Pages

Friday, March 20, 2015

Report claims Israel times its attacks when news media is distracted by other stories

Here's almost certainly another example of using statistics to come up with a predetermined conclusion. From CBS Marketwatch:

One of the Israeli government’s secrets to manipulating the American media has been revealed for the first time by new research.

Israel habitually launches its most unpopular and, sometimes, deadly attacks against Palestinians in the West Bank and Gaza to coincide with big news events here in the U.S., so that they don’t get too much public attention, according to the study.

The news management is so sophisticated that the Israeli government is especially good about avoiding damaging “follow-up” or “day two” stories about its attacks — stories most likely to include awkward human interest details about the casualties and their families.

So finds a study conducted by Ruben Durante, professor at Sciences Po in Paris, and Ekaterina Zhuravskaya, professor at the Paris School of Economics.

The researchers looked at Israel’s military interventions in Palestine over an 11-year period, from 2000 to 2011, and then compared them to what was going on in the news at the time. That included looking at whether there was big other news, and whether that other news was scheduled — such as, say, the Super Bowl — or unscheduled, such as an earthquake or tsunami or plane crash somewhere in the world.

“We find that Israeli attacks are more likely to occur prior to days with very high news pressure driven by clearly predictable events,” they found. There were statistically significant upticks in Israeli military action in the West Bank and Gaza Strip before big holidays or sporting events, but not before things that the Israeli military could not anticipate.
The study is here.

While the statistical analysis is beyond me, if I am understanding this correctly they are basing the definition of whether something is a predictable event after the fact, based on keywords they found on news stories the day after Israeli attacks compared to keywords found on other "high news pressure" days.

They are claiming that news days that the difference between the two charts shows that Israeli attacks coincide with predictable new events. Israeli leaders cannot predict hurricanes or earthquakes, but they can anticipate elections.

I think that this analysis is flawed.

Firstly, keywords like "campaign" and "war" and "Bush" do not necessarily correlate with known predictable news events.  While the paper mentions things like the Super Bowl as a predictable high-pressure news event, it isn't mentioned on this list.

Secondly, it seems like their definition of "high news pressure" is very narrow. If words like "Katrina" and "tsunami" are showing up on the second list, that indicates that over the eleven year period of the study there were relatively few high-news pressure days, perhaps only a few every year. This would indicate that the sample size is very, very small to come up with these conclusions. Notice that "Obama" doesn't even show up as a keyword on high pressure news days.

It also indicates that high-pressure news days cannot be predicted. Elections can be but the Florida story of 2000 clearly could not be. (The study dates started in 2001, so perhaps the stories were about publication of the recount. I don't recall that news being as wall-to-wall as hurricanes.)

Which means that the basic assumptions of the analysis are wrong. One cannot look at the keywords after the fact to determine what the Israelis would have been able to predict beforehand would become a hot news story. To say that the keywords in the first list imply predictability seems not at all scientific.

It seems to me that a proper methodology for a study like this would be to first generate an impartial list of predictably important US news events - Super Bowl, Oscars, major primaries, national elections, New Year's Eve - and then try to correlate Israel's actions against those, not to ex post facto determine that the word "Florida"is coming from a predictable news event.

You cannot claim that things are predictable by looking at their keywords afterwards. And if the number of high-pressure news days is as small as I think it is, then there is no way that Israel's leaders could ever predict what would be a major news story and what won't - see again how the Super Bowl does not appear on this list.

These is some slight evidence in the report that Israeli strikes might have been more likely to have occurred during holidays, but I think that someone with some real statistical expertise look at the report altogether and see if its methodology is rigorous or, as I suspect, sloppy.

(h/t Nephew of Ziyon)