Temporal Aggregation Bias Research Articles

Temporal Aggregation Bias and Monetary Policy Transmission

Temporal aggregation biases estimates of monetary policy effects. We hypothesize that information mismatches between private agents and the econometrician—the source of temporal aggregationbias—are as important as the more studied mismatch between private agents and the centralbank (the “Fed information effect”) in the study of monetary policy transmission. In impulse responsesfrom both local projections and an unobserved components model, we find that the responseof daily inflation to high-frequency monetary shocks confirms theoretical predictions. If thereis an adverse-signed response such that inflation increases in response to a contractionary monetaryshock, it is much less prominent than previously thought and explained by frequency mismatches ofshocks and dependent variables.

Inflation Measured Every Day Keeps Adverse Responses Away: Temporal Aggregation and Monetary Policy Transmission

Using daily inflation data from the Billion Prices Project [Cavallo and Rigobon (2016)], we show how temporal aggregation biases estimates of monetary policy transmission. We argue that the information mismatch between private agents and the econometrician —the source of temporal aggregation bias —is equally important as the more studied mismatch between private agents and the central bank (the “Fed information effect”). We find that the adverse response of daily inflation to high-frequency monetary policy shocks is short-lived, if present at all, in impulse responses from both local projections and an unobserved components model of inflation dynamics. To reconcile how one can obtain a sizable adverse response with monthly or quarterly data when only a limited adverse response exists at a higher frequency, we appeal to a simple monetary policy model and show how temporal aggregation bias can exacerbate initial impulse response functions. Because our modeling results are generic and macroeconomic indicators are published with a lag, we argue that temporal aggregation bias will be a key feature of the nascent field of high-frequency macroeconomics.

Temporal aggregation bias and Gerrymandering urban time series

Temporal aggregation bias and Gerrymandering urban time series

Identifying high-frequency shocks with Bayesian mixed-frequency VARs

We contribute to research on mixed-frequency regressions by introducing an innovative Bayesian approach. We impose a Normal-inverse Wishart prior by adding a set of auxiliary dummies in estimating a Mixed-Frequency VAR. Based on this new “high-frequency” identification scheme, we illustrate our method by identifying uncertainty shock for the U.S. economy. As the main findings, we document a “temporal aggregation bias” when we adopt a common low-frequency model instead of estimating a mixed-frequency framework. The bias is amplified in case of a large mismatching between the high-frequency shock and low-frequency business cycle variables.

Identifying High-Frequency Shocks with Bayesian Mixed-Frequency VARs

We contribute to research on mixed-frequency regressions by introducing an innovative Bayesian approach. Based on a new “high-frequency” identification scheme, we provide novel empirical evidence of identifying uncertainty shock for the US economy. As main findings, we document a “temporal aggregation bias” when we adopt a common low frequency model instead of estimating a mixed-frequency framework. The bias is amplified when we identify a higher frequency shock.

Financial distress and real economic activity in Lithuania: a Granger causality test based on mixed-frequency VAR

In this paper, we extend the monthly financial stress index for Lithuania, computed by the European Central Bank, to a daily frequency and we also include banking sector stress among its constituents, beyond bond, equity and foreign exchange markets. We investigate the causal relationship between the daily financial stress index and monthly industrial production growth, using a Granger causality test applied to a mixed-frequency VAR. Our results suggest evidence of Granger causality from financial stress to industrial production growth once the index is enriched by daily observations from the financial markets. Our findings, based on impulse response analysis, confirm the negative effect of financial stress on real economy found in the empirical literature through common frequency analysis. Finally, the comparison between common and mixed frequency analysis suggests that ignoring the information content of daily data would lead to a mild temporal aggregation bias that could affect the evaluation of financial stress shocks on industrial production.

What are the macroeconomic effects of high‐frequency uncertainty shocks?

SummaryThis paper evaluates the effects of high‐frequency uncertainty shocks on a set of low‐frequency macroeconomic variables representative of the US economy. Rather than estimating models at the same common low frequency, we use recently developed econometric models, which allow us to deal with data of different sampling frequencies. We find that credit and labor market variables react the most to uncertainty shocks in that they exhibit a prolonged negative response to such shocks. When looking at detailed investment subcategories, our estimates suggest that the most irreversible investment projects are the most affected by uncertainty shocks. We also find that the responses of macroeconomic variables to uncertainty shocks are relatively similar across single‐frequency and mixed‐frequency data models, suggesting that the temporal aggregation bias is not acute in this context.

Randomness or stock–flow: which mechanism describes labour market matching in Poland?

ABSTRACTI identify which theoretical model (random, stock–flow, or job queuing) best describes the matching mechanism in the labour market in Poland. The purpose of this work is to formulate policy recommendations aimed at increasing the number of matches. I use monthly registered unemployment data for the period January 1999–June 2013 and econometrically correct for temporal aggregation bias in the data. I extend known solutions and apply them directly to a job queuing model. Job seekers (from the pool) seek work among old and new job posts, but only a small fraction of the newly unemployed individuals find work quickly. Vacancies are the driving force in aggregate hiring, but the inflow is more important than the stock. The random model has greater explanatory power, although the results do not negate the non-random model. Hence, better information and higher inflows (especially of job offers) should facilitate matching.

Time Interval Bias: Its Impact on Advertising Budgeting

Abstract This paper considers what happens when two key parameters in advertising budgeting models (immediate and lagged effects of advertising) are estimated without recognition of time interval (or temporal aggregation) bias. It is demonstrated that careful attention should be paid to whether annual, biannual or quarterly advertising-sales data are used in estimating these key budgeting parameters, otherwise the measurement bias in the key budgeting parameters is likely to be severe. Most importantly, this measurement bias is shown to impact substantially the managerial accuracy and usefulness of dynamic advertising budgeting models.

Sticky prices and deviations from the Law of One Price: Evidence from Mexican micro-price data

This paper uses a novel dataset of weekly sampled store-level prices to study the impact of sticky prices on the volatility and persistence of intranational deviations from the Law of One Price (LOP). The volatility of LOP deviations is found to be increasing both on the distance separating two locations and on the degree of price stickiness. Sticky prices are also found to be systematically related to the persistence of LOP deviations: Half-lives of LOP deviations are systematically larger for goods with stickier prices. These observations are shown to accord well with the predictions of a dynamic general equilibrium model featuring real market segmentation and Calvo pricing. I also find evidence of remarkably fast convergence to the LOP (average half-lives of LOP deviations are in the order of 3–6weeks) and show that previously reported convergence estimates may be afflicted by a positive temporal aggregation bias.

Mixed Frequency Structural Vector Auto-Regressive Models

Summary A mismatch between the timescale of a structural vector auto-regressive model and that of the time series data used for its estimation can have serious consequences for identification, estimation and interpretation of the impulse response functions. However, the use of mixed frequency data, combined with a proper estimation approach, can alleviate the temporal aggregation bias, mitigate the identification issues and yield more reliable responses to shocks. The problems and possible remedy are illustrated analytically and with both simulated and actual data.

Mixed Frequency Structural VARs

A mismatch between the time scale of a structural VAR (SVAR) model and that of the time series data used for its estimation can have serious consequences for identification, estimation and interpretation of the impulse response functions. However, the use of mixed frequency data, combined with a proper estimation approach, can alleviate the temporal aggregation bias, mitigate the identification issues, and yield more reliable responses to shocks. The problems and possible remedy are illustrated analytically and with both simulated and actual data.

Mixed Frequency Structural Models: Identification, Estimation, and Policy Analysis

In this paper we show analytically, with simulation experiments and with actual data that a mismatch between the time scale of a DSGE model and that of the time series data used for its estimation generally creates identification problems, introduces estimation bias and distorts the results of policy analysis. On the constructive side, we prove that the use of mixed frequency data, combined with a proper estimation approach, can alleviate the temporal aggregation bias, mitigate the identification issues, and yield more reliable policy conclusions. The problems and possible remedy are illustrated in the context of standard structural monetary policy models.

Temporal aggregation and purchasing power parity persistence

This paper uses a unique new monthly US–UK real exchange rate series for the January 1794–December 2009 period to reexamine the academic debate over purchasing power parity (PPP). The consensus view described by Rogoff (1996) is that PPP holds in the long-run, but short-run deviations are very persistent, with half-lives ranging from 3 to 5 years. Most of the literature using long time series relies on the annual data developed by Lee (1976) and Lothian and Taylor (1996), which were both constructed from underlying higher-frequency data sources. Estimates of purchasing power parity persistence using these series may therefore be subject to temporal aggregation bias. We find evidence of aggregation bias which indicates the half-life of PPP deviations has been overestimated in much of the previous literature. We also find that estimates of the half-lives are further reduced once we account for the Harrod (1933)– Balassa (1964)– Samuelson (1964) effect. The result of aggregation bias appears to be robust even when considering the case that real exchange rates exhibit nonlinear dynamics.

Seasonal Hydrologic Forecasting: Do Multimodel Ensemble Averages Always Yield Improvements in Forecast Skill?

Abstract Multimodel techniques have proven useful in improving forecast skill in many applications, including hydrology. Seasonal hydrologic forecasting in large basins represents a special case of hydrologic modeling, in which postprocessing techniques such as temporal aggregation and time-varying bias correction are often employed to improve forecast skill. To investigate the effects that these techniques have on the performance of multimodel averaging, the performance of three hydrological models [Variable Infiltration Capacity, Sacramento/Snow-17, and the Noah land surface model] and two multimodel averages [simple model average (SMA) and multiple linear regression (MLR) with monthly varying model weights] are examined in three snowmelt-dominated basins in the western United States. These evaluations were performed for both simulating and forecasting [using the Ensemble Streamflow Prediction (ESP) method] monthly discharge, with and without monthly bias corrections. The single best bias-corrected model outperformed the multimodel averages of raw models in both retrospective simulations and ensemble mean forecasts in terms of RMSE. Forming an MLR multimodel average from bias-corrected models added only slight improvements over the best bias-corrected model. Differences in performance among all bias-corrected models and multimodel averages were small. For ESP forecasts, both bias correction and multimodel averaging generally reduced the RMSE of the ESP ensemble means at lead times of up to 6 months in months when flow is dominated by snowmelt, with the reduction increasing as lead time decreased. The primary reason for this is that aggregating simulated streamflows from daily to monthly time scales increases model cross correlation, which in turn reduces the effectiveness of multimodel averaging in reducing those components of model error that bias correction cannot address. This effect may be stronger in snowmelt-dominated basins because the interannual variability of winter precipitation is a common input to all models. It was also found that both bias correcting and multimodel averaging using monthly varying parameters yielded much greater error reductions than methods using time-invariant parameters.

A TEST BETWEEN STOCK‐FLOW MATCHING AND THE RANDOM MATCHING FUNCTION APPROACH*

This article tests between the standard “random matching function” approach and “stock‐flow” matching while controlling for temporal aggregation bias. Consistent with previous empirical work, the random matching function fits the matching data reasonably well. But match flows are more highly correlated with vacancy inflows than is consistent with the random matching approach. Instead the data support stock‐flow matching, where unemployed workers match directly with suitable new vacancies as such vacancies come on to the market.

Temporal Aggregation Bias and Inference of Causal Regulatory Networks

Temporal Aggregation Bias and Inference of Causal Regulatory Networks

Temporal Aggregation Bias and Inference of Causal Regulatory Networks

Time course experiments with microarrays have begun to provide a glimpse into the dynamic behavior of gene expression. In a typical experiment, scientists use microarrays to measure the abundance of mRNA at discrete time points after the onset of a stimulus. Recently, there has been much work on using these data to infer causal regulatory networks that model how genes influence each other. However, microarray studies typically have slow sampling rates that can lead to temporal aggregation of the signal. That is, each successive sampling point represents the sum of all signal changes since the previous sample. In this paper, we show that temporal aggregation can bias algorithms for causal inference and lead them to discover spurious relations that would not be found if the signal were sampled at a much faster rate. We discuss the implications of temporal aggregation on inference, the problems it creates, and potential directions for solutions.

Temporal aggregation and structural inference in macroeconomics

This paper examines the quantitative importance of temporal aggregation bias in distorting parameter estimates and hypothesis tests. Our strategy is to consider two empirical examples in which temporal aggregation bias has the potential to account for results which are widely viewed as being anomalous from the perspective of particular economic models. Our first example investigates the possibility that temporal aggregation bias can lead to spurious Granger causality relationships. The quantitative importance of this possibility is examined in the context of Granger causal relations between the growth rates of money and various measures of aggregate output. Our second example investigates the possibility that temporal aggregation bias can account for the slow speeds of adjustment typically obtained with stock adjustment models. The quantitative importance of this possibility is examined in the context of a particular class of continuous and discrete time equilibrium models of inventories and sales. The different models are compared on the basis of the behavioral implications of the estimated values of the structural parameters which we obtain and their overall statistical performance. The empirical results from both examples provide support for the view that temporal aggregation bias can be quantitatively important in the sense of significantly distorting inference. (This abstract was borrowed from another version of this item.)