We find that countries which are able to borrow at spreads that seem low given fundamentals (for example because investors take a bullish view on a country's future), are more likely to develop economic difficulties later on. We obtain this result through a two-stage procedure, where a first regression links sovereign spreads to fundamentals, after which residuals from this regression are deployed in a second stage to assess their impact on future outcomes (real GDP growth and the occurrence of fiscal crises). We confirm the relevance of past sovereign debt mispricing in several out-of-sample exercises, where they reduce the RMSE of real GDP growth forecasts by as much as 15 percent. This provides strong support for theories of sentiment affecting the business cycle. Our findings also suggest that countries shouldn't solely rely on spread levels when determining their fiscal strategy; underlying fundamentals should inform policy as well, since historical relationships between spreads and fundamentals often continue to apply in the medium-to-long run.
The widespread availability of internet search data is a new source of high-frequency information that can potentially improve the precision of macroeconomic forecasting, especially in areas with data constraints. This paper investigates whether travel-related online search queries enhance accuracy in the forecasting of tourist arrivals to The Bahamas from the U.S. The results indicate that the forecast model incorporating internet search data provides additional information about tourist flows over a univariate approach using the traditional autoregressive integrated moving average (ARIMA) model and multivariate models with macroeconomic indicators. The Google Trends-augmented model improves predictability of tourist arrivals by about 30 percent compared to the benchmark ARIMA model and more than 20 percent compared to the model extended only with income and relative prices.
I regress real GDP growth rates on the IMF’s growth forecasts and find that IMF forecasts behave similarly to those generated by overfitted models, placing too much weight on observable predictors and underestimating the forces of mean reversion. I identify several such variables that explain forecasts well but are not predictors of actual growth. I show that, at long horizons, IMF forecasts are little better than a forecasting rule that uses no information other than the historical global sample average growth rate (i.e., a constant). Given the large noise component in forecasts, particularly at longer horizons, the paper calls into question the usefulness of judgment-based medium and long-run forecasts for policy analysis, including for debt sustainability assessments, and points to statistical methods to improve forecast accuracy by taking into account the risk of overfitting.
Jin-Kyu Jung, Manasa Patnam, and Anna Ter-Martirosyan
Forecasting macroeconomic variables is key to developing a view on a country's economic outlook.
Most traditional forecasting models rely on fitting data to a pre-specified relationship between input
and output variables, thereby assuming a specific functional and stochastic process underlying that
process. We pursue a new approach to forecasting by employing a number of machine learning
algorithms, a method that is data driven, and imposing limited restrictions on the nature of the true
relationship between input and output variables. We apply the Elastic Net, SuperLearner, and
Recurring Neural Network algorithms on macro data of seven, broadly representative, advanced and
emerging economies and find that these algorithms can outperform traditional statistical models,
thereby offering a relevant addition to the field of economic forecasting.
Inflation forecasts are modelled as monotonically diverging from an estimated long-run anchor point, or “implicit anchor”, towards actual inflation as the forecast horizon shortens. Fitting the model with forecasts by analysts, businesses and trade unions for South Africa, we find that inflation expectations have become increasingly strongly anchored. That is, the degree to which the estimated implicit anchor pins down inflation expectations at longer horizons has generally increased. Estimated inflation anchors of analysts lie within the 3–6 percent inflation target range of the central bank. However, the implicit anchors of businesses and trade unions, who are directly involved in the setting of wages and prices that drive the inflation process, have remained above the top end of the official target range. Possible explanations for these phenomena are discussed.
This study documents a semi-structural model developed for Sri Lanka. This model, extended with a fiscal sector block, is expected to serve as a core forecasting model in the process of the Central Bank of Sri Lanka’s move towards flexible inflation targeting. The model includes a forward-looking endogenous interest rate and foreign exchange rate policy rules allowing for flexible change in policy behavior. It is a gap model that allows for simultaneous identification of business cycle position and long-term equilibrium. The model was first calibrated and then its data-fit was improved using Bayesian estimation technique with relatively tight priors.
Mr. Sergi Lanau, Adrian Robles, and Mr. Frederik G Toscani
We study inflation dynamics in Colombia using a bottom-up Phillips curve approach. This
allows us to capture the different drivers of individual inflation components. We find that the
Phillips curve is relatively flat in Colombia but steeper than recent estimates for the U.S.
Supply side shocks play an important role for tradable and food prices, while indexation
dynamics are important for non-tradable goods. We show that besides allowing for a more
detailed understanding of inflation drivers, the bottom-up approach also improves on an
aggregate Phillips curve in terms of forecasting ability. In the baseline forecast scenario, both
headline and core inflation converge towards the Central Bank’s inflation target of 3 percent
by end-2018 but these favorable inflation dynamics are vulnerable to large supply shocks.
Ms. Piyabha Kongsamut, Mr. Christian Mumssen, Anne-Charlotte Paret, and Mr. Thierry Tressel
How can information on financial conditions be used to better understand macroeconomic
developments and improve macroeconomic projections? We investigate this question for France
by constructing country-specific financial conditions indices (FCIs) that are tailored to movements
in GDP, investment, private consumption and exports respectively. We rely on a VAR approach to
estimate the weights of the financial components of each FCI, including equity market returns
(which turn out having a relatively strong weight across all FCIs), private sector risk premiums,
long-term interest rates, and banks’ credit standards. We find that the tailored FCIs are useful as
leading indicators of GDP, investment, and exports, and as a contemporaneous indicator of private
consumption. Credit volumes turn out to be lagging indicators of growth. The indices inform us on
macro-financial linkages in France and are used to improve the accuracy of quarterly forecasting
models and high-frequency “nowcast” models. We show that FCI-augmented models could have
significantly improved forecasts during and after the global financial crisis.
Macroeconomic forecasts are persistently too optimistic. This paper finds that common
factors related to general uncertainty about U.S. macrofinancial prospects and global demand
drive this overoptimism. These common factors matter most for advanced economies and G-
20 countries. The results suggest that an increase in uncertainty-driven overoptimism has
dampening effects on next-year real GDP growth rates. This implies that incorporating the
common structure governing forecast errors across countries can help improve subsequent
Macroeconomic analysis in Lebanon presents a distinct challenge. For example, long
delays in the publication of GDP data mean that our analysis often relies on proxy
variables, and resembles an extended version of the “nowcasting” challenge familiar to
many central banks. Addressing this problem—and mindful of the pitfalls of extracting
information from a large number of correlated proxies—we explore some recent
techniques from the machine learning literature. We focus on two popular techniques
(Elastic Net regression and Random Forests) and provide an estimation procedure that is
intuitively familiar and well suited to the challenging features of Lebanon’s data.