Longer Forecast Horizons Research Articles

Using the Entire Yield Curve in Forecasting Output and Inflation

In forecasting a variable (forecast target) using many predictors, a factor model with principal components (PC) is often used. When the predictors are the yield curve (a set of many yields), the Nelson–Siegel (NS) factor model is used in place of the PC factors. These PC or NS factors are combining information (CI) in the predictors (yields). However, these CI factors are not “supervised” for a specific forecast target in that they are constructed by using only the predictors but not using a particular forecast target. In order to “supervise” factors for a forecast target, we follow Chan et al. (1999) and Stock and Watson (2004) to compute PC or NS factors of many forecasts (not of the predictors), with each of the many forecasts being computed using one predictor at a time. These PC or NS factors of forecasts are combining forecasts (CF). The CF factors are supervised for a specific forecast target. We demonstrate the advantage of the supervised CF factor models over the unsupervised CI factor models via simple numerical examples and Monte Carlo simulation. In out-of-sample forecasting of monthly US output growth and inflation, it is found that the CF factor models outperform the CI factor models especially at longer forecast horizons.

Macroeconomic determinants of the term structure: Long-run and short-run dynamics

Abstract We propose a no-arbitrage term structure model with a Taylor rule and two macroeconomic variables, real activity growth and inflation, that each contain long-run and short-run components. Variance decompositions indicate that the impact of macroeconomic variables on the term structure differs from existing models. For short maturities, inflation is relatively more important than real activity growth at short forecast horizons. For longer maturity yields, the long-run component of inflation explains most of the long-horizon forecast variance, but real activity growth matters for short forecast horizons. Unlike existing macro models, the model implies plausible term premia and expectations of short rates. The long-run components also improve the prediction of bond excess returns relative to information in the yield curve and macro variables. Measures of in-sample and out-of-sample fit confirm the benefits of allowing for long- and short-run components.

Does Crowdsourced Research Discipline Sell-Side Analysts?

We examine whether increased competition stemming from technological innovation disciplines sell-side analysts. We document a decline in short-term forecast bias for firms added to Estimize, an open platform that crowdsources short-term earnings forecasts, relative to matched control firms; this decline is greater when (1) existing sell-side competition is smaller, (2) earnings uncertainty is higher, and (3) Estimize coverage is less biased and more accurate. We also document an increase in short-term forecast accuracy and representativeness. Finally, we find no change in bias for longer-horizon forecasts or recommendations, suggesting competition from Estimize rather than broad economic forces drives our results.

Asset Growth and Analysts’ Multi-Period Earnings Forecasts

Using analysts'' multi-period earnings forecasts, we examine whether analyst forecast bias is related to asset growth. We find that analyst forecasts are more optimistic for firms with higher asset growth. This relation is particularly noticeable for longer-term (e.g., two- and three-year-ahead) forecasts than for shorter-term (e.g., one-year-ahead) forecasts. Moreover, analyst optimism for high-growth firms is greater for 1) firms that have maintained similar levels of growth over recent periods, 2) firms with higher information uncertainty, and 3) forecasts with longer forecast horizons (e.g., forecasts issued far before the fiscal year-end). We examine to what extent analyst optimism for high-growth firms explains the asset growth effect (i.e., a negative association between asset growth and subsequent stock returns). Controlling for forecast bias in a growth-return regression substantially attenuates the asset growth effect, suggesting that forecast bias plays an important role in the asset growth effect. Path analysis further suggests that analysts'' long-term (but not short-term) forecast bias is an important mediator through which biased expectations about asset growth are incorporated into stock prices. Overall, our findings are consistent with the extrapolation bias explanation for the asset growth effect.

Bank liquidity creation and recessions

Abstract We investigate the relationship between bank liquidity creation and recessions in the U.S. For the 1984–2010 sample, we find that (i) lower bank on-balance sheet liquidity creation signals recessions four quarters into the future; (ii) off-balance sheet liquidity creation is not a robust predictor of recessions at longer forecast horizons; (iii) off-balance sheet liquidity creation falls in tandem with on-balance sheet liquidity creation one quarter prior to recessions, and aggregate, on- and off-balance sheet bank liquidity creation continue to decline during and up to five quarters after recessions; and (iv) liquidity creation of larger banks contains more information about future recessions than that of smaller ones.

Forecasting and trading high frequency volatility on large indices

The present paper analyses the forecastability and tradability of volatility on the large S&P500 index and the liquid SPY ETF, VIX index and VXX ETN. Even though there is already a huge array of literature on forecasting high frequency volatility, most publications only evaluate the forecast in terms of statistical errors. In practice, this kind of analysis is only a minor indication of the actual economic significance of the forecast that has been developed. For this reason, in our approach, we also include a test of our forecast through trading an appropriate volatility derivative. As a method we use parametric and artificial intelligence models. We also combine these models in order to achieve a hybrid forecast. We report that the results of all three model types are of similar quality. However, we observe that artificial intelligence models are able to achieve these results with a shorter input time frame and the errors are uniformly lower comparing with the parametric one. Similarly, the chosen models do not appear to differ much while the analysis of trading efficiency is performed. Finally, we notice that Sharpe ratios tend to improve for longer forecast horizons.

Combining Survey Long-Run Forecasts and Nowcasts with BVAR Forecasts Using Relative Entropy

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.

Data-driven short-term forecasting of solar irradiance profile

Autoregressive and non-linear models are frequently used for solar irradiance forecasting, and their results are found to vary depending on the forecast horizon, time resolution and location. Given that higher accuracies are often reported at lower time resolutions and areas with less cloud cover, it was found necessary to separately assess the applicability of models for tropical Singapore where high solar irradiance variabilities are experienced.This paper presents a short-range forecasting system of 30-min irradiance averages for 0.5 to 6 hours ahead based on per-min data of solar irradiance and ambient temperature. In addition, it explores the possibility of predicting volatility by looking at the distribution of solar irradiance in the next 30-min period with a novel approach that estimates the proportion of points within each of 21 bands defined to cover the range of irradiance. With it, upper and lower bound predictions for the period are obtained to calculate upside and downside risks posed by photovoltaic (PV) generation. Using persistence models for comparison and assessing accuracy across 8 locations, all models showed marked improvement, especially at longer forecast horizons. On average, MAE of point forecast models decreased by 9% (98 to 89 W/m2) and 58% (299 to 125 W/m2) for the 0.5 and 6-hour horizons respectively. For volatility models, MAE decreased from 4.8 to 3.7% in proportion predictions while errors of making upper and lower bound predictions outside the actual range of per-min fluctuations decreased from 43.0 to 10.4% and 30.6 to 3.4% respectively.

Biases in Analysts' Multiyear Forecasted Income Statements, Balance Sheets, and Cash Flow Statements

We evaluate sell-side equity analysts’ multiyear forecasted income statements, balance sheets and cash flow statements, and the profitability, efficiency and leverage ratios that they imply. Using both small- sample data extracted manually from Investext PDFs, and large-sample data taken from the I/B/E/S non-EPS archival detail history file, we find that analysts’ long-horizon financial statements contain many biases, many but not all of which are optimistic. Analysts make highly optimistic forecasts of long-horizon EPS, ROE, ROA, ROS and asset turnover, driven by overly bullish projections about revenues and all common-sized expenses except income tax, which they forecast pessimistically. Analysts are optimistic about both long-horizon operating cash flows and operating accruals, and while they are unbiased in their forecasts of long-horizon total assets, they underestimate long-horizon debt and overestimate long-horizon equity. Our regressions support the view that analysts strategically inflate their long-horizon forecasts of EPS the more intangible and hard-to-verify are the firm’s assets.

Asset Conversion Cycle and Accruals Persistence

We examine the relationship between the persistence of the accrual component of earnings and asset conversion cycle, i.e., the speed of accrual reversal. We argue that, as all accruals have to ultimately reverse, those reversing at a slower speed will exhibit more persistence. We use the operating cycle and investment cycle to measure the reversal speeds of working capital accruals and noncurrent accruals, respectively. Our results show that the investment cycle is positively associated with the persistence of noncurrent accruals. However, the operating cycle does not affect the persistence of working capital accruals, because by definition these accruals will reverse within one year. In addition, we find that analysts do not fully understand the effects of asset conversion cycles especially for long-horizon forecasts. Collectively, our study suggests that the lower persistence of accruals is not all attributable to estimation errors, and the asset conversion cycle is an important determinant of accrual persistence.

Backtesting Distributional Forecasts in Counterparty Risk II: Empirical Results for Long Horizon Testing

This paper provides empirical results supporting the theoretical ones by the first author on backtesting long-horizon distributional forecasts. The problem is quite general but for us it is motivated by the regulatory requirement of backtesting evolution models used in the measurement of counterparty risk. This is particularly relevant for banks that are capitalized under the IMM (internal models method) regime of the BCBS (Basel Committee on Banking Supervision). The main practical challenge there is that banks are required to backtest (forecasting) models up to horizons of at least one year. Our results give strong support to the proposed method and offers some guidelines for prudent use.

Forecasting Performance of Lumber Futures Prices

We test the forecasting power and information content of lumber futures prices traded on the Chicago Mercantile Exchange, from 1995 to 2013, at four forecast horizons. A Mincer-Zarnowitz regression finds evidence of statistically significant forecasting power at all forecast horizons. The results also support the presence of a time-varying risk premium for the shorter forecast horizons. A Granger causality test provides evidence that lumber futures prices lag spot prices in information assimilation over longer forecast horizons, while neither lagging nor leading over shorter forecast horizons.

Realized volatility forecast of agricultural futures using the HAR models with bagging and combination approaches

In order to reduce the uncertainty associated with a single predictor model, we incorporate the bagging and combination approaches into a HAR model with the lags of realized volatility and other potential predictors to forecast the realized volatility of agricultural commodity futures in China. We evaluate the performances of the two approaches by employing the mean square forecast error (MSFE) loss function, the modified DM test and the model confidence set (MCS) test at the multiple horizons over the three out-of-sample periods. We find that the realized forecasts from the HAR model with bagging and principal component (PC) combination approaches produce the lowest MSFE at relatively longer forecast horizons. We also find that the simple average of the forecasts from the HAR models with bagging and PC combination methods leads to a further reduction in MSFE, suggesting that they are the effective methods to forecast the realized volatility of agricultural commodity futures in China.

Forecasting Daily Political Opinion Polls Using the Fractionally Cointegrated Vector Auto-Regressive Model

SummaryWe examine forecasting performance of the recent fractionally cointegrated vector auto-regressive (FCVAR) model. We use daily polling data of political support in the UK for 2010–2015 and compare with popular competing models at several forecast horizons. Our findings show that the four variants of the FCVAR model considered are generally ranked as the top four models in terms of forecast accuracy, and the FCVAR model significantly outperforms both univariate fractional models and the standard cointegrated vector auto-regressive model at all forecast horizons. The relative forecast improvement is higher at longer forecast horizons, where the root-mean-squared forecast error of the FCVAR model is up to 15% lower than that of the univariate fractional models and up to 20% lower than that of the cointegrated vector auto-regressive model. In an empirical application to the 2015 UK general election, the estimated common stochastic trend from the model follows the vote share of the UK Independence Party very closely, and we thus interpret it as a measure of Euroscepticism in public opinion rather than an indicator of the more traditional left–right political spectrum. In terms of prediction of vote shares in the election, forecasts generated by the FCVAR model leading to the election appear to provide a more informative assessment of the current state of public opinion on electoral support than the hung Parliament prediction of the opinion poll.

Forecasting Stock Market Volatility: The Gains from Using Intraday Data

There is evidence that volatility forecasting models that use intraday data produce superior forecast accuracy as compared with that delivered by the models that use daily data. However, this evidence is still sparse and incomplete in the stock markets. This paper extends previous studies on forecasting stock market volatility in several important directions and comprehensively assesses the gains in forecast accuracy provided by intraday data. First, we use an extensive set of intraday data on 28 single stocks and 23 stock market indices. Second, in our study we use forecast horizons ranging from 1 day to 6 months. Third, we compare forecasting abilities of several competing models. We find that the amount of gains depends on the length of the forecast horizon, on the forecasting model, and on whether the volatility is forecasted for a single stock or a stock market index. The major finding of our empirical study is that intraday data allow one to increase the forecast accuracy by 10% to 30%. Thus, the gains from using intraday data are rather significant. Surprisingly, we find that the gains in predictive accuracy from intraday data persist over longer forecast horizons and are greater for stock market indices than for single stocks.

Forecasting Inflation Across Euro Area Countries and Sectors: A Panel VAR Approach

In this paper, we adopt a panel vector autoregressive (PVAR) approach to estimating and forecasting inflation dynamics in four different sectors—industry, services, construction and agriculture—across the euro area and its four largest member states: France, Germany, Italy and Spain. By modelling inflation together with real activity, employment and wages at the sectoral level, we are able to disentangle the role of unit labour costs and profit margins as the fundamental determinants of price dynamics on the supply side. In out‐of‐sample forecast comparisons, the PVAR approach performs well against popular alternatives, especially at a short forecast horizon and relative to standard VAR forecasts based on aggregate economy‐wide data. Over longer forecast horizons, the accuracy of the PVAR model tends to decline relative to that of the univariate alternatives, while it remains high relative to the aggregate VAR forecasts. We show that these findings are driven by the event of the Great Recession. Our qualitative results carry over to a multi‐country extension of the PVAR approach. Copyright © 2016 John Wiley & Sons, Ltd.

The Power of Aggregate Book-to-Market Innovations: Forecasting, Nowcasting, and Dating the Real Economy

Aggregate book-to-market (B/M) ratio reflects market-wide assessments of growth opportunities and productivity. In this paper, I find that aggregate B/M innovations predict time-series variations in the U.S. economy. More importantly, the predictive content of the innovations is incremental (or even superior in a long-horizon forecast) to that of the Survey of Professional Forecasters (SPF). A real-time dating algorithm that is based on the innovations accurately identifies the business cycle turning points for the last 40 years. Decomposing aggregate B/M into components of accounting conservatism and delayed recognition of growth reveals different implications for the macroeconomic information content of aggregate B/M.

Long-run restrictions and survey forecasts of output, consumption and investment

We consider the extent to which long-horizon survey forecasts of consumption, investment and output growth are consistent with theory-based steady-state values, and whether imposing these restrictions on long-horizon forecasts will enhance their accuracy. The restrictions that we impose are consistent with a two-sector model in which the variables grow at different rates in steady state. The restrictions are imposed by an exponential-tilting of simple auxiliary forecast densities. We show that imposing the consumption–output restriction yields modest improvements in the long-horizon output growth forecasts, and larger improvements in the forecasts of the cointegrating combination of consumption and output: the transformation of the data on which accuracy is assessed plays an important role.

Analystss Long-Horizon Earnings Forecast Properties and Long-Horizon Macroeconomic Forecast Optimism

We examine whether the properties of earnings forecasts – bias and dispersion are different across periods when macroeconomic forecasts are optimistic than non-optimistic, and whether this difference in analyst forecast optimism is stronger during recessionary periods. We find that the long-horizon earnings forecasts are more optimistically biased in periods when the macroeconomic forecasts are optimistically biased as well, and the bias is more pronounced during periods of recession. We also find that the long-horizon earnings forecast dispersion is lower in periods when the long-horizon macroeconomic forecasts are optimistic than in other periods. These results suggest that firms that meet or beat earnings forecasts when there is no recession and macroeconomic forecast is optimistic are likely to have opportunistically biased their long-term forecasts and walked them down, i.e. opportunistic; and that firms that meet or beat earnings forecasts when there is recession and macroeconomic forecast is optimistic are likely to be the ones that are positioned to perform well when the economy recovers. Consistent with this we find that premium for meeting or beating the analysts’ earnings forecasts is highest in periods when there is recession and macroeconomic forecasts are optimistic; and there is no premium when there is no recession and macroeconomic forecast is optimistic. Collectively, the results show the interaction between the macroeconomic outlook and firm-level forecast properties.

On the Predictive Content of Nonlinear Transformations of Lagged Autoregression Residuals and Time Series Observations

Abstract Although many macroeconomic time series are assumed to follow nonlinear processes, nonlinear models often do not provide better predictions than their linear counterparts. Furthermore, nonlinear models easily become very complex and difficult to estimate. The aim of this study is to investigate whether simple nonlinear extensions of autoregressive processes are able to provide more accurate forecasting results than linear models. Therefore, simple autoregressive processes are extended by means of nonlinear transformations (quadratic, cubic, sine, exponential functions) of lagged time series observations and autoregression residuals. The proposed forecasting models are applied to a large set of macroeconomic and financial time series for 10 European countries. Findings suggest that these models, including nonlinear transformation of lagged autoregression residuals, are able to provide better forecasting results than simple linear models. Thus, it may be possible to improve the forecasting accuracy of linear models by including nonlinear components. This is especially true for time series that are positively tested for nonlinear characteristics and longer forecast horizons.