Kenya's Death Market

The Daily Nation publishes 30 deaths (on average) everyday. In publicly announcing the demise of the dearly departed ones, the bereaved intentions are untold. They may simply be informing relations and acquaintances on the passing of a mate, or myriad of other reasons such as fundraising for medical and funeral costs. Whatever the objectives, these intentions are an interesting study and an indicator on the expenditure on the death value chain. No better place offers information about intentions other than the obituaries. A gang of three Data Scientists ( Anthony Otieno, Emmanuel Kendagor, and I) collected 2017 obituaries data from The Daily Nation to examine the Kenyan death economy.

Data in the Papers

In a year, The Daily Nation will advertise an estimated 7,200 deaths. With a crude death rate of 5.8 per 1,000 people (2015), Kenya would have an estimated 272,600 deaths in 2017. Therefore, the deaths reported on the daily nation newspaper represent 2.6 percent of total deaths in the country. Thus, majority of deaths in the country are not announced through the newspapers. However, newspapers obituaries constitute what we referred to as ‘dominant deaths’— deaths that attract higher spending.

We begin by building features (variables) from the obituaries that are a pointer to expenditure such as word count, size (dimensions), color (black & white or colored), then to other discreet features e.g age, number of children, number of relatives, day of burial et cetera.

 

Straight to the newspapers, we translated the obituaries section to an Excel sheet. The resultant dataset had 17 variables — a snippet is shown below.

Intentions

Psychologists say everything we do is motivated by need, either subliminally or consciously. This brings us to the question,Why do we write obituaries? People have different reasons for writing an obituary. The main intention being to inform the deceased’s network of their passing. Another could be a call to fundraise and as a legal requirement just to mention a few. Completeness of a obituary may be an indication of intent.

In our study we defined a “complete” obituary as one with the following details.

  • Date of birth
  • Date of death
  • Burial date
  • Cause of death
  • Hospital of death
  • The morgue the body was kept
  • County of death
  • County of burial
  • Relations

With money being a major constrain, ,space becomes a scarce resource and one has to be utilised effectively by including variables that they deem most important. Only 13.8% of obituaries have a “complete” profile – we make the assumption that the 13.8% constitute obituaries that intend to merely inform public on death hence the provision of all details. There is a second category of the obituaries that explicitly have a call for fundraising. So we asked ourselves, what type of death would lead to a fundraising?

To that end we opted to build set up a classification task to determine factors that lead to a fundraised death. Our choice of algorithm is a regression trees. At the onset we opted to ignore location information since the good book ‘Weapons of Math Destruction’ 🙂 says location information introduces bias in prediction models. Straight into it we built the classification model below.

In the above model, the time between announcement of death on the newspaper and burial becomes the most important factor in deciding if an announcement will call for fundraising. If the difference between announcement and burial is greater than 2.5 days, there’s a 71% chance a fundraising will be included in the obituary. This shows that folks who put up their obituaries more days before the burial are always hinting at need of fundraising. Last minute obituaries on the other hand might have another unknown intention.

Folks with more relatives normally book smaller advertisement spaces and use a lot of words in the obituary which include a call for fundraising. It may seem that the inclusion of relative names in the obituaries is a subtle reminder for their contributions. This model however was only 60 percent accurate, so we opted to include more information to improve the accuracy – we went back to the location data and the resultant model is shown below.

Now the time between announcement and burials been displace by county of burial as the most important factor in predicting fundraising. Folks from Bomet, Bungoma, Embu, Isiolo, Kakamega, Kericho, Kiambu, Kirinyaga, Laikipia, Meru, Murang’a, Nairobi, Nakuru, Nandi, Narok, Nyandarua, Nyeri, Tana River and Trans Nzioa have a high propensity on not needing a fundraising. On the other hand, folks from Homa Bay, Kisumu,   Kitui, Machakos, Makueni, Migori, Nyamira, Siaya, Uasin Gishu and Vihiga have a 78 percent probability of requesting for fundraising.

This observation can be explained by the poverty index as shown below.

There may be an alternative explanation, i.e some communities are predisposed to funeral fundraising whether or not the deceased is wealthy or not. From our new model, the prediction accuracy improved to 70 percent with three factors (location, number of relatives, and time between announce and burial) having the highest prediction power. It is interesting that gender and age were poor predictors of fundraising.

This model can be useful for insurance companies selling life and last expense insurance products. According to a research done by Mburu John Murigi, the world’s insurance penetration is 6.5% with most of the penetration found in G7 countries at 65% penetration. Africa is still struggling with insurance penetration at a level of 3.65%. The Insurance Regulatory Authority (IRA) report of Kenya states Kenya’s insurance penetration at 2.8% for both life and non-life insurance. This shows that there is still quite a large market for Insurance companies in the country. A good strategy would be to target folks from the above mentioned counties who come from large families.

Expenses
Newspapers have different prices for various obituary size with the larger ones costing more than the smaller ones. For instance nation newspaper charges sh 25,000 for a 56cm2 obituary to sh 623,000 for a full page obituary. The mean size of an obituary published by The Daily Nation is 88.34cm2 consisting of an average 212 words. Going by their rates, the average cost per obituary comes to sh. 36,500.

Without factoring in the cost of meals and hiring of event planners, Kenyans invest about 170,000 per burial event. According to world population review a Kenyan dies every two minutes in which going by our data 88% of deaths are caused by illness. Kenyatta University morgue leads in body storage, storing 11% bodies of which 72% are to be buried the neighbouring counties of Kiambu, Murang’a, Nyeri and Meru counties.

Sammara

Once there was a wealthy merchant in Baghdad who had a servant that he liked so much. The servant came to him one day pale and trembling, and the merchant said, “Whatever is the matter?”. “Master” said the servant, “I saw death in the market today, and he pointed at me. I am sure that my time has come, but I think I can out run him … Please, master, give me a fine horse and the money to go to Sammara tonight.

The merchant said, “Of course. Take my best horse,” and he outfitted the servant with fine clothes and food besides. He saw the servant off from the gates of Baghdad. Then, as he was returning through the city, he ran into Death at the market. “My good fellow,” said the merchant sternly, “why did you point at my servant this morning? You scared him.” “Oh,” said Death contritely, “I’m quite sorry. It’s just that I was surprised to run into your servant here today, for I have an appointment in Samarra with him tonight.” — The Appointment in Samarra

Like the servant in the Appointment in Samarra Kenyans will die far away from home. Comparing the county of death and county of burial, 82 percent of Kenyans will die away from home. The youthful servant met his demise at home in Sammara but work took him in Bagdad – most Kenyans will meet their demise in their endeavours for work.

In the book Dracula by Bram Stoker, the writer states Denn Die Todten Reiten Schnell” (for the dead travel fast) – in Kenya they ride a bit slow, on average 9 days before they are buried.

Advertisements
  1. As always, that was a very good read. Best part, tying it to the practicality of selling life insurance and using numbers to justify why we take long to bury our loved ones. So something I have always been curious about is how do our religious beliefs affect these burial traditions. For instance in your sample, were there obituary entries for Islam or Hindu families?

    Like

    Reply

    1. The data is majority christians. Muslims don’t put up obituaries and also bury in 2 days. There were a few obituaries of hindus but not enough to draw separate conclusions.

      Like

      Reply

  2. Entertaining and thought provocative. Keep up the good work. You are an inspiration to many.

    Like

    Reply

    1. Thank you for the kind words.

      Like

      Reply

  3. Nice one Orwa!
    It’s interesting how this can be used by the insurance companies to create their models.

    Like

    Reply

    1. Thank you. I believe they can extend the experiment and use it to build insurance models.

      Like

      Reply

  4. Men, you are a serious data scavenger….any unstructured data can be analyzed to make sense

    Like

    Reply

    1. That’s the idea.

      Like

      Reply

      1. I long to go down to the alleys with you and become a data junkie too.

        Like

  5. I like. But Uasin Gishu is among top 10 counties but is likely to ask for fundraising. Consider also how important burials are in various cultures, also where (geographically) the relatives are (South Africa, USA) etc.
    Interesting analysis.

    Like

    Reply

  6. Very intriguing summary. Consider people whose ancestral home is far away from Nairobi incur greater costs for transporting the body and family to the burial, hence the need to advertise and fundraise. I’d put the average cost at about 300,000 to 500,000 for a burial.

    Like

    Reply

    1. Thank you. Travel to ancestral home is indeed a good pointer to fundraising – I couldn’t get the average cost for hearse.

      Like

      Reply

  7. Interesting, informative and intriguing. Always on point, big admirer of your Data Science Work.

    Like

    Reply

    1. Thank you

      Like

      Reply

  8. I have always loved the way you interact with data. On a more practical point, there is need for another morgue away from KU along the Thika superhighway. KU was apt for people in the Mt. Kenya region but traffic is becoming an issue.

    Like

    Reply

    1. Thank you. Thika town would be a good location.

      Like

      Reply

  9. Interesting analysis. But for many of the counties more likely to fundraise, especially the Luo and Luhya counties, it isn’t really poverty or distance. But the culture and ceremony of death that has been perfected over decades. Even if the family is wealthy, they will organize the nightly vigils and fundraising events. I have observed that countless times.

    Like

    Reply

    1. That’s correct.I made that argument in the article.

      Like

      Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

Advertisements
Advertisements
%d bloggers like this: