Table 1. Table of events of strong trends of early warning signals.

Event number	Duration of high-Kendall in Power spectrum	Nearest recorded crash (start date & source)	EWS (time to crash) or False alarm
1	17/08/1906 to 09/11/1906	Panic of 1907 (14/03/1907)	EWS (4 months)
2	26/07/1916 to 06/10/1916	none	False alarm
3	29/06/1926 to 05/05/1927	1929 Crash (24/10/1929)	EWS (3 years)
4	29/02/1929 to 14/10/1929	1929 Crash (24/10/1929)	EWS (9 months)
5	13/04/1937 to 23/02/1939	1937 recession (26/9/1937)	EWS (5 months)
6	30/07/1946 to 08/10/1947	none	False alarm
7	30/08/1951 to 15/05/1952	none	False alarm
8	26/11/1954 to 10/05/1957	none	False alarm
9	02/11/1959 to 14/04/1961	none	False alarm
10	26/09/1966 to 06/03/1968	none	False alarm
11	06/09/1973 to 09/03/1975	Oil shock (2/11/1973)	EWS (3 months)
12	15/07/1983 to 27/06/1984	Israel oil crisis (10/10/1983)	EWS (3 months)
13	12/11/1986 to 04/08/1988	1987 Crash (19/10/1987)	EWS (11 months)
14	12/10/1990 to 30/08/1991	none	False alarm
15	26/12/1995 to 09/08/2000	Dot-com crash (21/03/2000)	EWS (5 years)
16	15/02/2008 to 14/07/2009	2008 Crash (01/10/2008)	EWS (9 months)

Table of events of strong trends, as determined by Kendall-τ exceeding a threshold value of 0.9 in both variance and power spectrum (see caption for Fig 2). Event numbers in this table correspond to those identified in Fig 2. Note that there are several false alarms, as discussed in the text. In addition, there were persistent EWS for five years preceding 2000 crash. Whereas, in most cases of EWS the crashes occurred within an year or so of the signal. This indicates that occurrence of an EWS is not a predictive measure of when the crash will occur.