rate of interest. In the paper we show how this translates also into a contraction in risk (term) premia, which offers a potential amplification channel for the effect of news.
What really seems to matter for consumers is instead current labor market conditions. Firms shifting towards new technologies, or equivalently temporarily holding up and diverting investments, reduce total hours worked initially. In the paper, we show that wages also shrink over the same period. Both these contractions are tiny, and do not last more than a couple of quarters. Nonetheless, they bite hard on consumers, and are sufficient to worsen their projected outlook and to erode significantly their confidence in the short term. Consumers, as it happens, only seem to believe what they see.
Figure 3: Response of aggregate prices, the monetary authority, and the stock market to a technology news shock
Note: The figure reports the impulse response functions of the GDP deflator, the nominal short-term interest rate and the stock market index (Nasdaq) to a technology news shock identified using our external instrument. Results are based on a Bayesian Vector Autoregression estimated on quarterly US data and with 4 lags over the sample 1971:1-2016:12.
Figure 4: Response of labor market and consumer expectations about unemployment and index of consumer confidence to a technology news shock
Note: The figure reports the impulse response functions of hours worked, consumer expectations about unemployment one year hence, and the University of Michigan consumer confidence indicator to a technology news shock identified using our external instrument. Results are based on a Bayesian Vector Autoregression estimated on quarterly US data and with 4 lags over the sample 1971:1-2016:12.
Takeaways
Besides the direct application to technology news, our work offers valuable insights on the different speed at which different economic agents – firms, market participants, policy makers and consumers – incorporate signals about the future in their decision making. In turn, this has potentially more general implications for what concerns, for example, the study of aggregate behavior, or the design of policies whose implementation relies on the active management of aggregate expectations.
Silvia Miranda-Agrippino and Sinem Hacioglu Hoke work in the Bank’s Monetary and Financial Conditions Division and Kristina Bluwstein works in the Bank’s Macroprudential Strategy and Support Division.
If you want to get in touch, please email us at bankunderground@bankofengland.co.uk or leave a comment below.
Comments will only appear once approved by a moderator, and are only published where a full name is supplied. Bank Underground is a blog for Bank of England staff to share views that challenge – or support – prevailing policy orthodoxies. The views expressed here are those of the authors, and are not necessarily those of the Bank of England, or its policy committees.
