1 Census Evaluation in Kenya By M.G. Obudho & J. K. Bore Kenya National Bureau of Statistics.

1

Census Evaluation in Kenya

By

M.G. Obudho & J. K. BoreKenya National Bureau of Statistics

2

History of Census in Kenya

7 censuses including the 1948 and 1962 Post Independence Censuses:

1969, 1979,1989,1999 & 2009 Census Night maintained as 24/25th

August Success of consistencies attributed to

Gvt commitment and funding

3

History of Census Evaluation

First PES attempt was after 1989 Census, but was not implemented

Second attempt was made in 2000, after 1999 census

The 2000 PES was conducted but was not conclusively finalized

The next PES is scheduled for November 2009 to evaluate the August, 2009 Census

4

Kenya 2000 PES

Main objectives were to evaluate the coverage and error rates for 1999 Kenya Population and Housing census

Covered 7 out the 8 provinces in Kenya. North Eastern Province (2% of Kenyan

pop) was not covered due to nomadic nature)

5

Design for 2000 PES

Domains of estimation: National, Urban,Rural and 7 provinces

A single stage stratified cluster sample design Stratification by province and urban/rural Used the EAs as PSUs from the cartographic

mapping done prior to the 1999 Census All households within the sampled EA were

covered Sample size of 500 EAs out of the 62,000 EAs

realized

6

Sample distribution for 2000 PES

Serial Province Sample EAs

1 Nairobi 602 Central 783 Coast 504 Eastern 805 Nyanza 806 Rift Valley 947 Western 58 Total 500

7

Personnel for 2000 PES

The Central Bureau of Statistics, CBS (presently Kenya National Bureau of Statistics, KNBS) conducted the exercise

The technical officers from the Bureau HQs who were key in census managed PES

8

Personnel for 2000 PES cont….

Supervisors were chosen from those who worked well in census

Enumerators were picked from those who excelled in census & new ones

9

RESULTS OF 2000 PES

Though not fully concluded, the results were still used to adjust the census results

Vital lessons were learnt from it to inform future similar exercises

10

Challenges in 2000 PES Lack of independence for PES and census

personnel Lack of full integration of PES and census in

terms of planning and funding Lack of technical capacity at the Bureau to

handle the PES PES was done six months after census

which led to low match rate & memory recall Manual matching was used and was very

tedious and slow

11

Plan for the 2009 PES

The next PES is scheduled for November 2009 to evaluate 2009 census

Pilot PES will be conducted a month prior to PES

Capture by scanning the names of census respondents in the PES sampled EAs

12

Objectives of 2009 PES Quantitatively evaluate accuracy of the census in terms of

coverage and content error at the national, urban/rural and provincial levels.

Provide, if necessary, concrete statistical basis for adjustment of the census data

Act as a basis to provide and document lessons learnt for improving and implementing future censuses and surveys.

Provide information on erroneous enumeration and omissions during census enumeration

Enable KNBS staff to gain experience and enhance skills in census evaluation

13

The proposed Design for 2009 PES

Sample size was calculated based on:

- achieving a CV of 5% at province level

- a census omission rate of 5%

- a 10% adjustment for non-response

14

Design 2009 PES cont…….

Cover all the country’s 8 provinces A stratified single stage cluster sampling

design Stratification by provinces Sampling of EAs by PPS from about

96,000 EAs realized in 2009 census A total of 298 EAs are targeted An equal allocation to all provinces made

15

Design 2009 PES cont…….

On the design, two other options are being explored

In those options we may oversample the ASAL areas

Availability of funds will be the main determining factor for implementing any of these two other options

16

Sample distribution for 2009 PES

SRN ProvinceSample

(Pop)Pop per

Hh

Sampling Rate

(Hhlds)Sample

EAs

1 Nairobi 16,720 3.3 0.51 51

2 Central 16,720 4.0 0.35 42

3 Coast 16,720 4.7 0.49 36

4 Eastern 16,720 4.8 0.27 35

5 North Eastern 16,720 6.5 1.20 26

6 Nyanza 16,720 4.5 0.29 37

7 Rift Valley 16,720 4.7 0.17 36

8 Western 16,720 4.8 0.36 35

Total 133,760 4.5 0.31 298

17

Matching

Computer matching and supplemented with the manual one will be employed

Working closely with the US Census Bureau to identify the computer program

18

Concerns

If PES results suggests call for adjusting census figures downwards for a region, how do you deal with the wrath of the politicians from the affected areas?

If the PES results are not correct then any action to adjust census figures is dangerous; how do we safeguard against getting misleading results?

19

Ahsante Sana!

Thank You!