What households or firms expect for future macroeconomic outcomes affects their current choices of consumption, investment, or pricing decisions. Measures of expectations are therefore informative about the state of the economy. They are also often used as an input to economic policymaking. For example, central banks closely monitor long-run inflation expectations to assess whether they are well anchored. While private-sector expectations of economic variables are usually derived from asset prices or from the average forecast across respondents in surveys, an additional useful dimension of survey data is that they provide information about how individuals disagree about the same future economic outcome.
A growing theoretical literature has shown that incorporating disagreement in economic models goes a long way towards explaining fluctuations in economic activity, inflation, and asset prices over the business cycle (e.g. Woodford 2003 for producer prices, Scheinkman and Xiong 2003 for asset prices, and Angeletos and La’O 2013 for consumption). However, relatively little is known about the empirical properties of forecaster disagreement. In a recent working paper (Andrade et al. 2014) we document a novel set of facts about disagreement among professional forecasters. Specifically, we study disagreement about US output growth, US inflation, and the federal funds rate for different forecast horizons – including the very long run – over the last 30 years.
The average term structure of disagreement
Figure 1 below shows the average disagreement across time for a set of different forecast horizons ranging from one quarter to 6-to-11 years ahead, and for our three variables of interest as observed in the Blue Chip Financial Forecasts survey. This survey has been conducted monthly since the early 1980s, and asks participants ranging from broker-dealers to economic consulting firms to provide forecasts at different horizons. It is the longest-running survey that includes forecasts at short, medium and long horizons for the three key US macroeconomic variables (output, inflation, and the policy rate).
Figure 1. The average term structure of disagreement
Notes: This figure shows the term structure of disagreement averaged across time for real output growth, CPI inflation, and the federal funds rate for various forecast horizons from the Blue Chip Financial Forecasts survey. The longest horizon captures the average forecast for horizons from 6-to-11 years ahead. Disagreement is defined as the average forecast of the highest ten responses minus that of the lowest ten responses of survey participants.
The figure shows that agents disagree about output growth or inflation not just in the coming quarters but also at very long horizons. Put differently, these forecasters appear to have different views on long-run fundamentals such as potential output growth or the Federal Reserve’s inflation goal. Furthermore, the figure underscores that such ‘fundamental disagreement’ can be lower than, comparable to, or higher than disagreement for short forecast horizons. Perhaps most interestingly, forecasters disagree very little about the near-term outlook for the policy rate, but substantially about its long-run value. For example, at the longest forecast horizon the difference between the average forecast of the highest ten responses and the lowest ten responses has, on average, been a sizable two percentage points.
Fundamental disagreement as a result of imperfect information
This leads to the obvious question: What are the factors behind these facts about forecasters’ disagreement? We argue that forecasters face two important challenges stemming from imperfect information. First, they only have a cloudy view of current economic conditions. For example, measures of GDP are released with a delay and are subject to major revision. At the same time they only have noisy measures of underlying inflation trends. Second, they need to filter from the observed changes in economic conditions the temporary factors from the slow-moving permanent components of the variables of interest (e.g. changes to potential output growth). Because of these ‘information frictions’, each forecaster interprets the data differently, which results in forecast disagreement at all horizons.
As illustrated by each panel in Figure 2, our analysis shows that a simple model of imperfect information incorporating the features described above is able to replicate closely the shape of the average term structure that we observe in the data.
Figure 2. Model-implied average term structures of disagreement
Notes: This figure displays the model-implied (time) average of disagreement across different horizons for the generalised noisy information model (dark blue) and the generalised sticky information model (light blue) along with the Blue Chip Financial Forecasts survey (red). The top panel presents results for real output growth, the middle panel for CPI inflation, and the bottom panel for the federal funds rate. Open circles designate survey moments the calibration aims at replicating.
We offer two comments on Figure 2. First, what is striking about these results is that they do not rely on forecasters having different models of the economy and/or needing to learn about how the data are generated. Indeed, agents in our model agree on a statistical model that best describes the evolution of GDP growth, inflation, and the federal funds rate. Second, agents know that economic variables interact – movements in one of the variables affects their forecasts for all variables. For example, even if forecasters perfectly observe the federal funds rate, as shown in Figure 2, they still disagree substantially about its longer-run value. This is because they disagree about GDP growth and inflation, which are important determinants of the federal funds rate.
Why do people disagree about future monetary policy?
It is common to characterise the movement of the federal funds rate in terms of monetary-policy rules. Can we explain the upward slope of the average term structure of disagreement for the future federal funds rate in terms of such a rule? If so, what do we learn about the properties of this policy rule?
As Figure 3 below illustrates, the observed disagreement about the future federal funds rate is consistent with agents perceiving that the Federal Reserve relies on a policy rule that features a high degree of interest-rate smoothing and slowly moving long-run inflation and output targets. How do we explain this result? Because disagreement about the near-term output and inflation outlook is high, a policy rule without interest rate smoothing would imply much higher disagreement than what we observe in the data. Instead, interest-rate smoothing means that interest-rate forecasts in the near future are strongly determined by the current federal funds rate, which everyone observes. However, as the forecast horizon grows, disagreement about the slow-moving targets for output and inflation are increasingly important for interest-rate forecasts.
Figure 3. The average term structure of disagreement for different policy rules
Notes: This figure displays the term structure of federal funds rate disagreement averaged across time observed in the Blue Chip Financial Forecasts survey and the one implied by forecasters relying on a simple policy rule who face the ‘information frictions’ discussed in the text. īt, ḡt and ∏t are the long-run federal funds rate, output growth rate, and CPI inflation rate. Open circles designate survey moments used in the calibration. Results are based on 5,000 simulations.
Another lesson from Figure 3 is that even when one accounts for the disagreement about the inflation target and potential output growth, there is still a gap between model-implied and observed interest-rate disagreement (the difference between the solid black and solid blue lines). This suggests that although forecasters expect the Federal Reserve to achieve its goals, they still disagree about the policy rate in the long term.
This new evidence points to the importance of three features in the modelling of expectations. First, agents have imperfect information about the true state of the economy as advocated in models of information rigidities (e.g. Mankiw and Reis 2002, Sims 2003, or Woodford 2003). Second, changes in the state of the economy may come from either transitory or persistent components (or both), which forecasters must disentangle. Third, agents must take into account the dynamic interaction between variables.
Explicitly incorporating these elements in the expectation-formation process is critical to explaining the term structure of disagreement. Systematically integrating such information frictions and long-run uncertainty into macroeconomic models should therefore be a promising area for future research.
Authors’ note: The views expressed here are those of the authors and are not necessarily reflective of views at the Federal Reserve Bank of New York, the Federal Reserve System, the Banque de France or the Euro-system.
Andrade, P, R K Crump, S Eusepi, and E Moench (2014), “Fundamental disagreement”, FRBNY Staff Report 655.
Angeletos, G and J La’O (2013), “Sentiments”, Econometrica 81(2): 739–779.
Mankiw, G and R Reis (2002), “Sticky information versus sticky prices: a proposal to replace the new Keynesian Phillips curve”, Quarterly Journal of Economics 117(4): 1295–1328.
Scheinkman, J and W Xiong (2003), “Overconfidence and speculative bubbles”, Journal of Political Economy 111(6): 1183–1220.
Sims, C (2003), “Implications of rational inattention”, Journal of Monetary Economics 50(3): 665–690.
Woodford, M (2003), “Imperfect common knowledge and the effects of monetary policy”, in P Aghion, R Frydman, J Stiglitz, and M Woodford (eds.), Knowledge, information, and expectations in modern macroeconomics: in honor of Edmund S. Phelps, Princeton University Press.