Forecast Horizon Increases Research Articles

Forecast content and content horizons for some important macroeconomic time series

Abstract. The information content of statistical forecasts of approximately stationary quantities tends to decline as the forecast horizon increases, and there exists a maximum horizon beyond which forecasts cannot provide discernibly more information about the variable than is present in the unconditional mean (the content horizon). The pattern of decay of forecast content (or skill) with increasing horizon is well known for many types of meteorological forecasts; by contrast, little generally accepted information about these patterns or content horizons is available for economic variables. In this paper we estimate content horizons for a variety of macroeconomic quantities; more generally, we characterize the pattern of decay of forecast content as we project farther into the future. We find a wide variety of results for the different macroeconomic quantities, with models for some quantities providing useful content several years into the future, for other quantities providing negligible content beyond one or two months or quarters.

Multi-objective performance comparison of an artificial neural network and a conceptual rainfall—runoff model

A multi-objective comparison between an artificial neural network and the conceptual HBV rainfall—runoff model has been performed. The popular NSGA-II algorithm was used for calibration of both models. A combination of three objective functions was used to evaluate model performance. The results show that, for a small forecast lead time, the artificial neural network outperformed the HBV model on the objective functions for low and high flows, but the former was outperformed on a novel objective function related to the shape of the hydrograph. As the forecast horizon increases, the HBV model starts to outperform the ANN model on all objective functions. The main conclusion of this study is that, although the differences between the two model approaches make a straightforward and unequivocal comparison difficult, the multi-objective approach enables a more reliable evaluation of the two models than the single-objective approach.

Motivational and Cognitive Determinants of Buy-Side and Sell-Side Analyst Earnings Forecasts: An Experimental Study

This paper provides experimental evidence about the differences between buy-side analyst (BSA) and sell-side analyst (SSA) earnings forecasts, and investigates both motivational and cognitive determinants of these differences. Regarding motivational determinants, we argue that the SSA work environment contains greater incentives for optimistic forecasts than does the BSA work environment. Regarding cognitive determinants, we examine whether three characteristics of the information on which analysts base their forecasts (trend, variability, and recency) contribute to optimism. We also examine whether forecasts are more optimistic over longer forecast horizons. Results indicate that, as expected, SSAs make more favorable earnings forecast revisions than BSAs, and, consistent with prior research, analyst forecasts are greater as forecast horizon increases. In addition, while information variability does not contribute to optimism, differences in trend and recency do. Specifically, analysts act as if they discount both past earnings information with a decreasing trend and negative recent information when revising their forecasts. Directions for additional research on motivational and cognitive determinants of analyst forecasts are offered.

Explaining Bureaucratic Optimism: Theory and Evidence from U.S. Executive Agency Macroeconomic Forecasts

We offer a theory of intertemporal bureaucratic decision making which proposes that an agency's forecast optimism is related to the extent to which it discounts future reputation costs associated with bureaucratic incompetence. Agency forecasts of the distant future are more likely to be optimistic than short-term forecasts. We claim that unstable organizations will discount reputation costs at a steeper rate than stable organizations, and therefore will produce more optimistic forecasts. We test our theory using macroeconomic forecasts produced by the Office of Management and Budget (OMB) and the Social Security Administration (SSA) across six forecast horizons from 1979 to 2003. The statistical results are generally consistent with our theory: OMB generates more optimistic long-term forecasts than SSA. Further, differences in forecast optimism between these executive branch agencies widen as the forecast horizon increases. Our evidence suggests that more stable agencies place a premium on minimizing reputation costs. Conversely, less stable agencies are more likely to accommodate political pressures for forecast optimism. These findings underscore the importance of institutional design for understanding how executive agencies balance the conflicting goals of political responsiveness and bureaucratic competence within the administrative state.

Forecasting Canadian Time Series with the New-Keynesian Model

This paper documents the out-of-sample forecasting accuracy of the New Keynesian Model for Canada. We repeatedly estimate our variant of the model on a series of rolling subsamples, forecasting out-of-sample one to eight quarters ahead at each step. We then compare these forecasts to those arising from simple VARs, using econometric tests of forecasting accuracy. Our results show that the forecasting accuracy of the New Keynesian model compares favourably to that of the benchmarks, particularly as the forecasting horizon increases. These results suggest that the model can become a useful forecasting tool for Canadian time series. The principle of parsimony is invoked to explain our findings.

An Assessment of Bank of England and National Institute Inflation Forecast Uncertainties

In February 1996 the Bank of England and the National Institute of Economic and Social Research significantly increased the amount of information they published about the uncertainty surrounding their central projections of inflation. In effect, and in different ways, they each began to publish a density forecast of inflation, that is, an estimate of the probability distribution of possible outcomes for future inflation. The Bank represented this graphically, as a set of forecast intervals covering 10, 20, 30, …, 90 per cent of the probability distribution, coloured red, of lighter shades for the outer bands. This was done for inflation forecasts up to eight quarters ahead, and since the distribution becomes increasingly dispersed and the intervals ‘fan out’ as the forecast horizon increases, the chart became known as the ‘fan chart’ (or, rather more informally, and noting its red colour, the ‘rivers of blood’). The National Institute represented the distribution as a histogram, in the form of a table reporting the probabilities of inflation falling in various ranges. These intervals, or ‘bins’ of the histogram, have changed from time to time; those used currently are: less than 1.5 per cent, 1.5 to 2.0 per cent, 2.0 to 2.5 per cent, and so on. The forecasts refer to the fourth quarters of the current and following years, and from the beginning have included not only inflation but also real GDP growth. Fan charts for real GDP growth first appeared in the Bank's Inflation Report in November 1997.

Optimal Forecasts of Long-Term Returns: Geometric, Arithmetic, or Other Means

It is well known that an unbiased forecast of the terminal value of a portfolio requires compounding at the arithmetic mean rate of return over the investment horizon. Yet, the procedure applied to the standard unbiased estimator of the mean return, while maximum likelihood, produces very biased forecasts. We address the forecasting of long-term returns. First, we show that the geometric estimator often proposed as alternative to the MLE is also biased. We derive the unbiased forecast of long-term returns for given estimation period and horizon. Second, we derive an efficient, minimum mean squared error estimator for the long-term. Both estimators impose a penalty that decreases the annual compounding rate as the forecasting horizon increases. The minimum MSE estimator is however lower than the alternative MLE and unbiased estimator, even than the geometric estimator as H becomes larger. Third, we show that parameter uncertainty and forecasting horizon interact to produce optimal portfolio allocations in striking contrast with conventional wisdom: Longer investment horizons require lower, not higher, allocations to risky assets. Finally, we derive a Bayes estimate of long-term expected returns. While the small sample efficiency of the Bayesian method makes it ideal for this problem, we show that standard priors have unreasonable implications and need to be modified. Finally, we show how our results may adjust to non i.i.d. returns.

Forecasting real estate returns using financial spreads

This paper examines the predictability of real estate asset returns using a number of time series techniques. A vector autoregressive model, which incorporates financial spreads, is able to improve upon the out of sample forecasting performance of univariate time series models at a short forecasting horizon. However, as the forecasting horizon increases, the explanatory power of such models is reduced, so that returns on real estate assets are best forecast using the long term mean of the series. In the case of indirect property returns, such short-term forecasts can be turned into a trading rule that can generate excess returns over a buy-and-hold strategy gross of transactions costs, although none of the trading rules developed could cover the associated transactions costs. It is therefore concluded that such forecastability is entirely consistent with stock market efficiency.

Sources of fluctuations in Australian imports

Dynamic interactions among the real exchange rate, income and imports are modelled for Australia. Evidence of one cointegrating relationship is found among these series and base structural inferences on long-run identifying restrictions of the type proposed for vector-error correction models by King, Plosser, Stock and Watson (1991). Under the identifying restriction that a domestic supply shock has no effect on the real exchange rate in the long run, this shock explains most of the forecast-error variance of imports at all horizons except for one quarter. At this horizon, the domestic demand shock is most important. The foreign supply shock becomes more important as the forecast horizon increases and explains almost half of the forecast-error variance of imports in the long run.

Analysts' use of earnings forecasts in predicting stock returns: Forecast horizon effects

Little attention has been paid to a principal decision context in which analysts' earnings forecasts are prepared, namely, as an input to their recommendations. We use two data sets, Value Line, USA, and Research Evaluation Service, Canada, and examine the importance of analysts' earnings forecasts for their stock price forecasts via three hypotheses: (1) analysts' earnings forecasts are important for their stock price forecasts; (2) analysts' long-term earnings forecasts are more important than their short-term earnings forecasts for their predictions of stock prices over a particular stock price forecast horizon; (3) the importance of analysts' earnings forecasts for their stock price forecasts rises as the joint earnings and stock price forecast horizon increases. We show that: (1) when the earnings forecast horizon is the next fiscal year, forecasted earnings explain only 30% of the variation in forecasted price; (2) the importance of forecasted earnings for forecasted price rises as the earnings forecast horizon increases; (3) in the long run, (i.e. three to five years hence), forecasted earnings explain about 60% of the variation in forecasted price. Decision usefulness is an ex ante concept, but tests regarding the usefulness of earnings for stock price generally have used actual (not expectational) data. Our evidence suggests that earnings expectations are decision useful, where the decision context is sell-side analysts' stock price forecasts. Our results are potentially important to users of sell-side analyst research reports. When a stock recommendation is accompanied only by short-run earnings forecasts, investors need to closely examine estimates of non-earnings variables to assess the quality of stock recommendations. In contrast, when stock recommendations are accompanied by both short-run and long-run earnings forecasts, investors need to examine estimates of non-earnings information variables less closely.

The effect of parameter uncertainty on forecast variances and confidence intervals for unit root and trend stationary time‐series models

AbstractIn this paper I describe the effect of parameter uncertainty on the way conditional forecast variances grow as the forecast horizon increases. Without parameter uncertainty, forecast variances for the unit root model grow linearly with the forecast horizon while with the trend stationary model they are bounded. With parameter uncertainty, however, I find that for both the unit root and the trend stationary models, forecast variances grow with the square of the forecast horizon so that uncertainty grows at a much faster rate than without parameter uncertainty.

Exact and Approximate Multi-Period Mean-Square Forecast Errors for Dynamic Econometric Models

Both future disturbances and estimated coefficients contribute to the uncertainty in model-based ex ante forecasts, but only the first source is usually taken into account when calculating confidence intervals for practical applications. Schmidt (1974) and Baillie (1979) provide an easily computable second-order approximation to the mean-square forecast error (MSFE) for linear dynamic systems which recognizes both sources of uncertainty. To assess the accuracy of their approximation, and thus its usefulness, we compare it with three sets of estimates of the exact MSFE for the univariate AR(l) process: Monte Carlo estimates for OLS, analytically based values for OLS, and Monte Carlo estimates for maximum likelihood. We find that the Schmidt-Baillie formula is a good approximation to the exact MSFE, and that it helps explain why the exact MSFE can decrease as the forecast horizon increases. In fact, for dynamics typical to econometric models, the MSFE often has a maximum at a forecast horizon of one to twelve periods, i.e., at horizons that are of principal concern to forecasters and policy makers.

Econometric GNP forecasts: Incremental information relative to naive extrapolation

Recent studies of macroeconomic forecasts have focused primarily on the relative performance of individual forecasts and combinations thereof. We suggest that these forecasts be evaluated in terms of the incremental information that they provide relative to a simple extrapolation forecast. Using a Bayesian approach, we measure the incremental information contained in econometric forecasts of U.S. GNP relative to a random-walk-with-drift time series forecast. The results indicate that (1) substantial incremental gains can be obtained from econometric GNP forecasts for the current quarter, but that these gains decrease rapidly as the forecast horizon increases, and (2) after one econometric forecast has been consulted, subsequent such forecasts add little information.