WEBVTT

1
00:00:02.500 --> 00:00:04.750 line:16%
How many disabled people live in Nigeria?

2
00:00:04.940 --> 00:00:07.190 line:16%
You often hear thirty-five million.

3
00:00:07.360 --> 00:00:11.510 line:16%
That's a figure from the Joint National
Association of People with Disabilities in

4
00:00:11.560 --> 00:00:16.250 line:16%
2025, which itself is an estimate from the
global prevalence of fifteen percent of

5
00:00:16.300 --> 00:00:17.650 line:16%
the population in Nigeria.

6
00:00:18.100 --> 00:00:22.630 line:16%
That global rate was found in a study
by the World Bank and the World Health

7
00:00:22.680 --> 00:00:23.910 line:16%
Organization in 2011.

8
00:00:24.340 --> 00:00:25.910 line:16%
My name is Jerry Bannister.

9
00:00:25.960 --> 00:00:28.610 line:16%
I'm a statistician in Bauchi
and a wheelchair user.

10
00:00:28.660 --> 00:00:31.760 line:16%
Come along with me to see what
the surveys actually measured.

11
00:00:31.966 --> 00:00:35.896 line:16%
Nigeria's 2024 Demographic and Health
Survey asked the Washington Group

12
00:00:35.946 --> 00:00:38.336 line:16%
questions. That's the
international standard.

13
00:00:38.646 --> 00:00:42.796 line:16%
Three in a hundred adults have a lot of
difficulty or completely lack the ability

14
00:00:42.846 --> 00:00:44.696 line:16%
to engage in seeing, hearing,

15
00:00:44.746 --> 00:00:48.616 line:16%
walking, remembering, caring
for themselves or communicating.

16
00:00:49.066 --> 00:00:52.816 line:16%
The official report, counting
those with any difficulty at all,

17
00:00:52.866 --> 00:00:56.416 line:16%
gets about fifteen in a hundred
people that are aged fifteen and over.

18
00:00:56.466 --> 00:00:58.116 line:16%
But that's a different question.

19
00:00:59.246 --> 00:01:03.476 line:16%
The report already publishes state
figures for men and women separately,

20
00:01:03.526 --> 00:01:05.276 line:16%
without any margins of error.

21
00:01:05.506 --> 00:01:07.996 line:16%
I combined them into one figure per state,

22
00:01:08.046 --> 00:01:10.196 line:16%
which roughly doubles each state's sample,

23
00:01:10.246 --> 00:01:14.896 line:16%
and gave every state a confidence interval
using the survey data and sampling

24
00:01:14.946 --> 00:01:15.464 line:16%
weights.

25
00:01:15.504 --> 00:01:17.474 line:16%
The states are not alike.

26
00:01:17.524 --> 00:01:20.554 line:16%
In Taraba, it's about eleven in a hundred.

27
00:01:20.604 --> 00:01:23.454 line:16%
In Lagos, it's under one in a hundred.

28
00:01:23.504 --> 00:01:26.104 line:16%
Seven states are clearly
above the national figure,

29
00:01:26.154 --> 00:01:29.354 line:16%
and twelve clearly below after
correcting for multiple testing.

30
00:01:29.404 --> 00:01:31.514 line:16%
But two things I expected didn't hold.

31
00:01:31.564 --> 00:01:34.854 line:16%
Neighbouring states don't show
a clear pattern of being alike.

32
00:01:34.904 --> 00:01:37.074 line:16%
A spatial test was conducted,

33
00:01:37.124 --> 00:01:41.274 line:16%
Moran's I, and it found none of
the claims survives correction.

34
00:01:41.784 --> 00:01:46.474 line:16%
As a follow-up, a small-area model meant
to sharpen the state figures failed the

35
00:01:46.524 --> 00:01:49.514 line:16%
checks I set in advance,
so I also left it out.

36
00:01:50.554 --> 00:01:53.124 line:16%
So, does disability go with poverty?

37
00:01:53.594 --> 00:01:57.044 line:16%
I registered the plan publicly
before I split poverty by disability.

38
00:01:57.594 --> 00:01:59.124 line:16%
Our results were expected.

39
00:01:59.394 --> 00:02:02.494 line:16%
Using double machine learning
on the national poverty survey,

40
00:02:02.544 --> 00:02:06.584 line:16%
people in households with a member who
has severe difficulty are about twelve

41
00:02:06.634 --> 00:02:10.734 line:16%
percentage points more likely to be poor
on the national multidimensional poverty

42
00:02:10.784 --> 00:02:12.984 line:16%
index than people in
comparable households.

43
00:02:13.034 --> 00:02:15.484 line:16%
The range is six to seventeen percent.

44
00:02:15.534 --> 00:02:19.064 line:16%
When I shuffled who was
disabled a hundred times,

45
00:02:19.114 --> 00:02:21.604 line:16%
the fake results averaged
almost about zero.

46
00:02:21.749 --> 00:02:24.039 line:16%
But I'll be honest about the weak spots.

47
00:02:24.109 --> 00:02:27.139 line:16%
In the worst case, a hidden
factor I didn't measure,

48
00:02:27.189 --> 00:02:29.719 line:16%
even a small one, could explain it away.

49
00:02:29.809 --> 00:02:33.539 line:16%
A factor as strong as the head of
household's education would barely move

50
00:02:33.589 --> 00:02:36.739 line:16%
it, but it would be impossible
to rule out something stronger.

51
00:02:36.789 --> 00:02:41.659 line:16%
So it's an association,
and not proof of cause.

52
00:02:41.709 --> 00:02:43.609 line:16%
A smaller panel survey looks at work.

53
00:02:43.659 --> 00:02:47.179 line:16%
Among people whose severe difficulty
began between 2018 and 2023,

54
00:02:47.229 --> 00:02:51.899 line:16%
the proportion working fell about
twenty-six points more than among

55
00:02:51.949 --> 00:02:55.819 line:16%
comparable people whose difficulty
didn't begin during that time period.

56
00:02:55.869 --> 00:02:59.359 line:16%
That part is exploratory, and
it can't say which came first.

57
00:02:59.419 --> 00:03:04.489 line:16%
A lot of what I tested wasn't
significant, and the site says so.

58
00:03:04.539 --> 00:03:07.509 line:16%
The state chart tested
seventy-three comparisons:

59
00:03:07.559 --> 00:03:11.869 line:16%
neighbouring states and each state were
tested against the national figure.

60
00:03:11.999 --> 00:03:15.569 line:16%
Only about twenty held up after
corrections for multiple testing.

61
00:03:15.619 --> 00:03:20.369 line:16%
Taraba, being the highest, can't be told
apart from Abia, the second highest.

62
00:03:20.639 --> 00:03:23.169 line:16%
Looking at where the
poverty difference sits,

63
00:03:23.219 --> 00:03:25.389 line:16%
the deprivation gap is clear for health,

64
00:03:25.439 --> 00:03:30.129 line:16%
for work and shocks, but the survey
can't show one for education or living

65
00:03:30.179 --> 00:03:34.589 line:16%
standards. The panel survey also fails
to show a difference in how household

66
00:03:34.639 --> 00:03:35.969 line:16%
consumption changed.

67
00:03:36.139 --> 00:03:42.209 line:16%
As a check, the site can't deploy unless
every page and its types score a hundred

68
00:03:42.259 --> 00:03:44.259 line:16%
on the Lighthouse accessibility checks.

69
00:03:44.785 --> 00:03:50.710 line:16%
NVDA: The DHS 2024 survey estimates that
3.0 percent of adults aged 18 and over

70
00:03:50.750 --> 00:03:56.949 line:16%
living in households in Nigeria (95
percent confidence interval 2.8 percent to

71
00:03:56.989 --> 00:04:03.106 line:16%
3.2 percent) have a lot of difficulty
with, or cannot do at all, at least one of

72
00:04:03.146 --> 00:04:04.845 line:16%
six everyday activities.

73
00:04:05.185 --> 00:04:11.267 line:16%
NVDA: Dot plot: Disability among adults 18
plus by state, with 95 percent confidence

74
00:04:11.307 --> 00:04:17.545 line:16%
intervals. 37 dots, one per state, each
with its 95 percent confidence interval.

75
00:04:17.885 --> 00:04:22.545 line:16%
NVDA: The main result, region,
The main result, heading, level 2

76
00:04:22.585 --> 00:04:29.168 line:16%
People in households with a member aged 5
or older who has a severe difficulty are

77
00:04:29.208 --> 00:04:34.559 line:16%
11.7 percentage points more likely to be
multidimensionally poor than people in

78
00:04:34.599 --> 00:04:41.745 line:16%
comparable households (95 percent
confidence interval 6.1 to 17.2 points).

79
00:04:42.085 --> 00:04:44.635 line:16%
This is my open-source Python library.

80
00:04:44.785 --> 00:04:47.975 line:16%
It's called disability-access-charts.

81
00:04:48.025 --> 00:04:51.175 line:16%
It builds each chart with a
description written from the data,

82
00:04:51.225 --> 00:04:54.895 line:16%
and then checks that text against
the data, with intervals and all.

83
00:04:54.945 --> 00:04:56.035 line:16%
Let's take a look.

84
00:04:57.045 --> 00:05:00.295 line:16%
The check: “Taraba has a
higher rate than Lagos” passes,

85
00:05:00.365 --> 00:05:03.335 line:16%
but “Taraba has a higher
rate than Abia” is rejected.

86
00:05:03.725 --> 00:05:08.075 line:16%
The gap is small, and the
uncertainty doesn't allow it.

87
00:05:08.125 --> 00:05:11.825 line:16%
The site's descriptions are built only
from the differences that held up,

88
00:05:11.875 --> 00:05:15.675 line:16%
and this check runs in strict mode on
every chart before the site is built,

89
00:05:15.725 --> 00:05:16.795 line:16%
as a second guard.

90
00:05:17.195 --> 00:05:19.225 line:16%
Every chart describes itself,

91
00:05:19.275 --> 00:05:23.285 line:16%
and my open-source library checks the
numbers and comparisons in those words

92
00:05:23.335 --> 00:05:25.485 line:16%
against the data before the site is built.

93
00:05:25.535 --> 00:05:28.265 line:16%
It's free, every figure names its source,

94
00:05:28.315 --> 00:05:31.465 line:16%
and it hasn't yet been tested
by a regular screen-reader user.

95
00:05:31.555 --> 00:05:34.865 line:16%
If you use one, tell me
what works and what doesn't.

96
00:05:34.915 --> 00:05:35.915 line:16%
Thank you.
