Editing Earthquake prediction (section)

== Evaluating earthquake predictions ==
{{See also|Prediction#Prediction in science}}

Predictions are deemed significant if they can be shown to be successful beyond random chance.<ref>{{Harvnb|Mulargia|Gasperini|1992|p=32}}; {{Harvnb|Luen|Stark|2008|p=302}}.</ref> Therefore, methods of [[statistical hypothesis testing]] are used to determine the probability that an earthquake such as is predicted would happen anyway (the [[null hypothesis]]). The predictions are then evaluated by testing whether they correlate with actual earthquakes better than the null hypothesis.<ref>{{Harvnb|Luen|Stark|2008}}; {{Harvnb|Console|2001}}.</ref>

In many instances, however, the statistical nature of earthquake occurrence is not simply homogeneous. Clustering occurs in both space and time.<ref>{{Harvnb|Jackson|1996a|p=3775}}.</ref> In southern California about 6% of M≥3.0 earthquakes are "followed by an earthquake of larger magnitude within 5 days and 10&nbsp;km."<ref>{{Harvnb|Jones|1985|p=1669}}.</ref> In central Italy 9.5% of M≥3.0 earthquakes are followed by a larger event within 48 hours and 30&nbsp;km.<ref>{{Harvnb|Console|2001|p=1261}}.</ref> While such statistics are not satisfactory for purposes of prediction (giving ten to twenty false alarms for each successful prediction) they will skew the results of any analysis that assumes that earthquakes occur randomly in time, for example, as realized from a [[Poisson process]]. It has been shown that a "naive" method based solely on clustering can successfully predict about 5% of earthquakes; "far better than 'chance'".<ref>{{Harvnb|Luen|Stark|2008}}. This was based on data from Southern California.</ref>

[[Image:Alarm dilemma.png|thumb|upright=1.75|The Dilemma: To Alarm? or Not to Alarm? It is assumed that the public is also warned, in addition to the authorities.]]
As the purpose of short-term prediction is to enable emergency measures to reduce death and destruction, failure to give warning of a major earthquake, that does occur, or at least an adequate evaluation of the hazard, can result in legal liability, or even political purging. For example, it has been reported that members of the Chinese Academy of Sciences were purged for "having ignored scientific predictions of the disastrous Tangshan earthquake of summer 1976."<ref>{{Harvnb|Wade|1977}}.</ref> Following the [[2009 L'Aquila earthquake#Prosecutions|2009 L'Aquila Earthquake]], seven scientists and technicians in Italy were convicted of manslaughter, but not so much for failing to ''predict'' the earthquake, where some 300 people died, as for ''giving undue assurance'' to the populace – one victim called it "anaesthetizing" – that there would ''not'' be a serious earthquake, and therefore no need to take precautions.<ref>{{Harvnb|Hall|2011}}; {{Harvnb|Cartlidge|2011}}. Additional details in {{Harvnb|Cartlidge|2012}}.</ref> But warning of an earthquake that does not occur also incurs a cost: not only the cost of the emergency measures themselves, but of civil and economic disruption.<ref>{{Harvnb|Geller|1997|loc=§5.2|p=437}}.</ref> False alarms, including alarms that are canceled, also undermine the credibility, and thereby the effectiveness, of future warnings.<ref>{{Harvnb|Atwood|Major|1998}}.</ref> In 1999 it was reported<ref>{{Harvnb|Saegusa|1999}}.</ref> that China was introducing "tough regulations intended to stamp out 'false' earthquake warnings, in order to prevent panic and mass evacuation of cities triggered by forecasts of major tremors." This was prompted by "more than 30 unofficial earthquake warnings ... in the past three years, none of which has been accurate."{{efn|1=However, {{Harvtxt|Mileti|Sorensen|1990}} have argued that the extent of panic related to public disaster forecasts, and the 'cry wolf' problem with respect to repeated false alarms, have both been overestimated, and can be mitigated through appropriate communications from the authorities.}} The acceptable trade-off between missed quakes and false alarms depends on the societal valuation of these outcomes. The rate of occurrence of both must be considered when evaluating any prediction method.<ref>{{Harvnb|Mason|2003|p=48}} and throughout.</ref>

In a 1997 study<ref>{{Harvnb|Stiros|1997}}.</ref> of the cost-benefit ratio of earthquake prediction research in Greece, Stathis Stiros suggested that even a (hypothetical) excellent prediction method would be of questionable social utility, because "organized evacuation of urban centers is unlikely to be successfully accomplished", while "panic and other undesirable side-effects can also be anticipated." He found that earthquakes kill less than ten people per year in Greece (on average), and that most of those fatalities occurred in large buildings with identifiable structural issues. Therefore, Stiros stated that it would be much more cost-effective to focus efforts on identifying and upgrading unsafe buildings. Since the death toll on Greek highways is more than 2300 per year on average, he argued that more lives would also be saved if Greece's entire budget for earthquake prediction had been used for street and highway safety instead.<ref>{{Harvnb|Stiros|1997|p=483}}.</ref>