Channel 4 News watch

I'm a fan of C4 News. At its best, it can be offbeat, prepared to offer a different perspective, and at ease with a diverse and multiethnic Britain. This occasional blog, though, will be largely devoted to the matter which grates on some viewers even more than some of Jon Snow's ties: the programme's tendentious reporting of Middle Eastern politics.

Friday, October 13, 2006

On those Iraqi death rates ...

Okay, nowt to do with Channel 4, but just a bit of number-crunching. What was the death rate in Iraq before the 2003 war? The newly-released Lancet study produces a figure of 5.5/1000. To support this figure, the report's writers cite the CIA World Factbook. This is despite the fact that the CIA's figures for more recent years remain at around the same level, which, if true, would mean that virtually no excess deaths have resulted from the war.

Jon Pederson, whom the Lancet report praises, disputes the 5.5 figure here, citing a higher UNICEF figure.

Other UN bodies seem to concur. For instance, the WHO has figures for 2000-1 here. (Hat tip: A2) If you scroll down to 'Iraq', select '2001', and press 'Go!", then you'll bring up a table of death rates separated by age (irritatingly, this doesn't have a separate URL of its own):

The first three columns are the only ones you need to look at. The third column, 'nMx', is the second divided by the first, so it's the death rate for each age group.

I put this into a spreadsheet, added together the actual populations and the actual deaths, and got this:

As you can see, the WHO's overall death rate for Iraq in 2001 is 9.03/1000. But this, again, is quite a lot over the 5.5/1000 figure (for the following year) produced by the Lancet study.

And not surprisingly. Over one third of the 2001 deaths were of infants under 1 year old, and we are often told that children were bearing the brunt of the sanctions. The Lancet study says so itself, in Appendix E of its companion document. Here's a graph you can find there:

Now, it's true that the year axis doesn't go beyond 1998; but the mortality rate lines seem to do so. And nothing in what the rest of the Appendix says would lead you to believe that things had changed much for the better after 1998.

In the comments section to this post at Harry's Place, Tim has some thoughts on the strange implications of this:

if a 10% infant mortality rate applied in Iraq (A country of 25 million with 35 births per thousand) This would give a total of 85,000 deaths.
The Lancet Report claims a 5.5 crude death rate.
A total of 137,500 deaths.
If 85000 of those are under one year olds then the rest of the populataion had a crude death rate of 2.1%.

the lowest in the entire world ( by a factor of two)

And the UN's World Population Prospects study gives a rate for Iraq in 2000-2005 of 9.7/1000, and 10/1000 for the previous five years. (Hat tip: dd)

This WHO report, using the 2004 Iraq Living Conditions survey, gives a 2002 crude death rate figure of 7.8/1000.

If the point of all this needs to be spelled out, then here goes: if the pre-war figure was higher than 5.5/1000 - as all these UN figures seem to suggest - then the excess deaths must be lower.

Update: By the way, the 2004 Lancet study had an even lower pre-war crude mortality estimate, of 5/1000. Fred Kaplan criticised this figure at the time. You can read more debate about the figure here.


Blogger Heiko said...

As mentioned over on Tim Lambert's blog, very nice work!

I was thinking of taking the WHO figures and putting them into a spreadsheet myself, and decided it was too much bother.

As said there, I am now very sceptical of any set of figures for crude death rate.

What I am not sure about is what specifically is causing the inaccuracies.

The Lancet figures for "excess death" are so politically charged, because they well equate "excess death" with deaths caused by the coalition, which in turn easily is understood as people murdered by US soldiers.

Death rates by contrast don't have this baggage of culpability attached to it.

Look at the coverage the ILCS infant mortality numbers got say. Because nobody made an issue out of it, the fact that they showed a more or less constant rate over the last 20 years never made any headlines.

Therefore, there should be no author bias.

I realise that accusations of fraud are also heavily charged, but showing scepticism because of author bias is something different.

You can be perfectly honest, and unconsciously be making decisions heavily impacting a survey. Say, if you are sympathetic to the cause of insurgents, you might choose households that are disproportionately Sunni Arab, and disproportionately involved in insurgent activity.

That's just an example, generally, I see that there's lots of ways for things to go wrong, eg:

1. Choice of households
2. Explaining properly who should be counted as a household member at the time of death and who should not

5:17 AM  
Blogger Heiko said...

This comment has been removed by a blog administrator.

5:23 AM  
Blogger Lopakhin said...

Say, if you are sympathetic to the cause of insurgents, you might choose households that are disproportionately Sunni Arab

Or even - as someone else has mentioned elsewhere - if you *are* a Sunni Arab (and my understanding of the university which supplied the survey staff is that it is a Sunni institution, although I'm open to correction on that) - maybe you just tend to go to Sunni houses, as, given the insecurity there, you feel safer that way.

I see what you mean about the ILCS figures & its graph (figure 26 in my document).

As for the WHO figures, I'm going to try and find out where they come from. They do look quite precise. There are quite a few PDFs on the WHO site which I'm trying to plough through. Looking at this page:

It seems to say that they get these figures from the registration figures of the governments of the countries concerned. But I'm not certain about that - have to do a bit more reading on the subject. :o)

6:46 AM  
Blogger Heiko said...

I've tried it myself and have not been able to get good answers. Unicef have never replied to my email asking for information, and with WHO I didn't go so far as writing them a letter or an email.

I spent ages looking at the first Lancet survey and concluded that the numbers were far too imprecise to be useful. I think, even making all sorts of allowances for biases and errors, it made a case that violent death had gone up, but even that in the end I mostly believe because it's in agreement with what other sources of information tell us, not because I trust the survey.

So, I am not willing to waste any time looking at the new survey, I am far keener to find out more about mortality data in general, and how reliable surveys are.

For "war related" deaths, I accept the ILCS survey value as the best we've got in terms of survey data, though even with the ILCS, where I don't see political bias or the size of the survey as big issues, I am mistrustful of the reliability of the results.

I like it that they left it to the survey respondents to decide what is a "war related" death, even though that carries the problem that we aren't sure what Iraqis think is war related or not. But I think their judgment as to what is war related or not is preferable, as they are most directly concerned.

4:36 AM  
Blogger Lopakhin said...

On the ILCS, for what it's worth, I had occasion to write to Prof. Richard Garfield about this some time ago. Here's how he replied:

Some basics I can clarify:

Mortality rates clearly started to decline among children in Iraq in 2000 or 2001; they just didnt decline as much as they could have. The Min of Health in 2005 had only hospital data for comparisons, maeaning in a time of insecurity that no infomration was really available on the wider population of kids. Perhaps they are right that diarrheal death rates declined, I hope so; I havent seen these data.

The FAFO study
[i.e. ILCS] doesnt provide guidance on the question of child mortality rates. Their death reporting is so far from ALL OTHER FIELD_BASED SOURCES that it cannot be believed. No surpise, as it was a total of 4 qustions out of 100 pages of interview.

We are comparing the survey data, most importantly the 1999 UNICEF survey, with census data, to get a better view. In the latter years, 1995 - 1998, the two agree closely on mortality rates around 100 or higher, and a rise from pre-sancitons period of 10% or more. This is likely the best info we will have, unless the 2007 census is surprisingly well done.

I am acutally working on this issue right now and will eventually have a paper on it to share with you.


9:13 AM  
Blogger Heiko said...

Thanks, I see he doesn't believe them.

11:57 AM  

Post a Comment

<< Home