Workplace deaths in Bangladesh

Dear [redacted],

I have read your documents with keen interest and I stand with you in solidarity on the anniversary of the Rana Plaza tragedy; your organization seems to be making the world a better place. In your last email you mentioned that your organization didn’t have an IT person. Without promising anything, and with the caveats that they wouldn’t reside in Bangladesh, and might have to set up a new domain name if your IT person didn’t leave anything behind, do you want new webmaster, perhaps pro bono?

I have paid particular attention to your estimate of fatal workplace accidents, and their methodology. I do not expect the problems with relying on news reports to be new to you, but perhaps their extent might surprise you: I think you’re missing deaths, perhaps as little as 1000, and perhaps as much as 4000, or more.

I get the 1000 number from the Statistics Ministry of Spain for the year 2017 (, page 2. “Mortales” = Fatal, “In itinere” = “during transit”). In 2017, 496 people died in workplace accidents in Spain, and an additional 133 died in their way to work. Spain happens to be a developed country, which has reasonable workplace safety standards, and a statistics ministry I consider trustworthy. If you want to know more about how these statistics are compiled, I could research that.

Spain has a population of circa 46.5 million inhabitants, and Bangladesh has a population of circa 164.7 million people, so if Bangladesh were just a bigger version of Spain, by which I mean, if Bangladesh had just the same number of fatal accidents per million people because it had the same general conditions, you’d expect to see 496 * 164.7 / 46.5 = 1 757 deaths in Bangladesh every year. Furthermore, if you include deaths in road accidents while on the way to work (as you do in your reports), that number inflates to (496 + 133) * 164.7 / 46.5 = 2227 deaths. But Spain is a much richer country than Bangladesh, and its workplace standards are better, so that number must be an underestimate.

This article ( gives the workplace death toll for China: 38 000 during 2017. If Bangladesh had instead similar conditions to China (and if China’s numbers could be trusted), you’d expect to see, by the same process as above, 38 000 * 164.7 / 1 386 = 4516 deaths in Bangladesh, per year.

To a first approximation, you’d expect Bangladesh’s death rate to be somewhere between those two numbers. That is why I have been extremely surprised and confused to see that your estimates are much lower: 951, 1240, 1242 and 898 for the previous 5 years. This would seem to imply that Bangladesh is a safer country than Spain, which to my eyes seems unlikely. It could be the case that your estimates are so inexact that any variation in your numbers is due to the noise, the inaccuracy, the error of your estimation + really big accidents, and not due to any change in the conditions from year to year.

For this reason, I think that the raw absolute number might be very misleading. However, I do not think that your estimates are useless. In particular, I think that with some statistical analysis, some useful conclusions might nonetheless be extracted. It would be an honor for me to inaugurate your as yet empty Research Collaboration tab. Would it be possible for you to send me all the data you have, perhaps (Name of the accident or identification - Number of deaths - Number injured - Year - Sector) for each accident?

I am still, however, extremely worried about whether your estimates might inaccurate or misleading. In this article:, decreasing number of deaths are given as a reason for ejecting international safety inspectors:

Though around half the factories that make up $29bn industry are not covered by either agreement, deaths across the sector have fallen in the past five years from about 71 per year to 17, according to research from New York University’s Stern Centre (Comment: Can these figures be traced back to your organization?).

The government says its industry has now become safer than those of its competitors in Vietnam, China and India, and control must be handed back (Comment: is this actually true?)

I’m not confident at all in this last part, that is: that your numbers could be used to justify politically predetermined decisions; I do not claim to be very familiar with Bangladeshi politics. But it would be possible.

To conclude, even though the information I can get from reading reports without being there is somewhat limited, your organization seems to overall be doing superb work: to give just one example, I think that banning asbestos seems to be a particularly impactful goal. More broadly, “Organizing the unorganized”, as you put it on your 2015 report, is a worthy undertaking. My overall view of your organization is extremely positive.

One of the organizations which I most admire, Evidence Action, recently carried out a large experiment to test one of their programs, and they found out that they didn’t have a significant impact ( Another organization which I greatly respect, Give Well, has a mistakes page (, some of which are even embarrassing. I respect them more for admitting their mistakes and implementing measures to correct them, not less, and I think that’s part of why they’re great organizations, and not just merely good ones. For this reason, if anything of what I said above turns out to not be true, or to stem from confused thinking on my part, I’ll gladly add another item to my log of mistakes.


Nuño Sempere