A central use of reserves held at Federal Reserve Banks (FRBs) is for the settlement of interbank obligations. These obligations are substantial—the average daily total reserves used on two main settlement systems, Fedwire Funds and Fedwire Securities, exceeds $6.5 trillion. The total amount of reserves needed to efficiently settle these obligations is an active area of debate, especially as the Federal Reserve’s current quantitative tightening (QT) policy seeks to drain reserves from the financial system. To better understand the use of reserves, in this post we examine the intraday flows of reserves over Fedwire Funds and Fedwire Securities and show that the mechanics of each settlement system result in starkly different intraday demands on reserves and differing sensitivities of those intraday demands to the total amount of reserves in the financial system.  

Background

As part of the normal course of business, banks settle obligations amongst themselves by drawing upon reserves they hold at the FRBs. For cash-only obligations, banks most often use the Fedwire® Funds Service, a large-value real-time payments system. The total daily value of transfers made over Fedwire Funds is substantial, with the average daily value sent in the second quarter of 2024 equal to $4.5 trillion.    

Another settlement system operated by the FRBs that uses reserves is the Fedwire® Securities Service, a real-time delivery-versus-payment system. This system is linked to the FRBs’ book-entry ledger, on which the U.S. Treasury, Fannie Mae, Freddie Mac, and other agencies issue securities. Among other services, Fedwire Securities enables the simultaneous movement of securities against reserves across accounts. There is also a substantial daily movement of reserves over Fedwire Securities, with the average daily value of reserves delivered in the second quarter of 2024 equal to $2.1 trillion.    

For both settlement services, banks usually initiate transactions on behalf of their customers. Those customers, or the underlying nature of the transaction, can dictate that the transactions be completed by certain deadlines within the day. The intraday flow of reserves in and out of a bank’s account, however, can be quite large relative to a bank’s balance. A previous Liberty Street Economics post showed that in 2015, a year of abundant total reserves, the largest ten banks still faced outflows over Fedwire Funds (alone) that were larger than their beginning-of-period balance at the FRBs. Banks then, often need to strategically manage the timing of transactions over both settlement systems, balancing the demands of their customers for earlier settlement within the day against the bank’s balance of reserves at the FRBs (and the bank’s appetite to access intraday credit from the FRBs).

What Is Each System’s Intraday Liquidity Demands?

Fedwire Funds is designed to accommodate only credit payments, so the account holder sending reserves initiates the transfer. Accordingly, when an account holder’s balance is running low, that bank often strategically delays sending payments over Fedwire Funds until later that same day, with the expectation that incoming payments will replenish that bank’s balance (see “Intraday Liquidity Management: A Tale of Games Banks Play” and “How Abundant Are Reserves? Evidence from the Wholesale Payment System”). Hence, when a bank is facing a constraint on the reserves it is holding within the day, the bank is likely to respond by delaying payments made over Fedwire Funds until later in the day.

Fedwire Securities is designed so that the account holder sending securities (and receiving reserves) initiates the transaction. Given that account holders value holding reserves for intraday liquidity needs, even on the margin, the account holders that face obligations to deliver securities for a given day will often deliver those securities as early as possible.

These intraday demands on reserves drive the distribution of transactions within the day across each settlement service. The chart below shows the percent of total transactions by value that occur within 1.5-hour buckets over the Fedwire Funds operational day (9 pm of the day before to 7 pm of the current day, eastern time) for the second quarter for 2024. There is an uptick in payments made with the opening of business on the U.S. East Coast, but the busiest interval is from 3-4:30 pm. This late-day bunching of payments is consistent with banks managing intraday liquidity demands, an observation made in “Changes in the Timing Distribution of Fedwire Funds Transfers.”

Banks Settle More than Half of Fedwire Funds Transactions in the Last Third of the Day

Turning to Fedwire Securities, the following chart shows the percent of total transactions by value that occur over this settlement system’s operational day (8:30 am to 3:15 pm, Eastern Time) in the second quarter of 2024. The main takeaway is that more than 60 percent of total transactions by value occur right at the opening of Fedwire Securities. As with Fedwire Funds, this behavior is consistent with banks valuing reserves for intraday liquidity. Furthermore, the concentration of settlement at the opening creates substantial demands on reserves—in the first 30 minutes after opening, about $1.05 trillion of reserves, on average, are transferred over Fedwire Securities. 

Most Fedwire Securities Payments Occur in the First 30 minutes after Opening

How Does the Intraday Timing of Payments Change with the Level of Aggregate Reserves?

When the Federal Reserve changes the total amount of reserves held by banks, it directly affects the intraday demand for reserves by banks. In particular, when the Fed increases the total amount of reserves, banks’ concern over the intraday flows of reserves diminishes, as discussed previously. This dynamic is seen in the chart below that plots aggregate reserves alongside the time of day when half of the total value of Fedwire Funds payments have been sent (time of median payment). When there are low levels of aggregate reserves, banks delay payments by a substantial amount. Indeed, from 2015 to 2019, the steady decline in aggregate reserves from more than $2 trillion to less than $2 trillion is accompanied by an almost one-hour shift later in the time of the median payment. Furthermore, the substantial increase in aggregate reserves from 2019 to 2022 coincides with a large shift in payments being settled earlier in the day.

This association between intraday settlement timing and aggregate reserves is not seen with Fedwire Securities. Rather, the timing of these transactions remains bunched at opening from 2015 to now. Starting in the first half of 2018 and ending in the first half of 2019, there is a 30 minute earlier-in-the-day shift in the time of median payment, but this shift is due to changes in the custodial business and is unrelated to the level of total reserves.

The Timing of Fedwire Funds Payments Reacts to the Total Level of Reserves …

… Whereas the Timing of Fedwire Securities Payments Does Not.

Takeaway

The timing of payments over Fedwire Funds has garnered attention because shifts in the timing of payments are informative about banks’ demand for reserves and, consequently, the overall level of total reserves in the system. Banks also need substantial amounts of reserves to settle their obligations on Fedwire Securities, especially right at opening. Unlike what is seen in Fedwire Funds however, the intraday timing of settlements on Fedwire Securities is invariant to changes in the total amount of reserves in the system.

Adam Copeland is a financial research advisor in Money and Payments Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.  

Sarah Yu Wang is a research analyst in the Federal Reserve Bank of New York’s Research and Statistics Group.

How to cite this post:
Adam Copeland and Sarah Yu Wang, “The Dueling Intraday Demands on Reserves,” Federal Reserve Bank of New York Liberty Street Economics, October 21, 2024, https://libertystreeteconomics.newyorkfed.org/2024/10/the-dueling-intraday-demands-on-reserves/.

You may also be interested in:

Tracking Reserve Ampleness in Real Time Using Reserve Demand Elasticity

With Abundant Reserves, Do Banks Adjust Reserve Balances to Accommodate Payment Flows?

Intraday Liquidity Management: A Tale of Games Banks Play

How Abundant Are Reserves? Evidence from the Wholesale Payment System

As central banks shrink their balance sheets to restore price stability and phase out expansionary programs, gauging the ampleness of reserves has become a central topic to policymakers and academics alike. The reason is that the ampleness of reserves informs when to slow and then stop quantitative tightening (QT). The Federal Reserve, for example, implements monetary policy in a regime of ample reserves, whereby the quantity of reserves in the banking system needs to be large enough such that everyday changes in reserves do not cause large variations in short-term rates. The goal is therefore to implement QT while ensuring that reserves remain sufficiently ample. In this post, we review how to gauge the ampleness of reserves using the new Reserve Demand Elasticity (RDE) measure, which will be published monthly on the public website of the Federal Reserve Bank of New York as a standalone product.

In a Liberty Street Economics post in 2022, we proposed focusing on the slope of the reserve demand curve to assess the ampleness of reserves in the banking system. The reserve demand curve describes the price at which banks are willing to trade their reserve balances with one another as a function of aggregate reserves. In the U.S., this price is called the federal funds rate, which is the rate targeted by the Federal Open Market Committee (FOMC) in its communication of the monetary policy stance. The slope of the reserve demand curve measures the elasticity of the federal funds rate to changes in the level of reserves; that is, by how much this rate changes in response to variation in aggregate reserves. We call this quantity the Reserve Demand Elasticity.

Economic theory tells us that the RDE (that is, the slope of the reserve demand curve) is different at different reserve levels. We can then operationalize the notion of ample reserves in the Federal Reserve monetary policy framework as the region of the reserve demand curve where the slope is only modestly negative: for this range of reserve levels, the RDE is small. At higher reserve levels, where reserves in the banking system are abundant, the RDE is zero (the demand curve is flat); at lower levels, where reserves are scarce, the RDE is negative and large (the curve is steeply sloped). To operate in an ample reserves framework and avoid reserve scarcity, it is therefore important to identify the transition point between abundant and ample reserves.

Identifying the transition between these regions is challenging because banks’ demand for reserves fluctuates over time and, in turn, the supply of reserves may respond to sudden changes in banks’ demand. As explained in our paper, we developed an econometric methodology that addresses these challenges. The idea is simple: exploiting large fluctuations in aggregate reserves over the last decade, we move along the reserve demand curve and, for each day, estimate its slope (the RDE).

In the first post of a two-part Liberty Street Economics series published this summer, we propose building on our methodology to monitor reserve ampleness in real time. We estimate the RDE daily using only information available as of each day, making our estimates equivalent to real-time calculations. We can then construct a signal of ampleness of reserves by looking at when the real-time estimates of the RDE become negative at a given confidence level.

The chart below, taken as a screenshot from the new interactive RDE webpage, shows our real-time estimate of the RDE from January 2010 through October 9, 2024. The RDE was significantly negative in 2010-11 but then became indistinguishable from zero throughout 2012-17 as a result of the large amounts of reserves injected into the banking system by the Federal Reserve in response to the Global Financial Crisis. During the Federal Reserve QT of 2018-19, the RDE returned to be significantly negative at the 68 percent confidence level in August 2018 and at the 95 percent confidence level in March 2019—twelve to six months in advance of the money market turmoil of September 2019. After that episode, the RDE started to move back towards zero, as the Federal Reserve injected liquidity in the banking system to maintain control of short-term rates; since mid-2020, as the Federal Reserve used quantitative easing (QE) in response to the COVID-19 crisis, the RDE has been indistinguishable from zero. Our most recent estimates suggest that, although reserves have declined by over $1 trillion since their peak of $4.2 trillion in November 2021, they remain abundant.

Importantly, today we are launching Reserve Demand Elasticity as a standalone product, with new readings to be published each month on the New York Fed’s public website. The Reserve Demand Elasticity webpage offers the historical and latest estimates of the RDE, as well as information on the underlying data, methodology, and interpretation. The goal of this product is to equip the public with a novel tool that can help in monitoring the ampleness of reserves in the U.S. banking system and stimulate further analysis in this important area.

Gara Afonso is the head of Banking Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.

Domenico Giannone is an assistant director at the International Monetary Fund and an affiliate professor at the University of Washington.

Gabriele La Spada is a financial research advisor in Money and Payments Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.   

John C. Williams is the president and chief executive officer of the Federal Reserve Bank of New York.  

How to cite this post:
Gara Afonso, Domenico Giannone, Gabriele La Spada, and John C. Williams, “Tracking Reserve Ampleness in Real Time Using Reserve Demand Elasticity,” Federal Reserve Bank of New York Liberty Street Economics, October 17, 2024, https://libertystreeteconomics.newyorkfed.org/2024/10/tracking-reserve-ampleness-in-real-time-using-reserve-demand-elasticity/.

You may also be interested in:

Reserve Demand Elasticity

When Are Central Bank Reserves Ample?  

Scarce, Abundant, or Ample? A Time-Varying Model of the Reserve Demand Curve

---

Disclaimer
The views expressed in this post are those of the author(s) and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System, or the International Monetary Fund (IMF), its executive board, or IMF management. Any errors or omissions are the responsibility of the author(s).

While policies to combat climate change are designed to address a global problem, they are generally implemented at the national level. Nevertheless, the impact of domestic climate policies may spill over internationally given countries’ economic and financial interdependence. For example, a carbon tax charged to domestic firms for their use of fossil fuels may lead the firms to charge higher prices to their domestic and foreign customers; given the importance of global value chains in modern economies, the impact of that carbon tax may propagate across multiple layers of cross-border production linkages. In this post, we quantify the spillover effects of climate policies on forward-looking asset prices globally by estimating the impact of carbon price shocks in the European Union’s Emissions Trading System (EU ETS) on stock prices across a broad set of country-industry pairs. In other words, we measure how asset markets evaluate the impact of changes to the carbon price on growth and profitability prospects of the firms.

The EU ETS and Climate Policy Shocks

The EU ETS is based on a “cap and trade” principle, where firms are faced with a set limit (the cap) on the amount of greenhouse gases they can emit in a given year. The cap is expressed in tons of CO2-equivalent, and firms can bid on allowances to have larger limits via a centralized auction system. These allowances are then traded on the EU ETS market, thereby setting a market price for carbon emissions.

Empirical Strategy

To estimate the impact of a carbon shock emanating from the EU ETS on international stock markets over 2005-2019, we combine annualized monthly country-sector (U.S. dollars) stock returns data for twenty-six countries with the carbon shock time series. We hypothesize that a policy announcement that leads to an unexpected increase in carbon prices will have a negative impact on stock returns, as an increase in carbon prices will raise the cost of production for firms within the EU and these costs will be passed on through input prices across global input-output linkages. After demonstrating a negative impact of carbon shock on global stock prices, we test for this hypothesis by making use of the global production network data from the World Input-Output Database (WIOD), as in some of our previous work.

Kanzig (2023) exploits policy announcements concerning the EU ETS since its inception in 2005 to identify policy “surprises” that affect carbon prices via supply-side forces (for example, the tightening of allowances). He follows the common approach used to identify monetary policy shocks that relies on price changes in futures markets around announcements. These price change surprises are used to construct the shock series that enter our regressions discussed below.

We consider two climate shock series in our regressions. The first is the raw series, which, following Kanzig (2023), is normalized such that a one standard deviation change in the shock corresponds to a one percentage point change in the energy component of the EU’s CPI on impact. This shock only takes values of one for targeted industries in the EU, which we refer to as “dirty” EU sectors (labeled as such if they are in the top 10 percent of emissions-to-output ratio across all EU country-sector pairs as of 2014, such as utilities and transportation sectors and some heavy manufacturing).

The second shock series weights the raw carbon surprise series by each country-sector’s intermediate input usage originating from “dirty” EU sectors as a ratio of the country-sector’s total production. This weighted shock then varies across country-sectors, as well as over time, and is meant to capture the relative importance of direct supply chain linkages of a given country-sector with the sectors that are directly affected by climate policy shocks in the EU ETS.

In addition to the climate shock variable, we include in our empirical analysis changes in the VIX, the broad U.S. dollar index, and the U.S. two-year Treasury rate, as these variables are also expected to influence stock returns.

Baseline Results

The chart below presents the monthly impact of the raw climate shock on stock returns in the average country-sector, where we include one (dark blue) and two (light blue) standard deviation bands. A positive innovation in carbon prices has an immediate negative impact on stock returns in the first month. This impact is not long-lived, only being statistically significant in the first month, but it is also not reversed subsequently. The magnitude of the impact implies that the largest carbon price shock observed in the series (which corresponds to a 0.02 percentage point change in the energy component of the EU’s CPI on impact) leads to a 37 percent decline in annualized monthly stock returns worldwide on average (which is about one-half of an average monthly return and about 17 percent of monthly stock returns deviations). 

The Effect of the Carbon Policy Shock Is Significant on Impact

Sources: Authors’ calculations based on carbon price shocks from Kanzig (2023) and stock market returns data from Refinitiv Global Equity Indices database.

The chart below explores how much of the effect is due to input-output linkages that transmit carbon shocks to foreign stock markets. The chart presents the impulse response of stock returns to the carbon shocks as defined above but weighted by the volume of inputs from the “dirty” EU sectors. The dynamics are similar to those for the raw shock series, in that the majority of the impact of an unexpected increase in stock prices accrues immediately and is not offset over time. Quantitatively, the largest observed weighted carbon shock leads to about a 33 percent decline in stock returns. That is, nearly all of the effect we previously observed is due to the linkages through trade in intermediate goods.

Input-Output Linkages Play a Role in Propagating the Carbon Price Shock

Sources: Authors’ calculations based on carbon price shocks from Kanzig (2023), stock market returns data from Refinitiv Global Equity Indices database, and data from the World Input-Output Database.

Conclusion

This post presents some early evidence on the spillover of domestic climate policy on foreign stock markets, based on the case of EU ETS. Our work complements recent studies by Bolton and Kacperczyk (2023) and Bolton, Lam, and Muûls (2023) that examine the impact of climate policies on domestic markets. By exploiting unexpected changes in carbon prices, we show that while there are sizable spillovers, they are not particularly large in the context of stock return volatility. Moreover, while the effects are not offset in the following months, they do not persist beyond the month of the impact. In ongoing work, we are further investigating the importance of specific trade linkages in the transmission of carbon shocks across stock markets and other determinants of cross-sectional differences in the impact that we observe in the data. Stay tuned.

Julian di Giovanni is the head of Climate Risk Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.  

Galina Hale is a professor of economics at the University of California, Santa Cruz.

Neel Lahiri is a former research analyst in the Federal Reserve Bank of New York’s Research and Statistics Group, and currently a graduate student at the University of Chicago.

Anirban Sanyal is an assistant advisor in the Reserve Bank of India’s research group.

How to cite this post:
Julian di Giovanni, Galina Hale, Neel Lahiri, and Anirban Sanyal, “International Stock Markets’ Reactions to EU Climate Policy Shocks,” Federal Reserve Bank of New York Liberty Street Economics, October 10, 2024, https://libertystreeteconomics.newyorkfed.org/2024/10/international-stock-markets-reactions-to-eu-climate-policy-shocks/.

You may also be interested in:

How Exposed Are U.S. Banks’ Loan Portfolios to Climate Transition Risks?

CRISK: Measuring the Climate Risk Exposure of the Financial System

Is the Green Transition Inflationary?

A key question in economic policy is how labor market tightness affects wage inflation and ultimately prices. In this post, we highlight the importance of two measures of tightness in determining wage growth: the quits rate, and vacancies per searcher (V/S)—where searchers include both employed and non-employed job seekers. Amongst a broad set of indicators, we find that these two measures are independently the most strongly correlated with wage inflation. We construct a new index, called the Heise-Pearce-Weber (HPW) Tightness Index, which is a composite of quits and vacancies per searcher, and show that it performs best of all in explaining U.S. wage growth, including over the COVID pandemic and recovery. 

The Importance of On-the-Job Search for Labor Market Tightness 

Labor market slack is often measured using the unemployment rate or the vacancy-to-unemployment ratio. In a recent Staff Report (Heise, Pearce, and Weber, 2024), we build on the theoretical foundation by Bloesch, Lee and Weber (2024) who argue that wage inflation should instead be strongly related to quits and to vacancies per job seeker. The key argument is that on-the-job search is important for understanding labor market tightness: since most new hires come from other jobs rather than from unemployment, an appropriate measure of labor market tightness must include employed job seekers. Consequently, labor market tightness should be measured by vacancies per searcher, where searchers combine employed, unemployed, and non-employed job seekers, rather than just vacancies over unemployment, or the unemployment rate. 

The intuition behind this argument is that when vacancies per searcher is high, competition for workers induces firms to raise offered wages to remain competitive. At the same time, workers will have more opportunities to change jobs, leading to a higher quits rate. As a result, the quits rate and vacancies per searcher are key components of the wage Phillips curve and more empirically informative than the unemployment rate or other measures of slack. 

Our recent Staff Report confirms this prediction in U.S. data. Crucially, we define searchers as a weighted sum of the number of short-term and long-term unemployed, employed, and non-employed workers, where the weights are based on estimates of these different workers’ search intensities. We then show that quits and vacancies per searcher outperform other standard measures of labor market tightness as predictors of wage growth. The table below demonstrates this point by reporting results from simple univariate regressions of the U.S. wage Phillips curve, ranking indicators by their ability to fit U.S. wage data since 1990. We regress three-month wage growth from the Employment Cost Index (ECI) on the measure listed, where we standardize each of the measures to have mean zero and standard deviation of one to help make comparisons of estimated coefficients. Column “Coefficient” presents the estimated coefficients and column “Fit” shows the regression fit. 

 We also create a composite measure of labor market tightness that takes a weighted average of quits and vacancies per searcher, using regression coefficients from a regression of wage growth on these two variables as the weights. This composite index, which we call the HPW Tightness Index, is ranked first in the table, indicating that it outperforms all other individual variables. According to the “Fit” column, it explains about 60 percent of wage growth during our sample period. The regression coefficient indicates that a one standard deviation increase in the index is associated with a 0.21 percentage point rise in wage growth. 

Quits and Vacancies per Searcher Outperform Other Measures of Labor Market Tightness

The chart below demonstrates the fit of the HPW Index visually by plotting it against wage growth, measured using a three-period moving average of the three-month growth in the ECI (both series are normalized to have a mean of zero and variance of one for ease of comparison). We compare our measure against a common measure of labor market tightness: the Conference Board’s survey measure of consumers’ perception of job availability. Both the Conference Board measure and the HPW Index track wage growth well in the pre-pandemic period. However, in the pandemic period, our measure performs significantly better.   

The HPW Index Tracks Wage Growth Well Even During COVID

Source: Authors’ calculations. Notes: The HPW Tightness Index, based on quits and vacancies per searcher, tracks wage growth well even during the COVID pandemic and recovery. All series are normalized to have zero mean and variance of one for ease of comparison. Wage growth is measured using the employment cost index. “CB Jobs Availability” is taken from the Conference Board. COVID period and recovery 2020:Q1—2022:Q4 is shaded. 

No Evidence for Nonlinearities in Wage Inflation 

Given recent interest in nonlinear effects of labor market tightness on price inflation (Benigno and Eggertsson, 2024), we also investigate whether there is a nonlinear relationship between labor market tightness and wage inflation. We do not find any evidence of nonlinearities. Indeed, there is nothing unusual in the wage/tightness relationship, either during the period of extreme tightness in the aftermath of COVID, or later. This can be seen in the chart below, where we provide a scatterplot of the HPW Tightness Index against wage inflation. We find a near-linear relationship between the two variables. 

No Evidence of a Nonlinear Relationship Between Wage Growth and Labor Market Tightness 

Concluding Remarks

In summary, the HPW Tightness Index of quits and vacancies per searcher performs well in summarizing labor market tightness for the purposes of determining wage inflation, consistent with theoretical results in Bloesch, Lee and Weber (2024). The relationship remained strong during the COVID period and recovery, suggesting that the empirical relationship documented is robust to even large, unusual economic shocks. 

Chart data

Sebastian Heise is a research economist in Labor and Product Market Studies in the Federal Reserve Bank of New York’s Research and Statistics Group. 

Jeremy Pearce is a research economist in Labor and Product Market Studies in the Federal Reserve Bank of New York’s Research and Statistics Group. 

Jacob P. Weber is a research economist in Macroeconomic and Monetary Studies in the Federal Reserve Bank of New York’s Research and Statistics Group. 

How to cite this post:
Sebastian Heise, Jeremy Pearce, and Jacob P. Weber, “A New Indicator of Labor Market Tightness for Predicting Wage Inflation,” Federal Reserve Bank of New York Liberty Street Economics, October 9, 2024, https://libertystreeteconomics.newyorkfed.org/2024/10/a-new-indicator-of-labor-market-tightness-for-predicting-wage-inflation/.

You may also be interested in:

Wage Growth and Labor Market Tightness

Do Unexpected Inflationary Shocks Raise Workers’ Wages?

What Has Driven the Labor Force Participation Gap since February 2020?

Trading activity in benchmark U.S. Treasury securities now concentrates on the last trading day of the month. Moreover, this stepped-up activity is associated with lower transaction costs, as shown by a smaller price impact of trades. We conjecture that increased turn-of-month portfolio rebalancing by passive investment funds that manage relative to fixed-income indices helps explain these patterns.

Trading Volume Concentrates on the Last Day of the Month

Since 2020, trading activity in benchmark Treasury notes and bonds has been roughly 33 percent higher on the last trading day of the month, on average, as shown in the chart below.

Trading Volume Is Higher on the Last Day of the Month

Moreover, as shown in the next chart, this end-of-month effect has been growing over time and was essentially nonexistent in the daily data before 2015.

End-of-Month Effects Have Been Growing over Time

To gauge the end-of-month patterns as far back as possible, we base our analysis on data from an interdealer broker whose coverage goes back to the early 2000s. Treasury TRACE, in contrast, better measures the breadth of trading in the market, including for the dealer-to-customer market and seasoned securities (see this post), but starts in July 2017. While there is no reason to think that our analysis is biased despite not viewing all market activity, month-end patterns may differ in other segments of the market.

Our analysis controls for day-of-week effects. This could matter because Friday is the last trading day of the month about three times more often than other weekdays (Friday is the last trading day for months that end on Saturday or Sunday). That said, activity on Friday is comparable to that of other weekdays, with benchmark note and bond volume about 4 percent lower than average. By comparison, volume on Monday is 12 percent lower than average, whereas volume on Tuesday, Wednesday, and Thursday is higher than average by 2 percent, 5 percent, and 7 percent, respectively.

Our analysis does not control for differences in end-of-month effects across months. Most notably, trading tends to be 16 percent lower on the last trading day of December because of the shortened holiday trading hours, lower trading desk staffing levels, and possibly positions that have been realigned in advance of that day. It follows that volume has been 37 percent higher on the last day of the month since 2020 if Decembers are excluded. We don’t find significant differences in end-of-month patterns for other months.

The end-of-month effect is robust to security type, as shown in the chart above. Two- and five-year notes are issued monthly, on the last day of the month, which might induce some monthly pattern in trading activity. However, the effect is roughly as strong for the ten-year note, which is issued mid-month. (Moreover, even for the two- and five-year notes, the more relevant auction days are concentrated two to three days before the end of each month, as shown in this paper’s Figure A1). Patterns are somewhat stronger for the less actively traded benchmark securities, which explains why the month-end effect in the first chart above is somewhat bigger than the effects observed in the second chart for the two-, five-, and ten-year notes (for the same 2020-2024 sample period).

Why Do These End-of-Month Effects Occur?

End-of-month effects have been studied across various asset markets and geographies, focusing mostly on prices and returns. For example, Ariel (1987) and Lakonishok and Smidt (1988) find higher U.S. equity market returns in the last few days of the month. Hartley and Schwarz (2019) and Etula et al. (2020) identify higher end-of-month U.S. Treasury returns and attribute them to price pressure from institutional investors’ trades and portfolio rebalancing. The former paper shows that net month-end purchases of Treasuries by insurers exceed net purchases on any other day of the month. Moreover, it finds that insurers that benchmark their performance more closely to indices show greater net purchases of the securities that are added to the index and that these purchases are concentrated on the end-of-month rebalancing date.

Evidence on end-of-month price patterns and returns, however, doesn’t immediately translate to higher than usual volume on the last trading day of the month. While not studying the impact on the aggregate market, Dick-Nielsen and Rossi (2019) examine the effects of corporate bond index rebalancing, also occurring on the last day of the month, and find that the volume of bonds that are excluded from the index is four to five times higher than normal.

Similar channels might be at play in the U.S. Treasury market. The high concentration of volume on the last trading day of the month and the increasing concentration over time coincide with the growth of passive funds that track index changes. For example, although still small as a fraction of U.S. Treasuries outstanding, exchange-traded funds (ETFs) that track Treasuries grew more than ten-fold between 2013 and mid-2024 as shown in the chart below, surpassing the two-fold growth of Treasuries outstanding over the same period. Asset managers are increasingly managing relative to indices that are rebalanced at the end of each month and this may be causing investors to increasingly trade at that time.

U.S. Treasury ETF Assets Under Management Are Growing Rapidly

How Is End-of-Month Liquidity Affected?

The relationship between trading volume and liquidity is not a simple one. Volume and volatility are positively correlated, and volatility and liquidity are negatively correlated (see this study, for example). One therefore might expect a negative relationship between volume and liquidity, and that’s what is seen at times of market turmoil, such as around the near-failure of Long-Term Capital Management (see this paper), during the 2007-09 financial crisis (see this paper), and during the COVID-19-related disruptions of March 2020 (see this paper).

In the case of month-ends, however, the purported reasons for the higher end-of-month activity are not information-based, and the higher volume is not associated with higher volatility. Higher volume that arises independent of volatility is associated with improved liquidity, consistent with larger Treasury issues, and the most recently issued Treasuries, being more actively traded and more liquid (see this paper and this paper).

It follows that liquidity tends to be markedly better on the last trading day of the month. Since 2020, price-impact coefficients for benchmark notes have been about 26 percent lower, on average, as shown in the chart below, indicating better liquidity.

Price Impact Is Lower on the Last Day of the Month

Moreover, this end-of-month liquidity improvement has been increasing in magnitude over time, in a manner akin to that for trading volume, as shown in the next chart.

End-of-Month Price Impact Effects Have Been Increasing in Magnitude over Time

Similar but weaker patterns are observed for other measures of market liquidity. Quoted depth at the inside tier has been about 6 percent higher on the last trading day of the month since 2020, on average, implying better liquidity, as compared to 9 percent lower between 2005 and 2009 (percent differences are first calculated for each of the two-, five-, and ten-year notes, and then averaged across them). Bid-ask spreads have been about 1 percent narrower on the last day of the month, on average, suggesting slightly better liquidity, as compared to 2 percent wider between 2005 and 2009. The weak end-of-month effects for spreads in particular are likely attributable to minimum tick sizes, which cause spreads to vary little outside times of market stress (see this paper, for example).

Implications

Based on this post’s findings only, one might conclude that the last trading day of the month is an especially good time to trade because of the day’s higher trading volume and lower transaction costs. However, the evidence of periodicity in returns from other studies suggests that advantageous times to trade vary for other reasons and differ between buyers and sellers. These monthly patterns also change over time, as shown in this post, warranting close watching of these patterns going forward.

Henry Dyer is a research analyst in the Federal Reserve Bank of New York’s Research and Statistics Group.

Michael J. Fleming is the head of Capital Markets Studies in the Federal Reserve Bank of New York’s Research and Statistics Group. 

Or Shachar is a financial research advisor in Capital Markets Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.

How to cite this post:
Henry Dyer, Michael Fleming, and Or Shachar, “End‑of‑Month Liquidity in the Treasury Market,” Federal Reserve Bank of New York Liberty Street Economics, September 24, 2024, https://libertystreeteconomics.newyorkfed.org/2024/09/end-of-month-liquidity-in-the-treasury-market/.

You may also be interested in:

Has Treasury Market Liquidity Improved in 2024?

Unlocking the Treasury Market through TRACE

Measuring Treasury Market Liquidity

The Evolution of Treasury Market Liquidity: Evidence from 30 Years of Limit Order Book Data

Economic surveys are very popular these days and for a good reason. They tell us how the folks being surveyed—professional forecasters, households, firm managers—feel about the economy. So, for instance, the New York Fed’s Survey of Consumer Expectations (SCE) website displays an inflation uncertainty measure that tells us households are more uncertain about inflation than they were pre-COVID, but a bit less than they were a few months ago. The Philadelphia Fed’s Survey of Professional Forecasters (SPF) tells us that forecasters believed last May that there was a lower risk of negative 2024 real GDP growth than there was last February. The question addressed in this post is: Does this information actually have any predictive content? Specifically, I will focus on the SPF and ask: When professional forecasters indicate that their uncertainty about future output or inflation is higher, does that mean that output or inflation is actually becoming more uncertain, in the sense that the SPF will have a harder time predicting these variables?

Measuring Uncertainty in Probabilistic Surveys—a Quick Recap

In my companion post I covered some material that is going to be helpful for understanding this post. Let me summarize it for you:

• Probabilistic surveys ask respondents not only for point predictions (that is, one number, such as the answer to the question “What is output growth going to be in 2024?”), but also try to elicit the entire probability distribution of possible outcomes. The SPF asks for both point forecasts and probabilities. In particular, SPF forecasters provide probabilistic forecasts for real GDP growth and GDP deflator inflation for the current and the following year. This is the survey used in this post.
• There are a variety of approaches for translating the answer to probabilistic surveys into measures of uncertainty. In this post I am going to use the approach developed in a recent Staff Report with my coauthors Federico Bassetti and Roberto Casarin. For some questions the approach used matters as discussed in the first post in this series.
• Professional forecasts massively disagree on how much uncertainty there is out there, and they also change their assessment of uncertainty over time (as is the case for the SCE inflation uncertainty measure mentioned above).
• These differences can in principle be explained by a theory called the noisy rational expectations (RE) hypothesis: forecasters receive both public and private signals about the state of the economy, which they do not observe. Differences in the variance of the private signals explain differences in uncertainty across forecasters (basically, some folks are better forecasters than others), and changes in the variance of either the public (for example, due to something like COVID) or private signals can explain why they change their mind about uncertainty.

Does Subjective Uncertainty Map into Objective Uncertainty, That Is, Forecast Errors?

Still, under rational expectations it better be that forecasters that are more uncertain are actually worse forecasters, in the sense that they make on average worse forecast errors. And that if they feel economic uncertainty has declined, they should be able to actually predict the economy better if RE holds. By regressing ex-post forecast errors (specifically, the logarithm of their absolute value) on subjective uncertainty (specifically, the logarithm of its standard deviation) across forecasters and time, we test whether subjective uncertainty maps into objective uncertainty. That is, we test whether it is the case that when ex-ante uncertainty increases by, say, 50 percent, either over time or across forecasters, ex-post uncertainty also increases by the same amount. As a byproduct, we also test whether the noisy rational expectations hypothesis holds water.

In the Staff Report I mentioned we run this regression for the panel of individual SPF forecasters from 1982 to 2022 for both output growth and inflation, and for eight different forecast horizons—from eight to one quarter-ahead (recall that SPF forecasters provide probabilistic forecasts for growth and inflation for the current and the following year; hence the eight quarter-ahead uses the surveys conducted in the first quarter of the year before the realization; the one quarter-ahead uses the surveys conducted in the fourth quarter of the same year). The thick dots in charts below show the OLS estimates and the whiskers indicate 90 percent posterior coverage intervals based on Driscoll-Kraay standard errors.

Do Differences in Subjective Uncertainty Map into Differences in Forecast Accuracy?

Source: Author’s calculations.

For output growth, for horizons longer than five quarters we cannot reject the hypothesis that the coefficient is zero and the point estimates are all small: In sum, we find no relationship between subjective uncertainty and the size of the ex-post forecast error for horizons beyond one year. As the forecast horizon shortens, the relationship becomes tighter, and for horizons of three quarters or less we cannot reject the hypothesis that the coefficient is one, that is, we cannot reject the noisy RE hypothesis. (Recall that the coefficient measures the extent to which ex-post uncertainty changes when ex-ante uncertainty changes, either over time or across forecasters. A value of zero means that changes in subjective or ex-ante uncertainty have no bearing whatsoever on ex-post forecast errors; a value of one means that ex-ante and ex-post uncertainty move in unison, as RE would imply.)

For inflation, the OLS estimates hover between 0.2 and 0.5 for longer horizons, but increase toward one as the horizon shortens, with estimates that are also not significantly different from one for horizons of three quarters or less. The appendix of the Staff Report shows that these results are broadly robust to different samples and specifically do not depend on including the COVID period.

The chart below shows the estimates controlling for time, forecaster, and both time and forecaster-fixed effects. The results with time-fixed effects (top panels) test whether cross-sectional differences in subjective uncertainty across forecasters map into similar differences in forecast errors, controlling for factors that affect all forecasters (such as recessions and the COVID period). For output, the answer is absolutely not at long horizons, although for short horizons the correspondence between the two becomes tighter, if not quite one-to-one. For inflation the relationship is also far from one-to-one at longer horizons, but for horizons less than two quarters the coefficient is indistinguishable from one.

The bottom panels show the results controlling for forecaster-fixed effects. They show that when subjective uncertainty changes over time—either because of aggregate uncertainty shocks or because the quality of their private signal has changed—on average it has no bearing for forecast accuracy for longer horizons, but maps one-to-one into corresponding changes in the absolute forecast errors for short horizons. The results with forecaster-fixed effects shed light on whether forecasters correctly anticipate periods of macroeconomic uncertainty. They clearly do not for any horizon longer than one year. But when they are already in a high uncertainty period (that is, for short horizons), this is reflected in their subjective uncertainty.

Controlling for Fixed Effects

What do we conclude from the evidence discussed in this series of posts? Professional forecasters clearly do not behave on average like the noisy rational expectations model suggests. But the evidence also indicates that any proposed theory of deviations from rational expectations better account for the fact that such deviations are horizon dependent. Whatever biases forecasters have, whether due to their past experiences or other factors, seem to go away as the horizon gets closer. In other words, professional forecasters cannot really predict what kind of uncertainty regime will materialize in the future, but they seem to have a good grasp of what regime we are in at the moment.

Marco Del Negro is an economic research advisor in Macroeconomic and Monetary Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.

How to cite this post:
Marco Del Negro, “Can Professional Forecasters Predict Uncertain Times?,” Federal Reserve Bank of New York Liberty Street Economics, September 4, 2024, https://libertystreeteconomics.newyorkfed.org/2024/09/can-professional-forecasters-predict-uncertain-times/.

You may also be interested in:

Are Professional Forecasters Overconfident?

A Bayesian Approach for Inference on Probabilistic Surveys

 The post-COVID years have not been kind to professional forecasters, whether from the private sector or policy institutions: their forecast errors for both output growth and inflation have increased dramatically relative to pre-COVID (see Figure 1 in this paper). In this two-post series we ask: First, are forecasters aware of their own fallibility? That is, when they provide measures of the uncertainty around their forecasts, are such measures on average in line with the size of the prediction errors they make? Second, can forecasters predict uncertain times? That is, does their own assessment of uncertainty change on par with changes in their forecasting ability? As we will see, the answer to both questions sheds light of whether forecasters are rational. And the answer to both questions is “no” for horizons longer than one year but is perhaps surprisingly “yes” for shorter-run forecasts. 

What Are Probabilistic Surveys? 

Let’s start by discussing the data. The Survey of Professional Forecasters (SPF), conducted by the Federal Reserve Bank of Philadelphia, elicits every quarter projections from a number of individuals who, according to the SPF, “produce projections in fulfillment of their professional responsibilities [and] have long track records in the field of macroeconomic forecasting.’’ These people are asked for point projections for a number of macro variables and also for probability distributions for a subset of these variables such as real output growth and inflation. The way the Philadelphia Fed asks for probability distributions is by dividing the real line (the interval between minus infinity and plus infinity) into “bins” or “ranges”—say, less than 0, 0 to 1, 1 to 2, …—and asking forecasters to put probabilities to each bin (see here for a recent example of the survey form).  The result, when averaged across forecasters, is the histogram shown below for the case of core personal consumption expenditure (PCE) inflation projections for 2024 (also shown on the SPF site).

An Example of Answers to Probabilistic Surveys 

So, for instance, in mid-May, forecasters expected on average a 40 percent probability that core PCE inflation in 2024 will be between 2.5 and 2.9 percent. Probabilistic surveys, whose study was pioneered by the economist Charles Manski, have a number of advantages compared to surveys that only ask for point projections: they provide a wealth of information that is not included in point projections, for example, on uncertainty and risks to the outlook. For this reason, probabilistic surveys have become more and more popular in recent years. The New York Fed’s Survey of Consumer Expectations (SCE), for example, is a shining example of a very popular probabilistic survey. 

In order to obtain from probabilistic surveys information that is useful to macroeconomists—for example, measures of uncertainty—one has to extract the probability distribution underlying the histogram and use it to compute the object of interest; if one is interested in uncertainty, that would be the variance or an interquartile range. The way this is usually done (for instance, in the SCE) is to assume a specific parametric distribution (in the SCE case, a beta distribution) and to choose its parameters so that it best fits the histogram. In a recent paper with my coauthors Federico Bassetti and Roberto Casarin, we propose an alternative approach, based on Bayesian nonparametric techniques, that is arguably more robust as it depends less on the specific distributional assumption. We argue that for certain questions, such as whether forecasters are overconfident, this approach makes a difference. 

The Evolution of Subjective Uncertainty for SPF Forecasters  

We apply our approach to individual probabilistic surveys for real output growth and GDP deflator inflation from 1982 to 2022. For each respondent and each survey, we then construct a measure of subjective uncertainty for both variables. The chart below plots these measures for next year’s output growth (that is, in 1982 this would be the uncertainty about output growth in 1983). Specifically, the thin blue crosses indicate the posterior mean of the standard deviation of the individual predictive distribution. (We use the standard deviation as opposed to the variance because its units are easily grasped quantitatively and are comparable with alternative measures of uncertainty such as the interquartile range, which we include in the paper’s appendix. Recall that the units of a standard deviation are the same as those of the variable being forecasted.) Thin blue lines connect the crosses across periods when the respondent is the same. This way you can see whether respondents change their view on uncertainty. Finally, the thick black dashed line shows the average uncertainty across forecasters in any given survey. In this chart we plot the result for the survey collected in the second quarter of each year, but the results for different quarters are very similar. 

Subjective Uncertainty for Next Year’s Output Growth by Individual Respondent

The chart shows that, on average, uncertainty for output growth projections declined from the 1980s to the early 1990s, likely reflecting a gradual learning about the Great Moderation (a period characterized by less volatility in business cycles), and then remained fairly constant up to the Great Recession, after which it ticked up toward a slightly higher plateau. Finally, in 2020, when the COVID pandemic struck, average uncertainty grew twofold. The chart also shows that differences in subjective uncertainty across individuals are very large and quantitatively trump any time variation in average uncertainty. The standard deviation of low-uncertainty individuals remains below one throughout most of the sample, while that of high-uncertainty individuals is often higher than two. The thin blue lines also show that while subjective uncertainty is persistent—low-uncertainty respondents tend to remain so—forecasters do change their minds over time about their own uncertainty.  

The next chart shows that, on average, subjective uncertainty for next year’s inflation declined from the 1980s to the mid-1990s and then was roughly flat up until the mid-2000s. Average uncertainty rose in the years surrounding the Great Recession, but then declined again quite steadily starting in 2011, reaching a lower plateau around 2015. Interestingly, average uncertainty did not rise dramatically in 2020 through 2022 despite COVID and its aftermath, and despite the fact that, for most respondents, mean inflation forecasts (and the point predictions) rose sharply. 

Subjective Uncertainty for Next Year’s Inflation by Individual Respondent 

Are Professional Forecasters Overconfident? 

Obviously, the heterogeneity in uncertainty just documented flies in the face of full information rational expectations (RE): if all forecasters used the “true” model of the economy to produce their forecasts—whatever that is—they would all have the same uncertainty and this is clearly not the case. There is a version of RE, called noisy RE, that may still be consistent with the evidence: according to this theory, forecasters receive both public and private signals about the state of the economy, which they do not observe. Heterogeneity in the signals, and in their precision, explains the heterogeneity in their subjective uncertainty: forecasters receiving a poor/more precise signal have higher/lower subjective uncertainty. Still, under RE, their subjective uncertainty better match the quality of their forecasts as measured by their forecast error—that is, forecasters should be neither over- nor under-confident. We test this hypothesis by checking whether, on average, the ratio of ex-post (squared) forecast errors over subjective uncertainty, as measured by the variance of the predictive distribution, equals one. 

The thick dots in charts below show the average ratio of squared forecast errors over subjective uncertainty for eight to one quarters ahead (the eight-quarter-ahead measure uses the surveys conducted in the first quarter of the year before the realization; the one-quarter-ahead measure uses the surveys conducted in the fourth quarter of the same year), while the whiskers indicate 90 percent posterior coverage intervals based on Driscoll-Kraay standard errors.

Do Forecasters Over- or Under- Estimate Uncertainty? 

We find that for long horizons—between two and one years—forecasters are overconfident by a factor ranging from two to four for both output growth and inflation. But the opposite is true for short horizons: on average forecasters overestimate uncertainty, with point estimates lower than one for horizons less than four quarters (recall that one means that ex-post and ex-ante uncertainty are equal, as should be the case under RE). The standard errors are large, especially for long horizons. For output growth, the estimates are significantly above one for horizons greater than six, but, for inflation, the 90 percent coverage intervals always include one. We show in the paper that this pattern of overconfidence at long horizons and underconfidence at short horizons is robust across different sub-samples (e.g., excluding the COVID period), although the degree of overconfidence for long horizons changes with the sample, especially for inflation. We also show that it makes a big difference whether one uses measures of uncertainty from our approach or that obtained from fitting a beta distribution, especially at long horizons.

While the findings are in line with the literature on overconfidence (see the volume edited by Malmendier and Taylor [2015]) for output for horizons greater than one year, results are more uncertain for inflation. For horizons shorter than three quarters, the evidence shows that forecasters if anything overestimate uncertainty for both variables. What might explain these results? Patton and Timmermann (2010) show that dispersion in point forecasts increases with the horizon and argue that this result is consistent with differences not just in information sets, as the noisy RE hypothesis assumes, but also in priors/models, and where these priors matter more for longer horizons. In sum, for short horizons forecasters are actually slightly better at forecasting than they think they are. For long horizons, they are a lot worse at forecasting and they are not aware of it.

In today’s post we looked at the average relationship between subjective uncertainty and forecast errors. In the next post we will look at whether differences in uncertainty across forecasters and/or over time map into differences in forecasting accuracy. We will see that again the forecast horizon matters a lot for the results. 

Marco Del Negro is an economic research advisor in Macroeconomic and Monetary Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.

How to cite this post:
Marco Del Negro , “Are Professional Forecasters Overconfident? ,” Federal Reserve Bank of New York Liberty Street Economics, September 3, 2024, https://libertystreeteconomics.newyorkfed.org/2024/09/are-professional-forecasters-overconfident/.

You may also be interested in:

A Bayesian Approach for Inference on Probabilistic Surveys

Long-run trends in increased access to credit are thought to improve real activity. However, “rapid” credit expansions do not always end well and have been shown in the academic literature to predict adverse real outcomes such as lower GDP growth and an increased likelihood of crises. Given these financial stability considerations associated with rapid credit expansions, being able to distinguish in real time “good booms” from “bad booms” is of crucial interest for policymakers. While the recent literature has focused on understanding how the composition of borrowers helps distinguish good and bad booms, in this post we investigate how the composition of lending during a credit expansion matters for subsequent real outcomes.

Bank Lending and Nonbank Lending Do Not Always Go Hand in Hand

We start by documenting that credit extended by the banking sector and credit extended by the nonbanking sector do not always move together. The chart below plots the time series of three-year growth in bank credit (credit given by banks to the private sector) and three-year growth in nonbank credit (credit given by nonbanks to the private sector) in the U.S. While there are some periods when the two series move together—for example, following the 2008 Global Financial Crisis—most of the time, growth in bank and nonbank credit evolve separately. That is, for most years since 1950, bank lending to the private nonfinancial sector in the U.S. has moved asynchronously to nonbank lending.

Bank and Nonbank Credit Growth in the U.S. Are Asynchronous…

In contrast, the chart below shows that, in Japan, bank and nonbank credit move together much more closely than in the U.S., with only a few periods in which growth in nonbank lending is disjointed from growth in bank lending.

…while Bank and Nonbank Credit Growth in Japan Evolve Closely Together

To more systematically investigate the synchronicity between bank and nonbank credit, we plot one-year growth in bank credit against one-year growth in nonbank credit for a large number of countries and years. The chart below shows that a large number of country-year observations are quite far from the 45-degree line, which means that the growth rate of one type of credit is quite different from the growth rate of the other type.

The chart also shows two additional features of bank versus nonbank lending. First, although there are periods in which both bank debt and nonbank debt move in the same direction (i.e., they have the same sign), a considerable number of country-year observations feature opposite signs. That is, one type of lending is expanding while the other is contracting, which suggests a substitution between bank and nonbank lending.

Second, overall booms in private credit can be driven by either bank or nonbank expansions. The blue diamonds highlight the country-years that correspond to the start of booms in overall private credit following the definition of credit booms in Verner (2022). As the illustration shows, a number of booms are financed by just one type of lender. That is, we observe a number of country-year observations identified as the beginning of a credit boom with very little subsequent expansion in one type of lending.

Distinct Evolution in Bank and Nonbank Credit Seen Across Country-Years

Does Lender Composition Matter for Real Outcomes?

We next document that the composition of lending during a credit expansion matters for subsequent real outcomes. We do this by first computing three-year growth in bank credit (credit given by banks to the private sector) and three-year growth in nonbank credit (credit given by nonbanks to the private sector) and then estimating a predictive regression for cumulative annualized real GDP growth going forward.

The chart below shows the results for a panel of thirty-three countries, including both advanced and emerging economies, from 1966 to 2020. Starting with the red line, the illustration shows that growth in nonbank credit predicts negative GDP growth in the short and medium term (one to four years) but that the effect on GDP growth in the medium to long term (five to ten years) is not significantly different from zero. On the other hand, the blue line shows that growth in bank credit predicts negative GDP growth in the medium to long term (three to ten years).

Put together the results seem to indicate that bank credit growth is associated with more persistently negative real outcomes.

Expansions in Bank Credit Have a Prolonged Adverse Impact on Average Future Real GDP Growth

Growth in Bank Lending Is a Better Predictor of Extreme GDP Growth Events

We further find that bank and nonbank credit expansions predict differentially the downside risk to growth—that is, the probability of extreme negative real GDP growth realizations. The blue line in the chart below thus shows that the likelihood of an extreme negative real GDP growth realization—which we define as year-on-year real GDP growth below -2 percent—increases following expansions in bank credit for horizons of one to three years. Importantly, at the same horizon, growth in nonbank credit actually lowers the probability of a large drop in real GDP growth. In particular, a one-standard-deviation-higher growth rate in bank credit increases the probability of real GDP growth below -2 percent in two years’ time by 2.5 percentage points relative to a baseline 6 percent probability in our sample. In contrast, a one-standard-deviation-higher growth rate in nonbank credit lowers the probability of real GDP growth below -2 percent in two years’ time by 1.9 percentage points.

Expansions in Nonbank Credit Predict a Lower Probability of Extreme Negative Growth Outcomes at the Two-to-Three-Year Horizon

Conclusion

While the academic literature has shown that who borrows (households vs. firms, or firms in the tradable vs. nontradable sectors) during credit expansions matters for subsequent real outcomes, we show here that the composition of the lender also matters. To more fully explore why and how lender composition matters, in our Staff Report, we investigate how variation in the composition of the financial sector across countries and over time translates into overall credit booms and the relative expansions in bank and nonbank credit. We focus in particular on nonfinancial firm borrowing, collecting national accounts data on both nonfinancial and financial sectors’ balance sheets, and studying the general equilibrium sensitivities of debt growth to aggregate conditions through the lens of a credit supply-demand model.

Nina Boyarchenko is the head of Macrofinance Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.

Leonardo Elias is a financial research economist in Macrofinance Studies in the Federal Reserve Bank of New York’s Research and Statistics Group. 

How to cite this post:
Nina Boyarchenko and Leonardo Elias, “The Disparate Outcomes of Bank‑ and Nonbank‑Financed Private Credit Expansions,” Federal Reserve Bank of New York Liberty Street Economics, August 20, 2024, https://libertystreeteconomics.newyorkfed.org/2024/08/the-disparate-outcomes-of-bank-and-nonbank-financed-private-credit-expansions/.

You may also be interested in:

Financing Private Credit

The Changing Landscape of Corporate Credit

The Federal Reserve Bank of New York’s July 2024 SCE Labor Market Survey shows a year-over-year increase in the average reservation wage—the lowest wage respondents would be willing to accept for a new job—to $81,147, but a decline from a series’ high of $81,822 in March 2024. In this post, we investigate how the recent dynamics of reservation wages differed across individuals and how reservation wages are related to individuals’ expectations about their future labor market movements.

Reservation Wages 

The SCE Labor Market Survey, which has been fielded every four months since March 2014 as part of the broader Survey of Consumer Expectations (SCE), provides information on consumers’ experiences and expectations regarding the labor market. The data, together with a companion set of interactive charts showing a subset of the data that we collect, are published every four months by the New York Fed’s Center for Microeconomic Data. As with other components of the SCE, we report statistics not only for the overall sample, but also by various demographic categories, namely age, gender, education, and household income. The underlying micro (individual-level) data for the full survey are made available with an eighteen-month lag. 

Our measure of reservation wage comes from the following question in the SCE Labor Market Survey:  

Suppose someone offered you a job today in a line of work that you would consider. What is the lowest wage or salary you would accept (BEFORE taxes and other deductions) for this job? 

This question is asked to all respondents (that is, to those who are employed, unemployed, or out of the labor force). For those who are out of work, this measure provides information on the tradeoff between out-of-work transfers (such as unemployment insurance or means-tested government transfers) and expected salaries. For those who are currently employed, this measure is informative of the tradeoff between their current total compensation package (including the salary and non-wage amenities) and alternative compensation packages potentially available at other employers.  

The chart below shows that the average reservation wage has increased by 31.4 percent between March 2020 and July 2024. Note, however, that this measure does not account for inflation. Deflating the series using the Consumer Price Index (CPI) indexed to 1 for March 2020, we find that the average real reservation wage increased by 8.2 percent during the same time period, while in fact it declined in the four years prior to the pandemic. This shows that even though part of the increase in respondents’ reservation wages is due to inflation, there has still been a rise in the minimum compensation respondents require to accept (new) job offers in real terms. However, it is worth noting that the average reservation wage in real terms has been essentially flat since early 2021.

The Average Reservation Wage Increased Faster Than Inflation since 2020

Next, we examine how this upward trend in reservation wages varied by respondents’ education and employment status. The left panel of the chart below shows that the growth in average reservation wages, relative to March 2020, was primarily driven by the respondents without a college degree up until March 2022. This implied a compression in reservation wages across education levels, since those with lower education have lower reservation wages.   

Between mid-2022 and the end of 2023, reservation wages have grown faster among college graduates, reversing the previous trend in the reservation wage compression. However, since the end of last year, the reservation wages of respondents without a college degree have accelerated again. In July 2024, the reservation wages of those without a college degree were closer to those of college graduates than they were before the onset of the pandemic. 

In the right panel of the chart below, we show that since mid-2022 the average reservation wage of the non-employed grew faster than that of employed respondents. This stands in contrast to the dynamics in the previous two years, as we had discussed in an earlier Liberty Street Economics post. Overall, as of July 2024, the average reservation wage growth of non-employed respondents has caught up with, and in fact exceeded, the average of employed respondents since the onset of the pandemic. 

Non-employed Consumers and Those without a College Degree Experienced Faster Reservation Wage Growth 

Reservation Wages and Expectations about Labor Market Flows 

We next examine how reservation wages are linked to expected labor market movements. Every four months, the SCE Labor Market survey elicits respondents’ expected likelihood of being non-employed, employed, or employed with the same employer (if employed) in the subsequent four months. In the table below, we relate these probabilistic expectations to reservation wages, controlling for the respondents’ time-varying observable characteristics and for individual fixed effects.  

The results show that workers with a 1 standard deviation ($44,614) higher reservation wage report 2.72 percentage points (or 32 percent) lower likelihood of moving to a new employer in the subsequent four months (column 1). On the other hand, column 2 shows that workers’ expectations about moving into non-employment do not statistically differ based on their reservation wages. For non-employed workers (including those who are unemployed and out of the labor force), we also observe that the average likelihood of moving into employment over the subsequent four months does not statistically differ based on respondent’s reservation wages.

Reservation Wages Are Meaningfully Related to Households’ Expected Job-to-Job Movements 

	Employed		Employed		Non-Employed
	(1)		(2)		(3)
	Probability of Moving to a New Job		Probability of Moving into Non-Employment		Probability of Moving into Employment
Reservation Wage ($1,000)	-0.061*** (0.013)		-0.011 (0.009)		-0.022 (0.031)
Demographic Controls
Individual Fixed Effects
Dep. Var. Mean	8.172		3.143		14.078
R-squared	0.612		0.604		0.774
Observations	7,253		7,245		1,164

Conclusion 

Results of the July 2024 SCE Labor Market Survey show a slight decline in the average reservation wage to $81,147 from a series’ high $81,822 in March. However, we find that the average reservation wage increased faster than inflation since the onset of the pandemic. Overall, the patterns suggest a compression in the reservation wage distribution by education and employment status. We also document that reservation wages are meaningfully related to households’ expectations about their future labor market movements.  

Gizem Kosar is a research economist in Consumer Behavior Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.

Davide Melcangi is a research economist in Labor and Product Market Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.

Sasha Thomas is a research analyst in the Federal Reserve Bank of New York’s Research and Statistics Group.

How to cite this post:
Gizem Kosar, Davide Melcangi, and Sasha Thomas, “An Update on the Reservation Wages in the SCE Labor Market Survey,” Federal Reserve Bank of New York Liberty Street Economics, August 19, 2024, https://libertystreeteconomics.newyorkfed.org/2024/08/an-update-on-the-reservation-wages-in-the-sce-labor-market-survey/.

You may also be interested in:

SCE Labor Market Survey

SCE Labor Market Survey Shows Average Reservation Wage Continues Upward Trend

Survey of Consumer Expectations

 
The Federal Reserve (Fed) implements monetary policy in a regime of ample reserves, where short-term interest rates are controlled mainly through the setting of administered rates, and active management of the reserve supply is not required. In yesterday’s post, we proposed a methodology to evaluate the ampleness of reserves in real time based on the slope of the reserve demand curve—the elasticity of the federal (fed) funds rate to reserve shocks. In this post, we propose a suite of complementary indicators of reserve ampleness that, jointly with our elasticity measure, can help policymakers ensure that reserves remain ample as the Fed shrinks its balance sheet.

Complementary Measures of Ampleness of Reserves

As we explain in yesterday’s post, one could operationalize the notion of ample reserves as the region of the reserve demand curve where the slope is only modestly negative, which means that the elasticity of the fed funds rate to shocks in the supply of reserves is small. At higher reserve levels (abundant reserves), the elasticity is zero (that is, the curve is flat); at lower levels (scarce reserves), the elasticity is negative and large (the curve is steeply sloped).

The ampleness of central bank reserves, however, affects not only the fed funds rate but also other important money-market variables and bank liquidity management. In today’s post, we introduce four new indicators of reserve ampleness, which are complementary to our estimates of the slope of the reserve demand curve and work as an external validity check of our measure of reserve ampleness. Importantly, these indicators do not rely on the same sources of information as our elasticity measure because they do not use variation in the fed funds rate or quantity of reserves; rather they look at other variables that, based on economic theory and institutional details, should be affected by the ampleness of reserves.

• Late Payments
  The first indicator is the share of interbank payments settled after 5 p.m. Throughout each business day, banks use their accounts at the Fed to make and receive payments, transferring reserves over a system known as Fedwire^® Funds Service. As the supply of reserves declines and transitions from abundant to ample, banks have an incentive to postpone their outgoing payments, pushing the settlement towards the end of the business day to ensure they have sufficient reserves to settle their transactions (as explained in this paper).

• Banks’ Intraday Overdrafts
  If banks’ ability to postpone outgoing payments is limited, they may increase their use of intraday credit provided by the Fed, known as daylight or intraday overdraft. A bank incurs an intraday overdraft when the balance in its account at the Fed is negative during the business day. As reserves become less abundant and banks more liquidity constrained, we would expect to observe more intraday overdrafts. To reflect broad liquidity conditions in the banking system, we take the average dollar value of banks’ intraday overdrafts as an indicator of the relative ampleness of reserves.

• Domestic Borrowing in the Fed Funds Market
  As discussed in this post, domestic banks predominantly borrow in the fed funds market when they need short-term liquidity, whereas U.S. branches of foreign banks actively borrow in that market to earn the spread between the rate of interest on reserves (IOR) and the fed funds rate, even when they are awash with reserves. An increase in the volume of fed funds borrowing by domestic banks could therefore signal that reserves are becoming less abundant as banks are more liquidity constrained.

• Upward Pressure in Repo Rates
  Rates on overnight repurchase agreements (repos) can influence the fed funds rate because repos and fed funds are close substitutes for many market participants. Moreover, as discussed in several papers (see here, here, and here), lower reserve levels can increase repo rates by tightening the liquidity constraints of large banks active as repo intermediaries. We measure upward pressure in repo rates as the share of overnight Treasury repo transacted at or above IOR.

Looking at the Suite of Measures Jointly

By construction, all these indicators should increase as aggregate reserves transition from being abundant to ample and then scarce. Indeed, the chart below shows that, over time, our four complementary indicators of reserve ampleness have been consistent with our real-time estimate of the fed-funds-rate elasticity to reserve shocks. In each panel, we plot one of the indicators and the real-time elasticity estimate: in all cases, the complementary indicators increase as the fed-funds-rate elasticity decreases, indicating a decline in reserve ampleness. In particular, all complementary indicators start to trend upward—suggesting that reserves were transitioning from being abundant to ample—around the end of 2017 or beginning of 2018, when the elasticity becomes negative for the first time since 2014. The indicators then reach a maximum between September 2019 and March 2020, as our real-time elasticity measure reaches its minimum, indicating the greatest scarcity of reserves over the last ten years. Since then, they have gone back to their 2014-17 levels, suggesting that reserves have been abundant over the last four years, consistent with the elasticity being zero.  

Indicators Increase as the Elasticity and Reserve Ampleness Decline

The following spider web chart shows the four complementary indicators together with (the absolute value of) our measure of the elasticity at four different points in time: 2018:Q3, 2019:Q1, September 2019, and 2024:Q1. For each indicator, the point on the respective axis in the innermost dashed pentagon represents the level of most abundant reserves during 2014-2024:Q1 according to that indicator; the point on the indicator’s axis in the outermost pentagon corresponds to the level at which reserves were most scarce. As of the first quarter of 2024, all indicators suggest that reserves remain abundant and far from their 2018-19 conditions.

Reserves Remain Abundant as of 2024:Q1

In Sum

Monitoring reserve ampleness is a key task of the Fed, particularly amid the ongoing balance sheet reduction process. In yesterday’s post, we proposed a measure of reserve ampleness based on estimating the slope of the reserve demand curve (that is, the elasticity of the fed funds rate to changes in aggregate reserves) at the daily frequency. In this post, we propose a suite of complementary indicators based on different data that describe conditions in money markets and bank liquidity, whose evolution over time is consistent with our estimates of the fed-funds-rate elasticity. These indicators, jointly with our real-time elasticity estimates, can help inform policymakers on whether aggregate reserves in the banking system are becoming less abundant and potentially scarce.

Gara Afonso is the head of Banking Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.

Kevin Clark is a Capital Markets Trading Principal in the Federal Reserve Bank of New York’s Markets Group.

Brian Gowen is a Capital Markets Trading Principal in the Federal Reserve Bank of New York’s Markets Group.

Gabriele La Spada is a financial research advisor in Money and Payments Studies in the Federal Reserve Bank of New York’s Research and Statistics Group.   

JC Martinez is a research analyst in the Federal Reserve Bank of New York’s Research and Statistics Group.

Jason Miu is a Capital Markets Trading Associate Director in the Federal Reserve Bank of New York’s Markets Group.

William Riordan is a Capital Markets Trading Advisor in the Federal Reserve Bank of New York’s Markets Group.

How to cite this post:
Gara Afonso, Kevin Clark, Brian Gowen, Gabriele La Spada, JC Martinez, Jason Miu, and Will Riordan, “­­A New Set of Indicators of Reserve Ampleness,” Federal Reserve Bank of New York Liberty Street Economics, August 14, 2024, https://libertystreeteconomics.newyorkfed.org/2024/08/a-new-set-of-indicators-of-reserve-ampleness/.

You may also be interested in:

When Are Central Bank Reserves Ample?  

How Abundant Are Reserves? Evidence from the Wholesale Payment System

Measuring the Ampleness of Reserves

How the Federal Reserve’s Monetary Policy Implementation Framework Has Evolved

Shortly after the collapse of Silicon Valley Bank (SVB) in March 2023, a consortium of eleven large U.S. financial institutions deposited $30 billion into First Republic Bank to bolster its liquidity and assuage panic among uninsured depositors. In the end, however, First Republic Bank did not survive, raising the question of whether a reallocation of liquidity among financial institutions can ever reduce the need for central bank balance sheet expansion in the fight against bank runs. We explore this question in this post, based on a recent working paper.

Our Model

The theoretical laboratory for questions about bank runs is the Nobel‑winning model of Diamond and Dybvig (1983), where a representative bank is susceptible to runs by patient depositors who do not need to withdraw early but may choose to do so. We add to this model many banks with differing cash positions and an interbank market where banks can lend to each other to study the scope for effective liquidity reallocation in the face of depositor withdrawals.

As is common in bank-run models, the actions of patient depositors depend on their beliefs about what other patient depositors will do. The original Diamond-Dybvig framework has two equilibria: one where the bank fails because patient depositors believe that enough other patient depositors will withdraw early and one where the bank survives because patient depositors believe that enough other patient depositors will not withdraw early.

Our model with multiple banks admits an interesting equilibrium in between these two cases, where some but not all banks fail. If an individual bank can borrow enough on the interbank market to honor withdrawals by all its depositors, then the bank is run-proof and its patient depositors should not withdraw early, regardless of what other patient depositors do. This motivates us to consider as a benchmark a “conservative equilibrium” where patient depositors believe that all other patient depositors will withdraw early if and only if the bank is not run-proof at the prevailing interest rate for interbank loans.

The Case for Intervention

The interbank rate plays a critical role in the model outlined above. Specifically, the rate is critical to how many banks fail and how many survive because it affects depositor assessment of individual bank solvency. If total liquidity in the system is high relative to the needs of impatient depositors who experience shocks and must withdraw early, then supply and demand forces in the interbank market deliver an interest rate that is low enough for all banks to be run-proof. As the needs of impatient depositors rise relative to total liquidity, the interest rate in the interbank market also rises. The higher the interbank rate, the more expensive it is to obtain additional liquidity to honor early withdrawals by patient depositors. The marginal bank that could withstand a run by borrowing on the interbank market can no longer do so profitably, as the amount it needs to borrow is too high to be fully repaid at the higher interest rate. The minimum level of initial cash that a bank must have to be run-proof thus rises, as does the number of failing banks.

This relationship between the interbank rate and the number of bank failures when liquidity conditions are tight implies an important externality. If a bank were to lend (more) on the interbank market, the interest rate would fall and increase the number of run-proof banks. This positive effect is not internalized by a bank choosing how much to lend, leading to too many bank failures compared to what could be achieved with more interbank lending.

A Blueprint for Intervention

Understanding that the problem is insufficient lending among banks, a central planner would design a system of taxes and transfers that pulls liquidity away from some banks and redistributes it to others. The attempt at liquidity reallocation by the consortium of large U.S. financial institutions in March 2023 had a similar flavor. Among these institutions were banks that had likely experienced inflows from depositors running from SVB, so cash was effectively being redistributed from SVB to First Republic Bank. However, this redistribution occurred slowly and unsystematically. What exactly does the planner’s more effective system look like? Under certain conditions, we find that it mimics a system of centrally allocated IOUs (“I Owe Yous”) that overrides the decentralized market.

To fix ideas, suppose the planner allows some banks to issue IOUs to other banks and dictates that these IOUs must be accepted when issued. The planner only allows IOUs to be issued by banks that can ultimately repay them in cash, with the planner also setting the interest rate for repayment. The IOUs cannot be used to honor withdrawals away from the banking system, which are the withdrawals implicit in Diamond-Dybvig and hence the withdrawals we have focused on so far. Instead, the IOUs can be used to honor withdrawals out of one bank and into another, which are additional withdrawals that can exist in an environment with many banks such as ours. These “within-system” withdrawals arise most simply from checks written by depositors at one bank to depositors at another, but they could also reflect transfers between banks from the settlement of derivatives contracts.

A simple example illustrates how these IOUs reallocate liquidity. Imagine that depositors at Bank A write $100 worth of checks to depositors at Bank B. Instead of delivering cash, Bank A can issue $100 in IOUs to Bank B, deferring cash settlement to a later date. If the interest rate on IOUs is 5 percent, then Bank A will pay Bank B $105 in cash at this later date. By allowing Bank A to issue IOUs, the planner has elicited a loan of $100 from Bank B, which may not have occurred in the decentralized market because of the externality described earlier.

We demonstrate that IOUs can be allocated to a subset of banks to achieve a better outcome than a decentralized market when within-system withdrawals are large in volume and generate most of the cross-sectional variation in banks’ cash positions before depositors begin to withdraw away from the system. By eliciting more lending from some banks, the planner can set a lower interest rate than what prevails in the decentralized market and thus make more banks run-proof. While the banks that are taxed in this arrangement are incrementally worse off, the unconditional probability of bank survival is discretely higher, so banks are better off in expectation and willing to opt into the planner’s arrangement ex ante.

A Historical Precedent

At first glance, translating these IOUs from theory to practice appears challenging. However, a close examination of the historical record reveals that an instrument with many of the same features was used by the New York Clearinghouse (NYCH) before the creation of the Federal Reserve.

The NYCH was an association of all major banks in New York City. Its primary function was to facilitate the check-clearing process, but during banking panics, it became the de facto leader in liquidity management absent a central bank. The NYCH deployed a system of loan certificates among its members at the onset of the Panic of 1873, the first major panic of the National Banking Era. These certificates resemble the IOUs in our model: They were allocated to member banks by the NYCH; allocations were limited by the amount of collateral a bank could pledge to ensure ultimate repayment; they could only be used to defer check-clearing obligations with other members; they could not be refused as interim payment for such obligations; and the interest rate at which they had to be repaid was set by the NYCH.

Calibrating our model to historical data allows us to tease out the value of these loan certificates during the Panic of 1873. Controlling for other interventions by the NYCH, we find that liquidity conditions were tight enough for loan certificates to increase social welfare by 2 percent relative to a decentralized interbank market. A 2 percent increase is notable, as it fills roughly half the gap between the decentralized equilibrium and the “first-best” level of welfare—which is an upper bound on the welfare that any policy, including a liquidity injection by a central bank, could hope to achieve.

Conclusion

In times of major banking distress, central bank liquidity injections are the first line of defense for restoring financial stability. However, our analysis suggests that a system of centrally allocated IOUs—not unlike the loan certificates deployed in New York City during the Panic of 1873—could reduce the size of the required injection by implementing a more efficient allocation of liquidity across banks and increasing bank survival rates.

Kinda Hachem is a financial research advisor in Macrofinance Studies in the Federal Reserve Bank of New York’s Research and Statistics Group. 

How to cite this post:
Kinda Hachem, “Reallocating Liquidity to Resolve a Crisis,” Federal Reserve Bank of New York Liberty Street Economics, August 12, 2024, https://libertystreeteconomics.newyorkfed.org/2024/08/reallocating-liquidity-to-resolve-a-crisis/.

You may also be interested in:

Crisis Chronicles: The Long Depression and the Panic of 1873

Can Discount Window Stigma Be Cured? 

Deposits and the March 2023 Banking Crisis—A Retrospective