The Economics of the Fed Put

Abstract

The effect of the stock market on monetary policy is a topic of long-standing importance, with policy makers having to decide both whether to accommodate stock market slumps and whether to lean against the wind in times of stock market booms. This question has received renewed interest in the years following the financial crisis. Some observers have criticized the Fed for being excessively driven by asset prices, the stock market in particular, rather than by economic data. For example, discussing rate increases, former Fed governor Kevin Warsh (2016) has stated: “[...] They look to me asset price dependent, more than they look [economic] data dependent. When the stock market falls like it did in the beginning of this year, they say: ‘Oh, we’d better not do anything.”’ In this paper, we empirically revisit the question of whether the Fed is responding to the stock market and, if yes, why.

Establishing a causal impact of the stock market on policy is difficult because the stock market and policy react to each other, and underlying macroeconomic news affect both. Rigobon and Sack (2003) use identification through heteroscedasticity and find that a 5% decrease in the stock market causes a 14-basis-point (bp) reduction in the federal funds rate (FFR) target at the next Federal Open Market Committee (FOMC) meeting. An alternative approach is to estimate Taylor rules that include past stock returns in addition to the Fed’s forecasts of macroeconomic variables. Based on such estimates, Bernanke and Gertler (1999) find a modest negative but statistically significant effect of stock returns on the FFR target. They argue theoretically that monetary policy should not respond to the stock market beyond its effect on inflation expectations and conclude that actual policy by the U.S. Federal Reserve is largely consistent with this advice.

Subsequent literature builds on the above approaches. The findings differ depending on the specification and period under study,¹ and the existing evidence leaves open issues regarding causality and mechanism. In terms of causality, even if the stock market were insignificant above and beyond growth and inflation forecasts in a Taylor rule, it still could be a determinant of policy if it affects macroeconomic expectations included in the rule. As for the mechanism, existing evidence does not speak directly to the economic channels through which the stock market affects Fed expectations and policy.

In this paper, we use textual analysis of FOMC documents to shed light on these questions. Our first objective is to determine whether the Fed finds the stock market informative in its policy making. Frequent Fed discussions of the stock market would suggest that it is an important input to policy making. By contrast, few stock market mentions in FOMC documents would imply that the Fed does not find the market to be informative and that the predictive power of the market for subsequent FFR target changes is coincidental, reflecting an omitted variables bias. Our second objective is to understand the economic mechanism for the Fed finding the stock market informative. Specifically, we seek to distinguish between the driver view of the stock market, whereby the Fed views stock market fluctuations as one of the factors affecting the economy, and the predictor view of the stock market, whereby the Fed attends to the stock market because it is a good predictor of future economic developments (growth, unemployment, etc.). We begin our analysis by reviewing statistical facts about the link between the stock market and policy. We show that since the mid-1990s, the Fed has engaged in a sequence of policy easings following large stock market declines in the intermeeting period. We refer to this pattern as a “Fed put,” by which we mean policy accommodation following poor stock returns. We then tie this fact to the Fed’s macroeconomic expectations, documenting a close comovement between intermeeting stock returns and updates to the Fed’s expectations about the real activity (output growth and unemployment), and a much weaker link to inflation expectations. The comovement emerges in the late 1990s and holds through the 2007/2009 financial crisis and beyond. The relationship is asymmetric; it is stronger for returns over the negative range. Estimating various specifications of the forward-looking Taylor rule, we show that negative stock returns are a significant predictor of the FFR target changes primarily via their correlation with the Fed’s downgrades of growth expectations and the Fed’s assessment of the level of current growth.

To interpret those empirical facts, we turn to the analysis of the minutes and transcripts of the FOMC meetings. To document that the Fed views the stock market as informative, in our baseline approach, we search for phrases related to the stock market (e.g., “stock market,” “equity prices,” “S&P 500”) in the FOMC minutes. We find 975 mentions of the stock market in the 184 FOMC minutes covering the 1994–2016 period. In relative terms, this number represents 14% of inflation mentions and 33% of (un)employment mentions. We classify their tone into positive or negative based on whether FOMC meeting attendees discuss the market going up or down. The tone relates to actual stock returns with the expected signs, with low (high) stock returns leading to more negative (positive) stock market mentions, both before and during the zero-lower bound period. The structure of the Fed documents allows us to measure how frequently actual decision makers at the FOMC (participants, in the minutes’ language) mention the stock market. We find that these mentions predict future policy and do so in an asymmetric way: mentions of stock market declines predict monetary easing, whereas there is no relationship between mentions of stock market gains and tightening. We verify the robustness of these findings using the FOMC transcripts.

In the next step, we use textual analysis to understand the mechanism for why the Fed pays attention to the stock market and, specifically, to assess the relative weight it puts on the driver versus predictor view of the stock market. Our classification of the minutes’ content indicates that 38% of stock market mentions align with the driver view, and 8% with the predictor view; further 11% discuss the stock market determinants, and 38% are purely descriptive summarizing recent stock market moves. For the driver view, the dominant mechanism is via the wealth effect of the stock market on consumption, corresponding to 22% of all stock market mentions. We find a much smaller role for the stock market affecting investment, mainly via the cost of capital channel. However, since consumption demand affects firms’ expected sales, and expected sales is a determinant of investment in the Fed’s model of the U.S. economy, the so-called “FRB/U.S. model” (Brayton et al., 2014), the consumption wealth effect implies an indirect effect of the market on investment.

The FOMC minutes possibly understate the importance of the predictor view of the stock market because of the way they are crafted. Therefore, we extend the analysis of the mechanism to the FOMC transcripts and focus on stock market mentions made by contemporaneous and future Fed chairs: Alan Greenspan, Ben Bernanke, and Janet Yellen. In the chairs’ statements, the driver view is nearly three times as frequent as the predictor view, with 43% versus 15% of mentions across the three chairs. More frequently than suggested by the minutes, the chairs engage in discussions of the determinants of stock market valuations, either trying to learn about economic shocks underlying stock market reactions or deliberating about the potential effects of a stock market correction on the economy.

In terms of the interpretation of the textual analysis results, when the Fed discusses the stock market as a driver of the economy (e.g., via the consumption wealth effect), this channel can be operative whether the Fed perceives stock market fluctuations as independent shocks (shocks to risk aversion or investor sentiment) or as reflecting underlying expectations of economic fundamentals. We find evidence that the risk premium plays a role in the Fed’s decision-making as indicated by Fed officials referencing the market mood, sentiment, and risk aversion. Since risk premiums likely have a substantial exogenous component over and above the expectations of the fundamentals, this finding is consistent with our interpretation of the textual evidence as emphasizing the market as a driver of the economy.

We provide several benchmarks to quantify the strength of the Fed’s reaction to the stock market. To this end, we compare the relationship between the stock market and the Fed’s updating of economic forecasts to that of the corresponding private sector forecasts, as well as to the predictive power of the stock market for realized economic variables (output, unemployment, and inflation). Over the post-1994 period, we find little evidence that the Fed is updating its growth or unemployment expectations too aggressively relative to either of these benchmarks. Together with the Taylor rule estimates, this evidence suggests that the Fed does not react to the stock market over and above what is justified by macroeconomic expectations.

While the Fed may be acting appropriately ex post, an important consideration related to the Fed’s response to the stock market pertains to the ex ante effects of its actions, in particular the moral hazard that they could entail. Analysis of articles published in the Wall Street Journal and the Financial Times indicates that while public awareness of the Fed put goes back to at least 2000, the strength of the put remains a debated issue with each incoming Fed chair.² To understand the policy makers’ own perceptions of the Fed put and the impact of moral hazard considerations on their decision-making, we gather evidence based on textual analysis of financial stability concerns in the Fed documents, reasons for dissents within the FOMC, and direct discussion of the Fed put in the FOMC transcripts. Although the FOMC is clearly aware of the potential moral hazard effects of loose policy, especially post-financial crisis, such concerns do not appear to have a major impact on actual policy choices.

Our work is related to the recent literature exploiting information in textual data to gain insight into the workings of monetary policy. Hansen, McMahon, and Prat (2017) study how central bank transparency influences monetary policy makers’ deliberations, and Hansen and McMahon (2016) analyze the effects of the Fed communication on asset markets and macroeconomic outcomes. Schmeling and Wagner (2017) show that changes in the tone of the ECB communication have a significant effect on asset prices. This work focuses on central bank deliberations and communication, whereas we explore the Fed documents to understand the mechanism through which the stock market affects the central bank decision-making. Cecchetti (2003) counts words related to the stock market and asset prices to argue that the FOMC pays attention to the stock market, but does not classify their tone, either positive or negative, which we show has an asymmetric relation to policy, and does not evaluate why the Fed pays attention to the stock market, which is our main focus. Peek, Rosengren, and Tootell (2016) use text to assess whether the Fed acts as if financial stability was its tertiary mandate. Based on a set of 32 noun phrases, they classify mentions as positive or negative from a financial stability perspective.³ We instead undertake an extensive classification of all stock market mentions in the FOMC minutes and by chairs in the transcripts to assess the relative weight that the Fed puts on the predictor versus driver view of the stock market, and we further distill the specific economic channels underlying the driver view.

1. Data and Variable Definitions

1.1 Defining target changes and intermeeting excess stock returns

The time series of the FFR target going back to September 27, 1982, is available from the FRED Economic Data. Since 1981, the Fed has held eight scheduled meetings per year roughly 6 to 8 weeks apart, and has publicly announced changes to the FFR target since 1994. Prior to 1994, the target series rely on Thornton (2005), who dates target changes based on when the public likely learned about them via the open market operations. This is generally one or two days after the FOMC’s decision. The first FOMC meeting for which we observe the FFR target is October 5, 1982. We define the FOMC cycle day as the number of days elapsed from a scheduled FOMC meeting. Thus, day 0 in FOMC cycle time is the day of a scheduled meeting (the last day for 2-day meetings), day |$-$|1 (+1) is the day before (after) a meeting, and so on.

Daily stock returns and Treasury-bill returns are from Kenneth French’s website. We denote intermeeting excess stock returns as |$rx_m$|⁠. From 1994 onward, we calculate the intermeeting return for the FOMC cycle |$m$| as the excess return of stocks over Treasury bills from day 1 of cycle |$m-1$| to day |$-$|2 of cycle |$m$|⁠, that is, excluding returns earned on day |$-1$| and 0 since these days may be particularly driven by monetary policy news, leading to reverse causality. For the pre-1994 period, we calculate intermeeting returns using returns from day 3 of cycle |$m-1$| to day |$-$|2 of cycle |$m$| to reflect the fact that investors did not know the decision until a day or two after the meeting. For all years, we additionally exclude excess returns earned on days of intermeeting moves because the Fed’s decisions likely influence the stock market on those days.⁴ We identify days of intermeeting moves as those when the FFR target changed outside of scheduled meetings.

To separately study the relation between monetary policy and bad versus good stock market news in the intermeeting period, we define a variable |$rx_m^{-} = \text{min}(0,rx_m)$| to capture movement in excess stock returns over the negative range and |$rx_m^{+} = \text{max}(0,rx_m)$| to capture variation in excess stock return over the positive range.

1.2 Selection of subsamples

In our subsequent analysis, we document differences in the relation between the stock market and Fed policy in the pre- and post-1994 period. The pre-1994 (post-1994) sample starts with the November 1982 (February 1994) meeting and ends with the December 1993 (December 2016) meeting. While it is difficult to point to one break-date event, several facts related to the Fed’s internal modeling and its public communication make 1994 a plausible demarcation line for our analysis. In terms of internal modeling, major modifications to the Fed’s models took place following the 1991 recession. By around 1993, it became clear that the models in use at that time were unable to explain the slow recovery and its relationship with the “financial headwinds” (Reifschneider et al., 1997).⁵ The new FRB/U.S. model became fully operational in mid-1996 (Brayton and Tinsley, 1996), with its key innovation to incorporate expectation formation and intertemporal decision-making of households and firms. In terms of communication, with the first meeting in 1994, the FOMC began making public announcements of their decisions. This moment coincides with a switch from quite frequent to rare intermeeting target moves before and after 1994, implying a change in the Fed’s reaction to events in the intermeeting period. Together, these developments suggest that the mid-1990s was a period of significant changes to the way Fed policy was conducted.

Table 1 presents summary statistics of the intermeeting stock excess returns in the pre- and post-1994 sample. The standard deviation of intermeeting returns is lower in the pre-1994 period (4.77% vs. 5.21%), but an F-test for the difference in standard deviations does not reject the null hypothesis that the volatilities in the two subsamples are equal against the alternative that they are different or against the alternative that volatility in the pre-1994 sample was lower than that in the post-1994 sample (⁠|$p$|-value for the one-sided test is 0.18). If we define quintiles of intermeeting returns based on all years, 18% of intermeeting returns fall in the lowest quintile in the pre-1994 period, compared to 21% in the post-1994 period. This suggests sufficient return distribution within each subsample to reveal whether the Fed accommodates low stock returns in a given subsample.

2. Stock Returns as a Predictor of Fed Growth Expectation Updates and Monetary Policy

This section reviews the statistical facts connecting intermeeting stock market excess returns, updates to the Fed’s macroeconomics expectations, and subsequent FFR target changes. The evidence supports a strong comovement between the stock market, the Fed’s expectations about the real economy, and its monetary policy decisions that emerges in the second half of the 1990s. Subsequently, we use textual analysis to argue that the statistical facts reflect the Fed viewing the stock market as informative for policy making.

2.1 Low stock returns predict policy accommodation

Using data for 1994–2008, in Figure 1, panel A, we plot the average cumulative change in the FFR target from meeting |$m-1$| to meeting |$m+X$| (for different values of |$X$|⁠) against average intermeeting excess stock returns, with both averages calculated by quintile of the intermeeting excess stock return. The quintiles of intermeeting returns are constructed over the 1994–2008 sample.⁶ Returns in the lowest quintile (averaging around |$-$|8%) are associated with an average reduction in the target of as much as 119 bps over the eight subsequent FOMC cycles from |$m-1$| to |$m+7$| (significant at the 1% level). No such pattern of Fed accommodation following low stock returns is seen before 1994 (Figure 1, panel B).

Figure 1

Changes in the FFR target conditional on the intermeeting stock excess returns

The figure plots the average cumulative FFR target change from day 0 of cycle |$m-1$| to day 0 of cycle |$m+7$| (approximately a 1-year period) by quintiles of the intermeeting excess return, |$rx_m$|⁠. The quintiles of intermeeting returns are constructed over the 1994–2008 sample.

Open in new tabDownload slide

Table 2 reports predictive regressions using |$rx_m^{-}$| and |$rx_m^{+}$| as the explanatory variables for the FFR target changes. The significantly positive coefficients on |$rx^{-}_m$| in the post-1994 sample indicate that more negative intermeeting returns forecast a stronger Fed accommodation. A 10% stock market decline predicts a reduction in the FFR target of 32 bps at the next meeting and 127 bps after 1 year. The results also point to an asymmetry, with positive intermeeting returns in most cases being insignificant. The right panel shows that the relationship is absent in the pre-1994 sample.⁷

The analysis of the FFR target over the 1994–2008 period is not informative as to whether the stock market has predictive power for monetary policy in the post-2008 period, during most of which the target was at the zero-lower bound (ZLB) and Fed policy focused on forward guidance and quantitative easing (QE). To speak to the ZLB period, we revisit the evidence in Cieslak, Morse, and Vissing-Jorgensen (2019, CMVJ) who study the effect of Fed policy on stock returns. CMVJ argue based on a series of facts that monetary policy news disproportionately arrives in “even weeks” in FOMC cycle time. They show that, over the 1994–2016 period, stock returns mean-revert on even-week days that follow particularly bad realizations of stock returns. They refer to this evidence as a “Fed put in stock returns” and interpret it as the Fed being able to reduce the risk premium in stocks via a promise to act as needed. This promise could be about both QE and the path for the policy rate. We would expect the Fed-induced even-week mean-reversion in stock returns to remain present during the ZLB if the Fed still accommodated following poor stock returns in that period (and accommodated more than expected by markets). Figure 2, panel A, illustrates the Fed put in stock returns over the 1994–2016 sample, plotting average excess stock returns on day |$t$| against prior 5-day excess stock returns: returns are high on even-week days that follow prior poor stock returns. Figure 2, panels B and C, split the 1994–2016 period into years 1994–2008 and 2009–2016. In both subperiods, stock returns are high (around 30 bps, statistically significant) on even-week days preceded by stock returns in the lowest quintile. The estimates for 2009–2016 suggest that the Fed put in stock returns is still present during the ZLB period. Extending CMVJ’s evidence, Figure 2, panel D, shows that the effect is absent pre-1994, coinciding with our finding in Figure 1 of no relationship between the stock market and target changes in that period.⁸

Figure 2

The Fed put in stock returns

The figure reports the average daily excess stock returns by even- and odd-numbered weeks of the FOMC cycle, conditional on returns realized over the previous week (by return quintiles, 1 = lowest; 5 = highest). Even and odd weeks are defined following CMVJ (2019). Return quintiles are defined over the 1994–2016 sample in panels A through C and over the 1982:11–1993 sample in panel D. Each plot contains 95% confidence intervals for the means. For readability of the graph, we truncate confidence intervals if they span outside of the |$-$|0.2% to 0.4% range.

Open in new tabDownload slide

A useful feature of the textual analysis in the following sections is that FOMC minutes are available up to the end of our sample, which further allows us to study whether and why the Fed paid attention to the stock market, both in the 1994–2008 period and in the 2009–2016 period.

2.2 Updates to the Fed’s growth expectations comove asymmetrically with stock returns

As a first step toward understanding the relation between the stock market and policy, we document how updates to the Fed’s macroeconomic expectations comove with the stock market. A few days before each scheduled FOMC meeting, the staff at the Federal Reserve Board prepares macroeconomic forecasts for several quarters ahead, collected in the so-called Greenbook (now called the Tealbook). Greenbooks become publicly available with a 5-year lag. Our forecast data set, available from the Philadelphia Fed’s website, ends in December 2012.⁹

To see how macroeconomic expectations evolve from one FOMC meeting to the next, we compute updates relative to expectations in the previous Greenbook for the same calendar quarter; that is, for variable |$Z$|⁠, the update is defined as |$\text{Updt}_m(Z_{qi}) = E_m(Z_{qi})-E_{m-1}(Z_{qi}),$| where |$qi$| is a calendar quarter (⁠|$q0$| is the current quarter, |$q1$| is the next quarter, etc.);¹⁰|$E_m(\cdot)$| denotes the Greenbook forecast at meeting |$m$|⁠. We regress the updates on past intermeeting returns. To study potential asymmetries, we allow for separate coefficients on positive and negative stock return realizations. We generally find that two return lags (for |$m$| and |$m-1$|⁠) are significant. The lagged return for |$m-1$| accounts for gradual expectations updating.

Table 3 presents the regression evidence for updates of real gross domestic product (GDP) growth, unemployment, and inflation expectations. We summarize the results by reporting the sum of coefficients on the intermeeting returns. The top panels of Table 3 document that, in the post-1994 sample, the Fed’s growth and unemployment expectations updates comove asymmetrically with the stock market, loading significantly on intermeeting returns over the negative range, with a smaller and less significant relation to positive returns. A 10% lower intermeeting return is associated with a reduction of the total expected growth rate over the next four quarters of slightly below 1 percentage point for real GDP growth (column 1),¹¹ and with an increase of 0.47 percentage points in the expected unemployment rate three quarters out (column 3). Before 1994, we observe a weak relationship, and, in particular, there is little downgrading of expectations following poor stock returns (columns 2 and 4). The bottom panel of Table 3 refers to the updating of the Fed’s inflation expectations. For two of the three inflation measures, more negative stock returns predict a lowering of inflation expectations.¹²

Figure 3, panel A, displays the time series of updates to real GDP growth expectations along with the fitted regression values, corresponding to spliced estimates from columns 1 and 2 of the top panel in Table 3. Before 1994, forecast updates display a weak relationship with stock returns. In contrast, post-1994, the stock market explains a large portion of the variation in growth updates (with a regression |$R^2$| of 38%), and the explanatory power comes primarily from negative return realizations. The tight relationship between stock returns and growth updates emerges in the second half of the 1990s and holds through the end of the Greenbook sample in 2012.

Figure 3

Greenbook real GDP growth expectations updates and intermeeting stock excess returns

Panel A plots fitted values from regressions of Greenbook real GDP growth expectations updates over four quarters on the current and lagged intermeeting stock excess returns over the positive and negative domain, corresponding to regressions in columns 1 and 2 of the top panel of Table 3. The regressions are estimated separately on the 1982:11–1993 and 1994–2012 sample (the vertical line in the graph represents the sample split date). Each regression also includes a constant and the lagged value of the dependent variable. Forecast updates are expressed as a percentage per annum. Panel B graphs average growth forecast updates over two meetings (⁠|$m$| and |$m+1$|⁠) by quintile of intermeeting stock excess returns |$rx_m$|⁠. The spikes represent the 95% confidence bands for the means. Quintiles of intermeeting returns are computed over the 1982:9–2012 sample (end of the Greenbook data).

Open in new tabDownload slide

Additionally, Figure 3, panel B, illustrates the magnitude of the asymmetry. The panel shows average growth expectations updates over two meetings (⁠|$m$| and |$m+1$|⁠) conditional on the quintiles of the intermeeting excess stock returns (⁠|$rx_m$|⁠), pre- and post-1994. We sum updates over two meetings to account for the lagged effect of stock returns, consistent with the specification in Table 3. The plot confirms that Greenbook growth forecasts decline significantly following stock market realizations in the lowest quintile in the post-1994 sample, and there is no similar relationship in the earlier period.

To put the above results into perspective, we also verify that the predictive power of intermeeting stock returns for the Fed’s growth expectations updates is stronger than the predictive power of any of the 38 macroeconomic indicators available in Bloomberg’s economic calendar or the Chicago Fed National Activity Index (CFNAI), a principal component of 85 macro series (for the details of this analysis, see Internet Appendix A). This fact could suggest that, compared to macro variables, the stock market provides policy makers with more up-to-date and forward-looking information about the economic outlook. We study this hypothesis in Section 4, in which we assess the weight policy makers place on the predictor view of the stock market.¹³

In sum, the strong relation between negative stock returns and the Fed’s expectations of real variables emerges from the mid-1990s, with a less clear pattern for inflation. These results complement recent evidence that shocks to financial conditions and realized economic growth are linked in the left tail of the distribution (Adrian et al., 2019, Berger et al., 2020, 2020). By themselves, however, the regressions above are not evidence that the Fed finds the stock market informative since the stock market and growth updates may be driven by common factors. The textual analysis in Section 4 establishes that the Fed does in fact view the stock market as informative for forming growth expectations and deciding on policy.

2.3 Fed policy reacts to growth expectations downgrades

Table 4 estimates different versions of Equation (1). Column 1 provides a statistical benchmark to document the stock market significance on a stand-alone basis. Intermeeting stock returns and lags of the FFR explain 54% of FFR target changes in column 1 (FFR lags alone explain 36%). The stock market significance is driven by the negative return realizations. The coefficients sum to 5.35 for |$rx_m^{-}$| and |$rx_{m-1}^{-}$| (⁠|$t$|-stat = 4.71 for the null that the sum of the two coefficients is zero) versus |$-1.80$| for |$rx_m^{+}$| and |$rx_{m-1}^{+}$| (⁠|$t$|-stat = |$-1.05$|⁠). Column 2 uses Greenbook variables to explain FFR changes. Splitting growth expectations updates into positive and negative shows that only negative updates are significant, |$\text{Updt}_m(g_{q1})^{-}=\text{min}(0,\text{Updt}_m(g_{q1}))$|⁠. Column 3 estimates a combined specification with Greenbook variables and intermeeting stock returns. The sum of coefficients on negative stock returns now drops from 5.35 in column 1 to 1.66 in column 3 (a 69% reduction) and becomes insignificant.¹⁶ The coefficients on negative growth update loses part of its statistical significance consistent with its correlation with intermeeting returns documented in Table 3.¹⁷

Columns 4–6 provide robustness to these results, following the branch of the literature that omits FFR lags in the Taylor rule (e.g., Orphanides 2019; Hartmann and Smets 2018). Without controlling for the lagged FFR, the negative intermeeting returns are significant up to five lags (we do not report lags that are insignificant). Including the Greenbook variables reduces the sum of coefficients on the intermeeting returns by 75%. To better understand these relations, columns 7 and 8 study the explanatory power of negative intermeeting returns for the real GDP nowcast, negative growth update, and inflation forecast. Lags of |$rx_m^-$| strongly predict the growth nowcast (⁠|$R^2 = 0.52$|⁠) and the growth update (⁠|$R^2=0.44$|⁠). For inflation, however, none of the return lags is significant (⁠|$R^2=0.04$|⁠).

In sum, across specifications, most of the explanatory power of the stock market for target changes can be rationalized by its correlation with the Greenbook growth update and growth nowcast. The Taylor rule results also point to an asymmetry in the policy reaction function in the post-1994 sample: accommodation in response to downgrades in growth expectations without symmetric tightening in response to upgrades. The latter fact is consistent with stable and low inflation over this period (see e.g., Gorodnichenko and Shapiro, 2007).

Interestingly, recent events related to the COVID-19 shock fit into our characterization of the stock market’s role in Fed’s policy making. The stock market turmoil that preceded the policy rate cuts in March 2020 is an instance of a lowest-quintile stock market decline occurring in the intermeeting period. Our analysis shows such declines to be predictive of subsequent policy rate cuts. The stock market declined before the economic downturn could be detected in macroeconomic data. The FOMC minutes from March 15, 2020, include a discussion of plummeting equity markets and explain that the staff projection for the U.S. economy were “downgraded significantly [...] in response to news on the spread of the coronavirus at home and abroad [...] along with recent financial market movements.” We turn to a systematic analysis of the content of the Fed documents next.

3. Does the Fed Pay Attention to the Stock Market?

The high explanatory power of the stock market for the Fed’s growth expectations and FFR target changes has two possible interpretations. The relation could be coincidental in the sense that the Fed views the stock market as uninformative but the econometrician finds it has explanatory power for target changes because the stock market is correlated with variables that drive the Fed’s decision-making. Alternatively, the Fed may perceive stock returns as informative and therefore react to them. This could be due to stock returns being viewed as a driver of the economy, or due to them being viewed as a useful predictor of economic variables the Fed cares about, notably growth.

We first seek to distinguish the coincidental interpretation from the informative one. To establish that the Fed does pay attention to the stock market directly, we perform a textual analysis of the minutes and transcripts of FOMC meetings. In the next section, we then turn to using textual analysis to understand the mechanism for why the Fed may view the stock market as informative.

3.1 Textual data: Minutes and transcripts of FOMC meetings

We collect texts of the minutes and transcripts of FOMC meetings from the Federal Reserve Board’s website. The longest sample we consider ranges from 1976 through 2016. FOMC meetings are highly structured events that always include

1. Staff Review of the Economic Situation;

2. Staff Review of the Financial Situation;

3. Staff Economic Outlook;

4. Participants’ Views on Current Conditions and the Economic Outlook;

5. Committee Policy Action.

We refer to parts 1–3 as representing the views of the staff and parts 4 and 5 as representing the views of the participants (the chair, vice-chair, other governors, and regional Reserve Bank presidents). The FOMC minutes are carefully crafted to “record all decisions taken by the Committee with respect to these policy issues and explain the reasoning behind these decisions,” as stated on the Federal Reserve Board’s website. The sections of the minutes corresponding to the above five parts of the FOMC meeting are typically 7–10 pages long. Since 2005, minutes have been published three weeks after the FOMC meeting. Before 2005, they were published three days after the next FOMC meeting. Minutes are available up to the end of our sample in 2016.¹⁸ The FOMC transcripts contain verbatim comments by individual staff members and meeting participants, with each transcript 200–300 pages long. They are released publicly with a 5-year lag, our sample ending in 2013. Our baseline results rely on human reading and hand coding of the minutes content, and we use both algorithm-based coding and human reading of the transcripts for robustness.

Figure 4 displays simple counts of stock-market-related phrases in the minutes (panel A) and in the transcripts (panel B) starting in 1976. Table 5, panel A, lists search phrases. The stock market is rarely mentioned before the mid-1990s, with the exception of a spike in October 1987 following the 1987 market crash. From the mid-1990s, the number of mentions increases and remains elevated through the end of the sample. Given the change in attention paid to the stock market in the mid-1990s, our subsequent textual analysis mainly focuses on the post-1994 sample. This is also the period during which our results thus far suggest that policy may find the stock market informative.

Figure 4

Counts of stock market mentions in FOMC documents

Panel A reports combined counts of stock market mentions in Records of Policy Actions and Minutes of Actions for the 1976–1992 sample and in FOMC minutes for the 1993–2016 sample. Panel B reports counts in the transcripts of FOMC meetings (solid-black line) and those combined with counts in the transcripts of FOMC conference calls (solid-gray lines). Counts in transcripts of conference calls in the intermeeting period are added to the counts in the transcripts of the next FOMC meeting. Vertical, thick dashed lines in both panels represent the ends of the tenures of Fed Chairs Miller, Burns, Volcker, Greenspan, and Bernanke.

Open in new tabDownload slide

3.2 Stock market mentions in FOMC minutes

We extract all paragraphs in the FOMC minutes between 1994 and 2016 that are related to the stock market. There are 975 stock market mentions. Table 5, panel A, reports the number of times each search phrase appears, and panel B reports the counts of stock market phrases by section of the minutes. The category “Other” refers to parts outside the regular minutes’ sections (e.g., summary of special sessions or conference calls). About half of the mentions are in the Staff Review of the Financial Situation section, with the other half split between the staff and the participants.

We read and classify the tone of stock market mentions based on the direction of the market’s evolution: positive (discussion of the stock market going up), negative (discussion of the stock market going down), neutral (stock market flat), and hypothetical (discussion of what would happen if the stock market were to move in a particular way). Table 5, panel C, displays counts by tone. Mentions where tone is unclear or cannot be determined are marked as “unclear.” Consistent with the stock market on average having increased over the sample, there are more positive than negative stock market mentions. Figure 5 displays the respective frequencies over time. Mentions in the Staff Review of the Financial Situation are frequently purely descriptive (more on our classification of the content below). Therefore, Figure 5 presents counts both including and excluding counts in that section. Peaks in the number of negative mentions in panel A often correspond to periods of market stress, as marked on the graph. The positive mentions in panel B are relatively more uniformly distributed over time.

Figure 5

Time series of positive and negative stock market phrases in FOMC minutes

The figure presents the time series of negative and positive stock market phrases in FOMC minutes based on human coding. We present the overall counts as well as excluding the counts in the Staff Review of the Financial Situation (minutes’ section 2). The sample period is 1994–2016. The triangles in panel A represent FOMC meetings that were preceded by intermeeting stock market returns in the lowest quintile.

Open in new tabDownload slide

3.2.1 Predicting the Fed’s attention to the stock market with past stock returns

To relate stock market mentions to stock returns, Figure 6 plots the average number of mentions against average intermeeting excess stock returns (with and without the Staff Review of the Financial Situation section), with averages calculated by quintiles of |$rx_m$| defined over the 1994–2016 sample. Lower intermeeting returns in the left panel predict more negative stock market mentions, especially in the lowest quintile of return realizations. Similarly, in the right panel, higher stock returns lead to more positive stock market mentions. There are relatively more negative mentions when the stock market performed very poorly than there are positive mentions when it did well, which is particularly visible when we exclude the Staff Review of the Financial Situation.

Figure 6

Intermeeting stock returns and negative and positive stock market mentions in FOMC minutes

The figure presents the average count of positive and negative stock market phrases conditional on the quintiles of intermeeting stock excess returns (x-axis labels report the average intermeeting return within a given quintile). Averages are for overall counts and for counts excluding the Staff Review of the Financial Situation section. The sample period is 1994–2016. The results are based on human coding of the minutes content. The spikes represent 95% confidence bands for the means. Internet Appendix Figure 2 provides a scatter plot of the data for the individual meetings.

Open in new tabDownload slide

Past intermeeting returns have strong forecasting power for the frequency of stock market mentions in the minutes with an |$R^2$| around .50 (Internet Appendix Table 1 provides the details). The relation has the expected signs. More negative (positive) past intermeeting returns predict more (fewer) negative stock market mentions. A 10% more negative return is associated with 6.4 more negative stock market mentions, a substantial effect relative to the mean (1.8 mentions) and standard deviation (2.6 mentions). The fact is present both before and during the ZLB period.

3.2.2 Predicting target changes by the Fed’s stock market mentions

Does the degree of the Fed’s attention to the stock market predict actual policy actions? Table 6 shows that negative stock market mentions in the minutes of the current and past FOMC meeting have statistically significant explanatory power for FFR target changes.¹⁹ The estimates in column 1 imply that a one-standard-deviation increase in the number of negative mentions (2.6 more mentions) is associated with a cumulative reduction in the FFR target of 32 bps. In contrast, there is no systematic relation between positive mentions and subsequent target changes.²⁰

If all explanatory power came from staff discussions, this may speak against the policy makers viewing the stock market as informative. In columns 2 through 4, we therefore estimate the regressions separately for the Staff Review of the Financial Situation (minutes’ section 2), other sections by the staff (sections 1 and 3), and participants (sections 4 and 5).

Importantly, negative stock market mentions predict target changes if we focus only the FOMC participants. While many mentions by the staff simply summarize recent developments in financial markets, Section 4 shows that participants most often describe a causal effect of the stock market on the economy.

To assess whether the minutes-based results accurately reflect the nature of Fed deliberations, we extend the analysis to the FOMC transcripts. We develop an algorithm to identify negative and positive stock market mentions in the transcripts (see Internet Appendix B for details). The results, reported in column 5 of Table 6, confirm that while there is no relationship between participants’ positive stock market mentions and target changes, their negative stock market predict target reductions.

To conclude, the Fed pays direct attention to the stock market rather than to just variables correlated with the stock market. Positive and negative stock market mentions move with intermeeting excess stock returns in the expected directions and the frequency of negative stock market mentions predicts FFR target reductions. These facts suggest that the Fed views the stock market as informative for policy making.

3.3 Stock market and broader financial conditions

The analysis so far may understate the FOMC’s concern with the stock market. The FOMC minutes often talk about “financial conditions” without explicitly referring to the stock market, and thus such cases are not accounted for in our analysis above. As described by then-President of the New York Fed William Dudley “(...) financial conditions can be broadly summarized by five key measures: short- and long-term Treasury rates, credit spreads, the foreign exchange value of the dollar, and equity prices” (Dudley, 2017).²¹

To assess the frequency of such references, we use an algorithm to code the number of negative and positive mentions of financial conditions that do not explicitly reference the stock market in the minutes (identifying 350 negative and 232 positive cases). Mentions of negative financial conditions spike during the financial crisis in 2008 and 2009, and their frequency is predictable by intermeeting stock returns. While negative financial conditions mentions predict FFR target changes over and above the stock market mentions, this result is driven by the financial crisis. Once 2008 is excluded, stock market mentions subsume the explanatory power of financial conditions for the target. Internet Appendix C reports detailed results.

To place the Fed’s attention to the stock market in perspective, Figure 7 graphs the frequency of mentions of interest rates, credit and spreads, and exchange rates, along with our series for stock market mentions. The focus on the stock market emerges in the second half of the 1990s, whereas mentions of interest rates, credit and spreads, and exchange rates are prevalent going back to the late 1970s.

Figure 7

Mentions of specific financial conditions in FOMC minutes

The figure displays counts of mentions of different variables determining financial conditions. The counts are obtained from FOMC minutes. Dashed vertical lines represent the end of tenures of subsequent Fed chairs. See Internet Appendix C for word lists for each of the concepts graphed.

Open in new tabDownload slide

4. Why Is the Stock Market Viewed as Informative by Fed Policy Makers?

To shed light on why the Fed views the stock market as informative, we analyze the content of the stock market mentions. We start with the FOMC minutes, and then extend the analysis to the transcripts studying the stock market mentions by the contemporaneous and future Fed chairs (Greenspan, Bernanke, and Yellen). Our goal is to distinguish between different mechanisms that could lead the Fed to pay attention to the stock market, that is, whether the Fed thinks of the stock market as a driver of the economy or as a predictor of future economic conditions. If the first possibility dominates, we would like to understand the economic channels through which the Fed believes the stock market affects the economy.

4.1 The content of stock market mentions in the FOMC minutes

We read and classify the 975 stock market mentions in the minutes into four main categories: (1) consistent with the driver view (describing the stock market as having a causal effect on the economy), (2) consistent with the predictor view (describing how the stock market reflects the economic outlook or investors’ expectations), (3) describing other determinants of stock valuation (factors affecting the stock market, other than the economic outlook, including the effect of monetary policy on the market) and discussing valuation levels, and (4) descriptive (simply summarizing recent stock market movements). In the driver-view category, we further distinguish mentions of the effect of the stock market on consumption, investment, demand (when there is no detail about which component of demand), and other mentions indicating that the market affects the economy but for which no precise mechanism is stated. We take an expansive approach to the predictor view to give this channel the best chance. Thus, we include not only cases in which the stock market is explicitly stated as an indicator of the outlook or predictor of economic conditions but also cases discussing how the stock market reflects investors’ expectations of economic conditions. We treat those separately from discussions of other stock market determinants, especially from how the stock market responds to current (as opposed to expected) economic data.²² A few mentions of the stock market in the context of financial stability are made, and we report them separately. Mentions that we cannot classify are collected in an “Other” category. Internet Appendix D contains sample excerpts from the minutes illustrating these categories.

Table 7 reports the classification results: 38% of mentions align with the driver view, and 8% with the predictor view, another 11% discuss stock market determinants, and 38% are purely descriptive. When the FOMC participants discuss the stock market, the driver view prevails, accounting for 80% of participants’ mentions. Overall, the majority of mentions in the driver-view category comes from the participants’ sections of the minutes. Instead, among the 8% of cases consistent with the predictor view and the 11% of cases discussing valuation determinants most can be attributed to the staff.