Can news-based economic sentiment predict bubbles in precious metal markets?

Maghyereh, Aktham; Abdoh, Hussein

doi:10.1186/s40854-022-00341-w

Financial Innovation

Table 11 Estimation results under different subsamples

From: Can news-based economic sentiment predict bubbles in precious metal markets?

	Dependent variable: Bubble
	Gold	Silver	Palladium	Platinum
Panel A: Before GFC (January 1985 to November 2007)
\(NESI_{t - 1}\)	− 0.8726**	− 0.2250	0.4289	− 0.5202***
\(NESI_{t - 1}\)	(0.0198)	(0.2810)	(0.5010)	(0.0000)
\(Inflation_{t - 1}\)	1.0110***	1.3990*	0.6475***	2.7806**
\(Inflation_{t - 1}\)	(0.0020)	(0.0600)	(0.0000)	(0.0150)
\(US{\text{DI}}_{t - 1}\)	− 0.7840	− 2.4983**	− 0.7669***	− 0.6896***
\(US{\text{DI}}_{t - 1}\)	(0.1660)	(0.0120)	(0.0040)	(0.0010)
\({\text{EFR}}_{t - 1}\)	− 0.0204	− 0.2623	− 0.1273	− 0.2918***
\({\text{EFR}}_{t - 1}\)	(0.8410)	(0.1580)	(0.2210)	(0.0020)
\(T - Spread_{t - 1}\)	− 0.2347	− 1.6647**	− 2.7865***	− 1.2880**
\(T - Spread_{t - 1}\)	(0.6610)	(0.0253)	(0.0010)	(0.0260)
\({\text{GEA}}_{t - 1}\)	0.0010	0.0181**	0.0055	0.0099**
\({\text{GEA}}_{t - 1}\)	(0.8460)	(0.0340)	(0.1930)	(0.0150)
\(Constant\)	1.0716	3.9195**	5.2929	4.0492
\(Constant\)	(0.4670)	(0.0437)	(0.3401)	(0.1460)
\(Observations\)	272	272	272	272
McFadden's pseud-R²	0.3998	0.5996	0.3219	0.4873
Log-likelihood	− 54.1213	− 39.6921	− 68.1614	− 74.8537
Hosmer–Lemeshow test	3.17	3.97	12.29	9.88
Hosmer–Lemeshow test	(0.9235)	(0.8595)	(0.1388)	90.2737)
Panel B: After GFC (December 2007 to August 2020)
\(NESI_{t - 1}\)	− 2.7319***	− 1.2156***	0.5930	− 1.0272***
\(NESI_{t - 1}\)	(0.0019)	(0.0027)	(0.7240)	(0.0000)
\(Inflation_{t - 1}\)	2.1302***	2.4118*	1.7276**	3.1534***
\(Inflation_{t - 1}\)	(0.0060)	(0.0703)	(0.0110)	(0.0000)
\(US{\text{DI}}_{t - 1}\)	− 5.4248*	− 4.8325***	− 1.3515	− 2.0290***
\(US{\text{DI}}_{t - 1}\)	(0.0590)	(0.0013)	(0.7320)	(0.0001)
\({\text{EFR}}_{t - 1}\)	− 1.2303**	− 1.2263**	− 2.1625*	− 0.9474***
\({\text{EFR}}_{t - 1}\)	(0.0296)	(0.0380)	(0.0646)	(0.0000)
\(T - Spread_{t - 1}\)	− 1.2499***	− 4.2208***	− 0.6202	− 2.3639
\(T - Spread_{t - 1}\)	(0.0072)	(0.0029)	(0.5510)	(0.0038)
\({\text{GEA}}_{t - 1}\)	0.0203**	0.0146*	0.0042	0.0394***
\({\text{GEA}}_{t - 1}\)	(0.0448)	(0.0512)	(0.5160)	(0.0094)
\(Constant\)	1.6553*	3.0170***	9.8635**	5.1021*
\(Constant\)	(0.0660)	(0.0015)	(0.0140)	(0.0790)
\(Observations\)	136	136	136	136
McFadden's pseud-R²	0.9532	0.8488	0.3632	0.9607
Log-likelihood	− 14.1578	− 16.13851	− 25.8458	− 45.083
Hosmer–Lemeshow test	7.75	0.04	1.76	6.89
Hosmer–Lemeshow test	(1.0000)	(1.0000)	(0.9874)	(0.4185)

The dependent variable is a binary that equals 1 (bubble dates) and 0 (none-bubble dates) identified by the GSADF procedure. Panels A and B report the results of the probit regressions based on the news-based economic sentiment index (NESI). The Hosmer–Lemeshow test is a statistical test for goodness of fit for probit regressions, which follows an \(\chi^{2}\) distribution. A large \(\chi^{2}\) value (with small p value \(< 0.05\)) indicates poor fit regression model. p values are given in brackets.*, **, and *** indicate significance at 10%, 5%, and 1% levels, respectively.

Back to article page