Forecasting with Economic News

Figures & data

Fig. 1 Histogram of the number of days between the end of the reference period and the date of the first release of the dependent variables, namely CPIAUCSL, GDPC1, INDPRO and PAYEMS.

Fig. 2 Time series of the sentiment measures for all verbal tenses. The daily series is smoothed by taking a moving-average of 30 days and then subsampled at monthly frequencies. The gray areas denote the NBER recessionary periods.

Fig. 3 Kernel density of the economic sentiment measures for all verbal tenses separately for periods of expansion (red) and recession (green) as defined by the NBER business cycle committee. The density is calculated on the monthly series.

Fig. 4 Statistical significance of the sentiment measures based on the double lasso penalized regression in Equation (1)(1) $$Y_t^d={\eta }_h{S}_{d-h}+X_{d-h}^{\prime }{\beta }_h+{ϵ}_{t,d-h},$$(1) at each horizon h. The statistic reported is the p-value of the sentiment coefficient η_h corrected for multiple testing. The colors depend on the p-value for those variables that are significant at least at 10%: the darker the tile’s color, the smaller the p-value. The gray-shaded area corresponds to the reference period.

Fig. 5 Significance of the sentiment measures based on the double lasso penalized quantile regression at each horizon h for quantiles 0.1 (left column), 0.5 (center) and 0.9 (right). The colors depend on the p-value corrected for multiple testing for those variables that are significant at least at 10%: the darker the tile’s color, the smaller the p-value. The gray-shaded area corresponds to the reference period.

Table 1 One-sided p-values of the aSPA multi-horizon test proposed by Quaedvlieg (2021).

Sentiment	GDPC1	INDP	PAYEMS	CPIAUCSL
AVERAGE	0.143	0.006	0.010	0.020
LASSO	0.675	0.987	0.051	0.987
ECONOMY	0.646	0.831	0.008	0.801
FINANCIAL SECTOR	0.923	0.980	0.180	0.030
INFLATION	0.988	0.972	0.921	0.791
MANUFACTURING	0.990	0.035	0.177	0.365
MONETARY POLICY	0.950	0.076	0.149	0.619
UNEMPLOYMENT	0.920	0.098	0.983	0.801

NOTE: The test compares the ARXS, Average, and LASSO forecasts against the ARX forecasts. Significant p-values at 10% are reported in bold.

Table 2 One-sided p-values of the aSPA multi-horizon test proposed by Quaedvlieg (2021).

Variable	Sentiment	$\tau =0.10$	$\tau =0.50$	$\tau =0.90$	One-sided
GDP	AVERAGE	0.968	0.237	0.115	0.956
	LASSO	0.533	0.968	0.351	0.995
	ECONOMY	0.990	0.913	0.977	0.988
	FINANCIAL SECTOR	0.988	0.989	0.848	0.986
	INFLATION	0.967	0.986	0.978	0.955
	MANUFACTURING	0.977	0.953	0.718	0.970
	MONETARY POLICY	0.965	0.904	0.944	0.961
	UNEMPLOYMENT	0.975	0.947	0.197	0.969
INDP	AVERAGE	0.019	0.008	0.033	0.589
	LASSO	0.993	0.435	0.992	0.997
	ECONOMY	0.987	0.982	0.044	0.993
	FINANCIAL SECTOR	0.953	0.956	0.983	0.008
	INFLATION	0.891	0.954	0.948	0.960
	MANUFACTURING	0.902	0.389	0.985	0.989
	MONETARY POLICY	0.989	0.974	0.995	0.973
	UNEMPLOYMENT	0.631	0.643	0.958	0.165
PAYEMS	AVERAGE	0.003	0.016	0.022	0.011
	LASSO	0.010	0.861	0.988	0.890
	ECONOMY	0.016	0.144	0.055	0.997
	FINANCIAL SECTOR	0.077	0.050	0.989	0.023
	INFLATION	0.987	0.972	0.985	0.943
	MANUFACTURING	0.032	0.852	0.908	0.483
	MONETARY POLICY	0.172	0.337	0.993	0.990
	UNEMPLOYMENT	0.849	0.997	0.999	0.009
CPI	AVERAGE	0.019	0.011	0.173	0.731
	LASSO	0.051	0.079	0.976	0.915
	ECONOMY	0.977	0.801	0.976	0.095
	INFLATION	0.012	0.010	0.341	0.933
	FINANCIAL SECTOR	0.889	0.027	0.969	0.281
	MANUFACTURING	0.915	0.199	0.865	0.985
	MONETARY POLICY	0.605	0.923	0.030	0.050
	UNEMPLOYMENT	0.941	0.925	0.973	0.894

NOTE: The test compares the ARXS, Average, and LASSO quantile forecasts against the ARX quantile forecasts for τ equal to 0.10, 0.50, and 0.9 (first three columns). We also evaluate the predictive ability on the one-sided interval from 0 to 0.10 (last column).

Quaedvlieg, R. (2021), “Multi-Horizon Forecast Comparison,” Journal of Business & Economic Statistics, 39, 40–53.

 Web of Science ®Google Scholar