Public trust is a cornerstone of central banking. Shocks to trust depress economic activity and may cause inflation expectations to increase. This column introduces a new measure of public trust in the Federal Reserve derived from nearly 4 million tweets about the institution. Macro-financial factors affect trust, as does Fed communication. But scandals that bring into question the integrity of key officials cause the largest erosion in trust. While the shocks have only a short-lived impact on the trust index, their broader economic effects are persistent.
It’s hardly a surprise that public trust in many central banks has taken a hammering of late. In recent years, inflation rates have hit double digits in many countries for the first time in a generation; we’ve also seen high-profile bank failures on both sides of the Atlantic and mounting losses from past quantitative easing programmes – e.g. the Bank of England is currently forecasting a lifetime loss of £130 billion, around 5% of annual GDP.
But how can we measure the public’s trust in a central bank and why does it matter? In a recent paper (Aikman et al. 2024), we introduce a new measure of public trust in the US Fed derived by applying generative AI to nearly 4 million tweets about the institution, its leadership, and its policy framework and decisions. In this column, we explain how this new measure was created and summarise its properties – including the factors that seem to affect it. We also analyse the effects of exogenous shocks to trust, using a narrative identification approach that maps a set of alleged ethical scandals that have embroiled members of the Federal Open Market Committee in recent years into negative shocks to trust.
What do we mean by trust?
The economics literature has long emphasised the importance of credibility in macroeconomic policymaking. Credibility here refers to the likelihood that the central bank will fulfil its stated commitments, e.g. commitments to reduce inflation from an undesirably high level or to maintain interest rates at a low level for an extended period to boost economic activity. Kydland and Prescott (1977), Barro and Gordon (1983), and others taught us that credibility in this sense hinges critically on making commitments that align with the incentives of policymakers. Credibility can also be fostered by the central bank’s track record of honouring previous commitments.
By trust, we have in mind a related but broader concept. Trust captures the public’s belief in the goodwill or integrity of the central bank, its leadership, and its staff. It also captures an expectation that the central bank will act in the public interest rather than prioritising special interest groups. And it encompasses the idea that the central bank is perceived to have the technical competency to meet the goals it has set. 1 While credibility can be established through formal mechanisms, trust relies on informal norms and social capital, making it more fragile.
But public trust is a cornerstone of central banking. It is well understood that trust of course underpins the public’s willingness to hold central bank money. It also reinforces the transmission mechanism of monetary policy and helps the central bank restore confidence during periods of economic crisis and uncertainty. And by cementing the legitimacy of the central bank to act in the public eye, it serves as a key bulwark against the threat of short-term political influence over monetary and financial policy decisions. See Ehrmann (2024) for a comprehensive review of the literature on trust and central banking.
How can we measure the public’s trust in the central bank?
Traditional approaches for measuring trust in central banks and other macroeconomic institutions rely on public surveys (e.g. van der Cruijsen et al. 2023, Goodhart and Vu 2024).
Despite their widespread use, surveys have various shortcomings as measures of central bank trust. One is the limited capacity they offer in practice to gauge why responses have changed – surveys are relatively infrequent, and the questions tend to be closed-end (e.g. ‘Do you tend to trust or not to trust the ECB?”). Another is the concern that respondents might interpret the question in quite different ways, with responses capturing views about related institutions or general gut reactions. Survey-based measures also suffer from well-known problems including social desirability bias, non-response bias, and the effects of question framing.
Our new measure of public trust in the Fed is based on a very large sample of tweets about the US Federal Reserve. Specifically, we downloaded every tweet about the Fed relating to its policy framework, competence, and ethics/governance. After filtering out tweets that solely provided objective information about forthcoming meetings, testimony, etc., we were left with around 4 million tweets, providing a rich source of daily information about Twitter/X participants’ views of the Fed. We then asked ChatGPT to categorise these tweets according to whether they were supportive, critical, neutral, or unrelated to the Fed. 2 Our measure is presented in Figure 1, graphed from start-2010 until end-2023.
Figure 1 Index of public trust in the Fed
Note: The lightly shaded grey line shows weekly values of the index, obtained by summing daily values Monday to Sunday. The bold red line shows a four-week moving average of the index. The y-axis shows the different between the positive and negative share of tweets, measured in decimals.
The chart shows the difference between the share of positive and negative tweets each day. It fluctuates significantly over the period – perhaps unsurprisingly given that this was a period that included the aftermath of the global financial crisis, public criticism of the Fed by various senior politicians, three different Fed chairs, the COVID-19 crisis and its aftermath, various quantitative easing programmes, a banking crisis, and the jump and subsequent decline in inflation.
The mean of the index is clearly negative, and there appears to be a mean shift in 2018 when the index declines significantly. It reaches its lowest point in the first quarter of 2020, a period that coincides with the early stages of the COVID-19 pandemic. Thereafter, it rebounds strongly up until mid-2021, when this recovery was halted in its tracks by the surge in inflation. There is some evidence that our trust measure approaches its pre-2018 mean in the final year of the sample, as the surge in inflation dissipates.
What affects public trust in the central bank?
Our measure of public trust in the Federal Reserve has an intuitive relationship with a wide range of plausible determinants, as we summarise in the figure below, which plots regression coefficients.
Figure 2 Determinants of trust
Note: This figure shows regression coefficients where the dependent variable in the regression is the trust index. The regression includes a constant and 5 lags of the dependent variable. It is estimated on daily data over the period 2010–2023. N=4.378. The adjusted R2 is 15%. For brevity, we only report coefficients that are both statistically and economically significant. All continuous right-hand side variables have been standardised. Macro-financial factors like the yield curve, inflation expectations, and market volatility all have statistically significant effects, with the largest impact coming from the Fed funds rate – a 1-standard-deviation increase in the Fed funds rate reduces net trust in the Fed by 11 percentage points.
Macro-financial factors like the yield curve, inflation expectations, and market volatility all have statistically significant effects, with the largest impact coming from the Fed funds rate: an increase of one standard deviation in the Fed funds rate reduces net trust in the Fed by 11 percentage points.
As might be expected, other monetary policy-related indicators and events also matter. The publication of the Monetary Policy Report and the annual Jackson Hole speech by the Fed chair both enter the regression with large positive coefficients. This highlights the importance of Fed communication in shaping public understanding and sentiment. The announcement of new Fed chairs, as well as their confirmation and swearing-in dates, also seem to have bolstered trust in the institution.
Events unrelated to monetary and financial policy also exert a strong influence on the trust index. We capture two such exogenous drivers in our narrative proxy. First, many senior Fed officials have been embroiled in ethical scandals, including several cases of alleged insider trading, and we code the days when the news of these alleged scandals broke via dummy variables. 3 Not all shocks to our trust metric are negative, however. Our proxy also includes the tribute to Janet Yellen (using the hashtag #PopYourCollar) celebrating her legacy as Fed chair, temporarily boosted trust. These events have the largest measured effect on the trust index – they reduce our measure by 20 percentage points. We also find that tweets by then President Trump that are perceived as threatening the Fed’s independence (see Bianchi et al. 2023) have large negative effects on our measure.
What is the effect of a shock to trust in the Fed?
We explore the effects of exogenous shocks on our trust index in the context of a vector autoregression model. The model is estimated on daily data and captures the dynamic relationship between our trust index and eight macro-financial indicators that were found to be significant drivers of it.
As an instrument to capture exogenous trust shocks, we use our narrative proxy, which has two components: first, it includes the #PopYourCollar tribute to Janet Yellen as she left office; second, it includes the reporting of alleged actions by Fed officials that contravened ethical guidelines. So, in our identification, we construct a narrative proxy that takes the value -1 on the days the news of the scandals was reported, 1 on the day the #PopYourCollar trend was launched, and zero otherwise. As reporting lags are random and these news move our index materially, the case for this being a valid instrument is very strong. 4 Figure 2 shows the impact of a negative one standard deviation fall in our trust index.
Figure 3 Narrative proxy
Notes: Impulse response functions to a trust shock identified using our narrative proxy (scandals + #PopYourCollar) as instrument. The dark and light shaded areas correspond to the 68% and the 90% high probability density (IIPD) sets, respectively.
A shock to trust leads to a sharp contemporaneous drop in the trust index, which however recovers quickly thereafter. The shock lowers the daily news sentiment index on impact; it also increases financial market uncertainty (VIX), causes the stock market to fall (NASDAQ), and leads to tighter business conditions (ADS). The break-even inflation rate over the next 5 years increases significantly before subsiding, while we see a temporary (small) rise in the current Fed funds rate and the Fed funds futures rate. This pattern of responses is consistent with a supply shock: trust shocks worsen the inflation–output trade-off, in line with the structural model of Bursian and Faia (2018), where trust emerges as an equilibrium in a game between betrayal-averse agents and the central bank.
Interestingly, despite their short-lived impact on our trust index, the broader economic effects of trust shocks are persistent – the impact of the shock is still felt in macro-financial variables six months after the shock.
How can we reconcile the persistence of these effects on the economy with the short-lived impact on our trust measure? One interpretation of these findings is that our trust index is not well suited to capturing persistence in the dynamics of trust itself. As a social media-based measure, attention will inevitably move on after a few days or, at most, weeks. As such, it might not capture latent erosions in trust, which would become apparent only when the next shock occurs. For this reason, we think our measure is perhaps best viewed as providing a fast-moving early read on shifts in trust.
Implications for policymakers and for policy
What message should central bank policymakers take from our results? First, ensuring the integrity and accountability of central bank officials is key to maintaining public trust. The events that cause the largest erosion in trust in our sample are scandals that bring into question the integrity of key officials. The number one priority for shoring up trust in the central bank must therefore be ensuring that central bank policymakers and staff have an appropriate formalised code of conduct, which includes restrictions on actively trading financial instruments that might be perceived as benefitting from inside information, while dealing swiftly with any incidents of unethical behaviour.
Second, our results reinforce the notion that proactive and transparent communication are important tools for bolstering the public’s trust in the central bank. We find large positive effects on trust from the publication of the Monetary Policy Report and the annual Jackson Hole speech by the Fed chair. We also find positive effects from the #PopYourCollar trend that was initiated by the New York Fed as a tribute to Chair Yellen in her final days in the role. But more fundamentally, it probably means being more open to criticism than has hitherto been the case for central banks. A good example here is the recent review of the Bank of England’s forecasting framework conducted by Ben Bernanke. While Bernanke’s report does not pull its punches and is openly critical about the state of the Bank’s forecast infrastructure, our overall sense is that the Bank is being applauded for having the self-confidence to invite this report in the first place.
Third, our analysis highlights the need for central banks to develop timely measures of the public’s trust. Shocks to trust depress economic activity and may cause inflation expectations to increase. While these shocks have not been of sufficient magnitude to influence US business cycles materially over the recent past, there is no guarantee that this will continue to be the case. With the potential for greater political interference in US monetary policy in the future, understanding how central banks should act to retain public trust has become ever more urgent.
Source : VOXeu
