Anxiety, Competing Narratives and the Macroeconomy: what is the role of policy in stabilising expect

by Francesco Carbonero, Jeremy Davies, Ekkehard Ernst, Sayantan Ghosal, Leaza McSorley

“It is certain that a large part of what we experience depends not all the actual circumstances of the moment so much as on the anticipation of future events”.

- W. S. Jevons, The Theory of Political Economy

Since the 2008 global financial crisis, automation and the rise of the gig economy, a revival of populism and nationalism, Brexit and now of course the COVID-19 pandemic have led to a perception the future is increasingly uncertain, resulting a shift towards increased anxiety.

Anxiety refers to a negative emotion regarding the anticipation of a future event. In other words, it constitutes a psychological payoff experienced by a decision-maker today in response to perceived future risk (Caplin and Leahy, 2001). Anxiety aversion implies that a higher perceived future risk will trigger a higher negative psychological payoff today. When anxiety rises, economic actors put more weight on (expected) negative outcomes, overreacting to negative news and discounting positive ones. As the psychological payoff associated with anxiety aversion is experienced today, a higher degree of impatience is associated with a higher degree of anxiety aversion. Risk aversion, on the other hand, captures the impact of current risk on current payoffs. So, a variation in the discount factor should have no effect on risk aversion but will change the degree of anxiety aversion.

Anxiety, multiple narratives, and aggregate welfare

Rousseau’s parable of the stag-hare hunting game (Rousseau, 1754) is useful here. Rousseau uses the game to contrast the gains of hunting hare, where the risk of non-cooperation is small or non-existent (in our case, investing in a riskless low return storage technology like a savings deposit) and the individual reward and social benefit equally small, against the gains of hunting the stag (in our case, investing in the risky asset whose returns depend on average level of investment), where maximum coordination is required and the risk of non-cooperation is greater but both the individual reward and social benefit (if successful) are much greater. The point is that hunting the hare is risk dominant as success in this activity does not depend on what other agents do even though coordinating to hunt the stag is Pareto dominant.

In our model, in the presence of multiple narratives about future risks arise, no individual decision-maker can predict what other decision-makers will do today: such strategic uncertainty in the presence of multiple narratives enhances perceived future risk and hence anxiety today. In turn, individual decision-makers become cautious and coordinate on risk-dominant outcomes. Hence, the level of anxiety is endogenously determined.

In such a situation, policy actions – rather than simple announcements – can act as a focal point in effect, as a lighthouse. Traditionally, such effects are thought to be the result of economic actors complying with certain regulations even though they are not directly affected. In our understanding, lighthouse policies can have the additional effect of guiding the economy through times of high economic anxiety by stabilising expectations and creating a degree of common knowledge about the future mitigating strategic uncertainty today, enabling welfare enhancing productive investments.

Measuring economic anxiety and its impact

We measure anxiety through a machine learning algorithm that applies sentiment analysison news articles published online by Daily Mail, Reuters and Press Association. First, we label a training set of articles with different values of anxiety from “no anxiety” to “high anxiety”. Then, we apply these labels to the whole archive of article the website dailymail.com. These labels are then rescaled on a score between zero and one.

To assess the plausibility of our anxiety score, we compare it with the rise of a general feeling of anxiety at the onset of the Covid-19 pandemic, unrelated to specific economic perspectives. Figure 1 plots the mean of survey responses together with the weekly mean of economic anxiety. Our economic anxiety indicator shows anxiety starts increasing rapidly in February, peaking mid-March. The figure depicts a good overlap between the two series, meaning that we are close other comparable measurement of anxiety. Moreover, economic anxiety maintains peak longer and reduces more slowly than the self-reported subjective anxiety recorded in the ONS Well-being and Quality of Life Survey. This signifies greater amplification coming from economic perspectives than from general personal conditions.

Figure 1: Comparison between economic anxiety (weekly mean of the daily indicator) and personal anxiety computed by the ONS (week level). Year 2020. Source: Authors’ calculations and ONS.

Note: The survey asks ”Overall, how anxious did you feel yesterday?” and answers are given on a scale of 0 to 10, where 0 is ”not at all anxious” and 10 is ”completely anxious”. Data is available on a weekly basis from March 2020 onwards. Source: ONS, retrieved on 22nd of October 2020.

Our theoretical analysis predicts that an increased level of anxiety reduces investment in risky assets and the volatility of asset prices. We test this prediction by estimating the impact of anxiety on stock market volatilities, using daily differences between maximum and minimum prices of FTSE250 from 2019. Figure 2 demonstrates indeed that both series show counter-cyclical movements: The spikes of anxiety mimic the drops in stock market volatility. This countercyclicality is confirmed by the negative correlation between the two series (ρ = −0.50). Furthermore, there seems to be a timid anticipation of anxiety with respect to stock market volatility.

Figure 2: Economic anxiety and stock market volatility. Daily data (HP filtered) over 2019. Left shaded area corresponds to period: 1st of January – 18th April. Right shaded area corresponds to the period: 8th of November – 16th of December). Source: authors’ calculations and investing.com.

Next, we apply a Vector Autoregression (VAR) model which exploits the joint dynamics of anxiety and stock market volatility as a direct test of the impact of anxiety on stock market outcomes. The causal impact is extracted by leveraging the key characteristics of anxiety, namely, its forward-looking generation. Figure 3 shows the short and cumulative impact of anxiety on stock market volatility. We find that a positive anxiety shock impacts negatively on stock market volatility. This effect is statistically significant with a 2-days lag, it remains significant for the subsequent two days and then become insignificant. The cumulative impact is negative and significant (the size is a cumulative drop of 3 log points in stock market volatility for an increase of one standard deviation in anxiety).

Figure 3: Impact of anxiety on stock market volatility. Vertical axis: orthogonal (cumulative) impulse response function of stock market volatility obtained from a reduced form VAR(2) of economic anxiety and stock market volatility. Horizontal axis: days following the shock of anxiety.

Lighthouse policies and anxiety

In times of crises then the role of policy to set in place a credible floor to the level of investment in risky assets. By stabilizing expectations and creating a common, credible narrative, lighthouse policies can help address economic anxiety. One example in this regard is to create a floor on expected returns through policy announcements. At the heyday of the European sovereign debt crisis in 2012, for instance, ECB President Mario Draghi, for instance, announced to undertake “whatever it takes” to save the euro. Without spending a single euro by intervening in the market, this simple – and credible – announcement created a floor against the speculation regarding the sovereign debt of individual member countries.

Similarly, when the coronavirus pandemic struck markets responded by flocking to safe havens/assets and a run towards gold and government bonds. The scale of the coronavirus shock resulted in central banks increasing the supply of government bonds to match the demand for safe assets and investments to prevent a sudden rise in spreads for assets considered to be high-risk. Even when bond yields turned negative investors where still choosing ‘safe’ government bonds at a negative rate of return because the alternatives – provided even greater risk of loss). Distinct from myopia, anxiety aversion (the psychological loss experienced today due perceived increased future uncertainty) was greater than monetary loss today (the guaranteed minor loss on government bonds), leading to a massive drop on stock market that could be reverted through a broad-based provision of liquidity across different asset classes.