• Selected
• Entire List

Abstract:
On November 26, 2001, the National Bureau of Economic Research announced that the U.S. economy had officially entered into a recession in March 2001. This decision was a surprise and did not end all the conflicting opinions expressed by economists. This matter was finally settled in July 2002 after a revision to the 2001 real gross domestic products showed negative growth rates for its first three quarters.

A series of political and economic events in the years 2000-01 has increased the amount of uncertainty in the state of the economy, which, in turn, has resulted in the production of less reliable economic indicators and forecasts. This paper evaluates the performances of two very reliable methodologies for predicting a downturn in the U.S. economy using composite leading economic indicators (CLI) for the years 2000-01. It explores the impact of the monetary policy on CLI and on the overall economy and shows how the gradualness and uncertainty of this impact on the overall economy have affected the forecasts of these methodologies. It suggests that the overexposure of the CLI to the monetary policy tools and a strong, but less effective, expansionary money policy have been the major factors in deteriorating the predictions of these methodologies. To improve these forecasts, it has explored the inclusion of the CLI diffusion index as a prior in the Bayesian methodology.

U.S. economy, predicting recessions, monetary policy, composite leading indicators, composite leading indicators diffusion index, Bayesian probability forecasting, classical statistical decision theory, information theory

Abstract:
On November 26, 2001, the National Bureau of Economic Research announced that the U.S. economy had officially entered into a recession in March 2001. This decision was a surprise and did not end all the conflicting opinions expressed by economists. This matter was finally settled in July 2002 after a revision to the 2001 real gross domestic products showed negative growth rates for its first three quarters.

A series of political and economic events in the years 2000-01 has increased the amount of uncertainty in the state of the economy, which, in turn, has resulted in the production of less reliable economic indicators and forecasts. This paper evaluates the performances of two very reliable methodologies for predicting a downturn in the U.S. economy using composite leading economic indicators (CLI) for the years 2000-01. It explores the impact of the monetary policy on CLI and on the overall economy and shows how the gradualness and uncertainty of this impact on the overall economy have affected the forecasts of these methodologies. It suggests that the overexposure of the CLI to the monetary policy tools and a strong, but less effective, expansionary money policy have been the major factors in deteriorating the predictions of these methodologies. To improve these forecasts, it has explored the inclusion of the CLI diffusion index as a prior in the Bayesian methodology.

U.S. economy, predicting recessions, monetary policy, composite leading indicators, composite leading indicators diffusion index, Bayesian probability forecasting, classical statistical decision theory, information theory

Abstract:
This paper presents a Bayesian methodology for estimating probability of a downturn in the economy and applies it to the 2007-2008 state of the U.S. economy with the focus on investigating the occurrence of a recession. In the methodological development, information theory (Kullback and Shannon) is used to assess the change in the amount of information when the economy evolves in moving from one state to another. For the application purposes, the latest information (March 2008) of the Conference Board's Composite Leading Economic Indicators (CLI) and Composite Leading Economic Indicators Diffusion Index (CLID) are used in estimating Bayesian probability of a downturn in U.S. The analysis shows that the models developed have been very accurate in predicting all the past recessions since 1959 correctly with only one false alarm in mid-1960s. When CLID is used as prior information for CLI as evidence, it helped the Bayesian probability of a downturn to be more decisive in its signaling, but at a cost of more false alarms. Overall, the models using only CLI have the best performance in terms of accuracy and false alarm. The best models have signaled for a downturn in the U.S. Economy as early as the Fall 2007.

2007-2008 U.S. economy, predicting recession, Bayesian probability forecasting, composite leading indicators, composite leading indicators diffusion index, information theory

Abstract:
Since the late 1970s/early 1980s, the entire U.S. economy has gone through some structural changes. Outside of the technological changes, the Federal Reserve monetary policies have probably been the main force behind these changes. These policies, known as soft-landing policies, focused on a stable growth in the economy by keeping the inflation low, with the ultimate objective of removing extreme fluctuations. To cope with these structural changes, the composite leading economic indicators have gone through a series of revisions since 1996 (CLI-96) with the objective of improving its quality information for the prediction of the near term state of the economy. In this research we look at two related questions: (1) Have these revisions in CLI-96 captured all the effects of the structural changes on the state of the economy since the early 1980s, or are there still some effects left behind and not captured? (2) Given that the latter part of question (1) is true, can this new information be measured and utilized to produce a better prediction of the near term state of the economy? Our research shows that CLI-96 did not capture all the effects of the structural changes in the economy. When the probability distribution of the CLI-96 growth rates for the expansion periods since 1983 is used, the two methodologies of the classical statistical decision theory and the Bayesian have predicted the 2001 recession correctly, but the overall reliability of their predictions has deteriorated. Overall, it can be concluded that the severity of the 2001 recession is more comparable to the economic slowdowns of 1993 and 1996 than to the recession of 1990.

U.S. economy, predicting recessions, monetary policy, composite leading indicators, Bayesian probability forecasting, classical statistical decision theory, information theory

Abstract:
This paper presents a new Bayesian methodology for predicting a turning point in an economic system. The methodology utilizes information-theoretic measurements for assessing likelihood functions for a turning point. This methodology shows that the total information of a likelihood function consist of two parts: one part measures the information closeness of the observed values to the probability distribution of a given state and one part measures the information content of the serial correlation in data. For application purposes, this methodology allows us to explicitly consider more than one recent observation in the analysis.

Bayes Rule, Kullback Information Measure, Composite Leading Economic

Abstract:
In an attempt to predict a peak in the U.S. economy using a classical statistical decision methodology and a Bayesian methodology and using the 1996 revised composite leading economic indicators, it is learned that the Bayesian models have generally outperformed the classical statistical ones and, among the Bayesian models, the two using two and three consecutive CLI growth rates are superior in reliability and in accuracy. These two models, however, failed to correctly predict the 2001 recession.

In investigating the reasons behind their failures, we learned that (1) if the concurrent data to the economic structure of 1983-1999 are used for the prediction, they have also been able to predict the 2001 recession correctly, but their overall reliability is not as strong as before, (2) given the overwhelming weight of the monetary policy tools in the CLI-1996 design and the combination of the economic and political events in year 2000, the less than expected effectiveness of the monetary policy since year 2001 has contributed to this failure, and (3) a possible structural change in the U.S. economy since year 2000 has also contributed to this prediction failure.

U.S. economy, predicting recessions, monetary policy, structural change, composite leading indicators, Bayesian probability forecasting, classical statistical decision theory, information theory

Abstract:
In an experts-assisted decision making paradigm, the information collection design becomes a strategic variable under a weak assumption that the final decision is dependent on the design used to collect information as well. As a result, the same information of the experts and the decision maker about the problem can potentially produce different final decisions for different information collection designs. The implication is that a decision maker can strategically select a design which serves his/her objective.

This paper uses a Bayesian estimation methodology for combining experts' information with the decision maker's prior. An information collection process is designed by setting constraints on this model. Several designs are developed here using such controlled factors as a one-stage versus a two-stage decision process, experts' rank ordering, and group versus individual lobbying/consultation.

An example is provided to illustrate the applicability of the concept. It is shown that the information produced in the process of producing a decision can also give insights into the impacts of the decision maker and the experts on the decision.

Experts-assisted decision making, Reaching a consensus, Bayesian method, Information theory

Abstract:
The problem of modeling the revision of the information of a decision maker based on the information of the expert sources is considered. The basic model assumes that the information of the decision maker and expert sources is in the form of the probability mass functions. The modeling approach is Bayesian estimation, which relies on Kullback entropy and Shannon entropy for information measurement, and produces a unique solution. Modeling of the problem not only considers information about the statistical dependence of the expert sources, but also uses information to measure the quality and importance of the individual expert sources in the form of rank ordering. The outcome shows that the effects of the dependence and rank ordering of the expert sources on the final decision cannot be isolated. In a special case where this isolation is possible, the effect of rank ordering decreases with the increase in the value of the correlation coefficient from -1 to +1, and the effect of the correlation never exceeds the effect of rank ordering. Sensitivity analysis is performed to explore other properties of the model related to the influence of the decision maker and expert sources. Extensions of the basic modeling to group decision making, group consensus, and mean value information are presented.

Bayesian Estimation, Information Theory, Kullback Entropy , Shannon Entropy, Group Decision Making, Consensus, Combining Information, Combining Forecasts, Rank Ordering

Abstract:
This paper uses a normative method to compare the performance of the composite leading economic indicators (CLI) after the measure was revised by the Conference Board in 1996 and 2001 with its prior design to check for a claim that the new design improves its performance in predicting a downturn in the U.S. economy. A comparison of the Bayesian probability forecasts of a downturn for these two designs has shown that the two consecutive post-1996 CLIs have the same amount of information as the four consecutive CLIs of the prior design for correctly predicting/detecting a recession in the U.S. economy, with no false alarm. This finding supports the Conference Board's claim.

Abstract:
In modeling a combination of forecasts all the information related to the past performance of the individual forecasts, including accuracy and correlation, is considered. In this paper I have extended the modeling to incorporate a rank ordering of the forecasts by a decision maker. This ordering could be based on the expectations of a decision maker or on the judgment of an expert about the relative future performance of the forecasts. The problem is set up as a likelihood function of the individual forecasts given the combined forecast. It is shown that this likelihood function is approximately an exponential function of a relative entropy information measure. The maximum likelihood combined forecast is a weighted linear function of the individual forecasts, where the weights are a function of the past performance of the individual forecasts, the correlations between the forecasts and the decision maker's ranking of the forecasts. It is shown that ranking is effective only when the forecasts are correlated: the greater the correlation, the more effective the ranking. A sample application of this methodology to forecasting U.S. hog prices shows that ordering forecasts according to their individual performances produces a very robust and accurate combined forecast; however, this forecast is not the most accurate among the combined forecasts.

Abstract:
This paper presents a paper for producing a probability forecast of a turning point in U.S. economy using Composite Leading Indicators. This methodology is based on classical statistical decision theory and uses information-theoretic measurement to produce a probability. The methodology is flexible using as many historical data points as desired. It is applied to producing probability forecasts of a downturn for 1970-1990 period. Four probability forecasts are produced using different amounts of information. The performance of these forecasts is evaluated using NBER downturn points and the scores measuring accuracy, calibration, and resolution. An indirect comparison of these forecasts with Neftci's sequential probability recursion is also presented. It is shown that the performances of our best two models are statistically different from the performance of the three-consecutive- month decline model and are the same as the one for the probit model. The probit model, however, is more conservative and failed to predict two of the recessions.

Abstract:
This paper presents a methodology for producing a probability forecast of a turning point in U.S. economy using Composite Leading Indicators. This methodology is based on classical statistical decision theory and uses information-theoretic measurement to produce a probability. The methodology is flexible using as many historical data points as desired. This methodology is applied to producing probability forecasts of a downturn in U.S. economy in 1970-1990 period. Four probability forecasts are produced using different amounts of information. The performance of these forecasts is evaluated using the actual downturn points and the scores measuring accuracy, calibration, and resolution. An indirect comparison of these forecasts with Diebold and Rudebusch's sequential probability recursion is also presented. It is shown that the performances of our best two models are statistically different from the performance of the three- consecutive-month decline model and are the same as the one for the best probit model. The probit model, however, is more conservative in its predictions than our two models.