-
Working Paper
05.30.2023 |
WP 22-23R
Theory and extant empirical evidence suggest that the cross-sectional asymmetry across disaggregated price indexes might be useful in forecasting aggregate inflation. Trimmed-mean inflation estimators have been shown to be useful devices for forecasting headline PCE inflation. But is this because they signal the underlying trend or because they implicitly signal asymmetry in the underlying distribution? We address this question by augmenting a "hard" to beat benchmark headline PCE inflation forecasting model with robust trimmed-mean inflation measures and robust measures of the cross-sectional skewness, both computed using the 180+ components of the PCE price index. Our results indicate significant gains in the point and density accuracy of PCE inflation forecasts over medium- and longer-term horizons, up through and including the COVID-19 pandemic. Improvements in accuracy stem mainly from the trend information implicit in trimmed-mean estimators, but skewness information is also useful. An examination of goods and services PCE inflation provides similar inference.
-
Working Paper
01.13.2023 |
WP 23-06
In the December 2022 Summary of Economic Projections (SEP), the median projection for four-quarter core PCE inflation in the fourth quarter of 2025 is 2.1 percent. This same SEP has unemployment rising by nine-tenths, to 4.6 percent, by the end of 2023. We assess the plausibility of this projection using a specific nonlinear model that embeds an empirically successful nonlinear Phillips curve specification into a structural model, identifying it via an underutilized data-dependent method. We model core PCE inflation using three components that align with those noted by Chair Powell in his December 14, 2022, press conference: housing, core goods, and core-services-less-housing. Our model projects that conditional on the SEP unemployment rate path and a rapid deceleration of core goods prices, core PCE inflation moderates to only 2.75 percent by the end of 2025: inflation will be higher for longer. A deep recession would be necessary to achieve the SEP’s projected inflation path. A simple reduced-form welfare analysis, which abstracts from any danger of inflation expectations becoming unanchored, suggests that such a recession would not be optimal.
-
Working Paper
01.09.2023 |
WP 23-03
What drove inflation so high in 2022? Can it drop rapidly without a recession? The Phillips curve is central to the answers; its proper (nonlinear) specification reveals that the relationship is strong and frequency dependent, and inflation is very persistent. We embed this empirically successful Phillips curve – incorporating a supply-shocks variable – into a structural model. Identification is achieved using an underutilized data-dependent method. Despite imposing anchored inflation expectations and a rapid relaxation of supply-chain problems, we find that absent a recession, inflation will be more than 3 percent by the end of 2025. A simple welfare analysis supports a mild recession as preferred to an extended period of elevated inflation, under a typical loss function.
-
Working Paper
08.15.2022 |
WP 21-23R
We develop a flexible semi-structural time-series model to estimate jointly several macroeconomic "stars" -- i.e., unobserved long-run equilibrium levels of output (and growth rate of output), the unemployment rate, the real rate of interest, productivity growth, price inflation, and wage inflation. The ingredients of the model are in part motivated by economic theory and in part by the empirical features necessitated by the changing economic environment. Following the recent literature on inflation and interest rate modeling, we explicitly model the links between long-run survey expectations and stars to improve the stars' econometric estimation. Our approach permits time variation in the relationships between various components, including time variation in error variances. To tractably estimate the large multivariate model, we use a recently developed precision sampler that relies on Bayesian methods. The by-products of this approach are the time-varying estimates of the wage and price Phillips curves, and the pass-through between prices and wages, both of which provide new insights into these empirical relationships' instability in US data. Generally, the contours of the stars echo those documented elsewhere in the literature -- estimated using smaller models -- but at times the estimates of stars are different, and these differences can matter for policy. Furthermore, our estimates of the stars are among the most precise. Last, we document the competitive real-time forecasting properties of the model and, separately, the usefulness of stars' estimates as steady-state values in external models.
-
Working Paper
08.03.2022 |
WP 22-23
Both theory and extant empirical evidence suggest that the cross-sectional asymmetry across disaggregated price indexes might be useful in the forecasting of aggregate inflation. Trimmed-mean inflation estimators have been shown to be useful devices for forecasting headline PCE inflation. But does this stem from their ability to signal the underlying trend, or does it mainly come from their implicit signaling of asymmetry (when included alongside headline PCE)? We address this question by augmenting a “hard to beat” benchmark inflation forecasting model of headline PCE price inflation with robust measures of trimmed-mean estimators of inflation (median PCE and trimmed-mean PCE) and robust measures of the cross-sectional asymmetry (Bowley skewness; Kelly skewness) computed using the 180+ components of the PCE price index. We also construct new trimmed-mean measures of goods and services PCE inflation and their accompanying robust skewness. Our results indicate significant gains in the point and density accuracy of PCE inflation forecasts over medium- and longer-term horizons, up through and including the COVID-19 pandemic. We find that improvements in accuracy stem mainly from the trend information implicit in trimmed-mean estimators, but that skewness is also useful. Median PCE slightly outperforms trimmed-mean PCE; both outperform core PCE. For point forecasts, Kelly skewness is preferred; but for estimating stochastic volatility, Bowley skewness is preferred. An examination of goods and services PCE inflation provides similar inference.
-
Working Paper
10.14.2021 |
WP 21-23
We develop a flexible semi-structural time-series model to estimate jointly several macroeconomic "stars" — i.e., unobserved long-run equilibrium levels of output (and growth rate of output), the unemployment rate, the real rate of interest, productivity growth, the price inflation, and wage inflation. The ingredients of the model are in part motivated by economic theory and in part by the empirical features necessitated by the changing economic environment. Following the recent literature on inflation and interest rate modeling, we explicitly model the links between long-run survey expectations and stars to improve the stars' econometric estimation. Our approach permits time variation in the relationships between various components, including time variation in error variances. To tractably estimate the large multivariate model, we use a recently developed precision sampler that relies on Bayesian methods. The by-products of this approach are the time-varying estimates of the wage and price Phillips curves, and the pass-through between prices and wages, both of which provide new insights into these empirical relationships' instability in US data. Generally, the contours of the stars echo those documented elsewhere in the literature — estimated using smaller models — but at times the estimates of stars are different, and these differences can matter for policy. Furthermore, our estimates of the stars are among the most precise. Lastly, we document the competitive real-time forecasting properties of the model and, separately, the usefulness of stars' estimates if they were used as steady-state values in external models.
-
Working Paper
10.22.2020 |
WP 20-31
We develop a flexible modeling framework to produce density nowcasts for US inflation at a trading-day frequency. Our framework: (1) combines individual density nowcasts from three classes of parsimonious mixed-frequency models; (2) adopts a novel flexible treatment in the use of the aggregation function; and (3) permits dynamic model averaging via the use of weights that are updated based on learning from past performance. Together these features provide density nowcasts that can accommodate non-Gaussian properties. We document the competitive properties of the nowcasts generated from our framework using high-frequency real-time data over the period 2000-2015.
-
Working Paper
06.16.2020 |
WP 20-17
We document asymmetric responses of consumer spending to energy price shocks: Using a multiple-regime threshold vector autoregressive model estimated with Bayesian methods on US data, we find that positive energy price shocks have a larger negative effect on consumption compared with the increase in consumption in response to negative energy price shocks. For large shocks, the cumulative consumption responses are three to five times larger for positive than for negative shocks. Digging into disaggregated spending, we find that the estimated asymmetric responses are strongest for durable goods, but asymmetries are also present in the responses of nondurables and services.
-
Working Paper
06.22.2018 |
WP 18-09
This paper constructs hybrid forecasts that combine both short- and long-term conditioning information from external surveys with forecasts from a standard fixed-coefficient vector autoregression (VAR) model. Specifically, we use relative entropy to tilt one-step ahead and long-horizon VAR forecasts to match the nowcast and long-horizon forecast from the Survey of Professional Forecasters. The results indicate meaningful gains in multi-horizon forecast accuracy relative to model forecasts that do not incorporate long-term survey conditions. The accuracy gains are achieved for a range of variables, including those that are not directly tilted but are affected through spillover effects from tilted variables. The forecast accuracy gains for inflation are substantial, statistically significant, and are competitive with the forecast accuracy from both time-varying VARs and univariate benchmarks. We view our proposal as an indirect approach to accommodating structural change and moving end points.
-
Working Paper
03.17.2017 |
WP 17-02
Financial data often contain information that is helpful for macroeconomic forecasting, while multistep forecast accuracy also benefits by incorporating good nowcasts of macroeconomic variables. This paper considers the role of nowcasts of financial variables in making conditional forecasts of real and nominal macroeconomic variables using standard quarterly Bayesian vector autoregressions (BVARs). For nowcasting the quarterly value of a variety of financial variables, we document that the average of the available daily data and a daily random walk forecast to fill in the missing days in the quarter typically outperforms other nowcasting approaches. Using real-time data and out-of-sample forecasting exercises, we find that the inclusion of financial variable nowcasts by themselves generally improves forecast accuracy for macroeconomic variables relative to unconditional forecasts, although we document several exceptions in which current-quarter forecast accuracy worsens with the inclusion of the financial nowcasts. Incorporating financial nowcasts and nowcasts of macroeconomic variables generally improves the forecast accuracy for all the macroeconomic indicators of interest, beyond including the nowcasts of the macroeconomic variables alone. Conditional forecasts generated from quarterly BVARs augmented with nowcasts of key financial variables rival the forecast accuracy of mixed-frequency dynamic factor models (MF-DFMs) and mixed-data sampling (MIDAS) models that explicitly link the quarterly data and forecasts to high-frequency financial data.
-
Working Paper
10.31.2016 |
WP 13-03R
In this paper we investigate the forecasting performance of the median consumer price index (CPI) in a variety of Bayesian vector autoregressions (BVARs) that are often used for monetary policy. Until now, the use of trimmed-mean price statistics in forecasting inflation has often been relegated to simple univariate or "Phillips-curve" approaches, thus limiting their usefulness in applications that require consistent forecasts of multiple macro variables. We find that inclusion of an extreme trimmed-mean measure—the median CPI—improves the forecasts of both core and headline inflation (CPI and PCE) across our set of monthly and quarterly BVARs. While the inflation forecasting improvements are perhaps not surprising given the current literature on core inflation statistics, we also find that inclusion of the median CPI improves the forecasting accuracy of the central bank's primary instrument for monetary policy—the federal funds rate. We conclude with a few illustrative exercises that highlight the usefulness of using the median CPI.
-
Working Paper
09.29.2016 |
WP 15-19R
We estimate an empirical model of inflation that exploits a Phillips curve relationship between a measure of unemployment and a subaggregate measure of inflation (services). We generate an aggregate inflation forecast from forecasts of the goods subcomponent separate from the services subcomponent, and compare the aggregated forecast to the leading time-series univariate and standard Phillips curve forecasting models. Our results indicate marked improvements in point and density forecasting accuracy statistics for models that exploit relationships between services inflation and the unemployment rate.
-
Working Paper
11.04.2015 |
WP 14-03R
Forecasting future inflation and nowcasting contemporaneous inflation are difficult. We propose a new and parsimonious model for nowcasting headline and core inflation in the U.S. consumer price index (CPI) and price index for personal consumption expenditures (PCE) that relies on relatively few variables. The model's nowcasting accuracy improves as information accumulates over a month or quarter, outperforming statistical benchmarks. In real-time comparisons, the model's headline inflation nowcasts substantially outperform those from the Blue Chip consensus and the Survey of Professional Forecasters. Across all four inflation measures, the model's nowcasting accuracy is comparable to the Federal Reserve Board's Greenbook.
-
Working Paper
10.14.2015 |
WP 15-19
We estimate an empirical model of inflation that exploits a Phillips curve relationship between a measure of unemployment and a subaggregate measure of inflation (services). We generate an aggregate inflation forecast from forecasts of the goods subcomponent separate from the services subcomponent, and compare the aggregated forecast to the leading time-series univariate and standard Phillips curve forecasting models. Our results indicate notable improvements in forecasting accuracy statistics for models that exploit relationships between services inflation and the unemployment rate. In addition, models of services inflation using the short-term unemployment rate (less than 27 weeks) as the real economic indicator display additional modest forecast accuracy improvements.
-
Working Paper
09.16.2015 |
WP 15-15
We examine how a combination of credit market and asset quality information can jointly be used in assessing bank franchise value. We find that expectations of future credit demand and future asset quality explain contemporaneous bank franchise value, indicative of the feedback in credit market information and its consequent impact on bank franchise value.
-
Working Paper
10.11.2014 |
WP 13-13R
This paper makes a fundamental contribution by studying loan-loss provisioning over the credit cycle as three distinct phases. Looking at the three distinct phases of the financial crisis—the precrisis period, crisis period, and postcrisis period—is important as loan-loss provisioning is driven by different factors in each, in part due to extensive shifts in (or in the application of) regulatory rule. We show evidence of forward-looking loan-loss provisioning by utilizing Senior Loan Officer Opinion Surveys (SLOOS), which provide useful controls for credit cycle information. Though the SLOOS data set is a restricted sample and generalizability to a broader sample could potentially be a stretch, we control for credit cycle factors as part of an identification strategy to sort out changes in the credit market equilibrium. We contribute to the growing literature on forward-looking loan-loss provisioning and early-in-the-cycle loss recognition by incorporating a broader range of available credit information.
-
Working Paper
05.20.2014 |
WP 14-03
Forecasting future inflation and nowcasting contemporaneous inflation are difficult. We propose a new and parsimonious model for nowcasting headline and core inflation in the U.S. price index for personal consumption expenditures (PCE) and the consumer price index (CPI). The model relies on relatively few variables and is tested using real-time data. The model’s nowcasting accuracy improves as information accumulates over the course of a month or quarter, and it easily outperforms a variety of statistical benchmarks. In head-to-head comparisons, the model’s nowcasts of CPI inflation outperform those from the Blue Chip consensus, with especially significant outperformance as the quarter goes on. The model’s nowcasts for CPI and PCE inflation also significantly outperform those from the Survey of Professional Forecasters, with similar nowcasting accuracy for core inflation measures. Across all four inflation measures, the model’s nowcasting accuracy is generally comparable to that of the Federal Reserve’s Greenbook.
-
Working Paper
09.11.2013 |
WP 13-13
This paper makes a fundamental contribution by studying loan-loss provisioning over the credit cycle as three distinct phases. Looking at the three distinct phases of the financial crisis—the precrisis period, crisis period, and postcrisis period—is important as loan-loss provisioning is driven by different factors in each, in part due to extensive shifts in (or in the application of) regulatory rule. We show evidence of forward-looking loan-loss provisioning by utilizing Senior Loan Officer Opinion Surveys (SLOOS), which provide useful controls for credit cycle information. Though the SLOOS data set is a restricted sample and generalizability to a broader sample could potentially be a stretch, we control for credit cycle factors as part of an identification strategy to sort out changes in the credit market equilibrium. We contribute to the growing literature on forward-looking loan-loss provisioning and early-in-the-cycle loss recognition by incorporating a broader range of available credit information.
-
Working Paper
02.26.2013 |
WP 13-03
In this paper we investigate the forecasting performance of the median CPI in a variety of Bayesian VARs (BVARs) that are often used for monetary policy. Until now, the use of trimmed-mean price statistics in forecasting inflation has often been relegated to simple univariate or "Philips-curve" approaches, thus limiting their usefulness in applications that require consistent forecasts of multiple macro variables. We find that inclusion of an extreme trimmed-mean measure—the median CPI—significantly improves the forecasts of both headline and core CPI across our wide-ranging set of BVARs. While the inflation forecasting improvements are perhaps not surprising given the current literature on core inflation statistics, we also find that inclusion of the median CPI improves the forecasting accuracy of the central bank's primary instrument for monetary policy—the federal funds rate. We conclude with a few illustrative exercises that highlight the usefulness of using the median CPI.
-
Working Paper
10.27.2011 |
WP 11-28
This paper presents a 16-variable Bayesian VAR forecasting model of the U.S. economy for use in a monetary policy setting. The variables that comprise the model are selected not only for their effectiveness in forecasting the primary variables of interest, but also for their relevance to the monetary policy process. In particular, the variables largely coincide with those of an augmented New-Keynesian DSGE model. We provide out-of sample forecast evaluations and illustrate the computation and use of predictive densities and fan charts. Although the reduced form model is the focus of the paper, we also provide an example of structural analysis to illustrate the macroeconomic response of a monetary policy shock.