In my The Fiji Times article last week, I commended the Fiji Bureau of Statistics for convincing the Government to publish the Household Income and Expenditure Report.
I commended the Bainimarama Government for releasing the poverty facts to the public, regardless of their political sensitivity. But I was clearly wrong on both counts.
The reputable Australian Bureau of Statistics collects data on ethnicity, such as on Aboriginal people or Asians or those Australians born in Fiji, even though all Australians are treated equally under the law.
No Government statistician in Australia ever loses his or her job because the ABS freely publishes such ethnicity data.
The reputable NZ Stats organisation collects all kinds of data on ethnicity, such as Maoris or Samoans or Asians or residents born in Fiji, even though all New Zealanders are treated equally under the law. No Government Statistician in NZ ever loses his or her job because NZ Stats publishes such data.
Last week, the Fiji Bureau of Statistics (FBS) published a Report on the findings of the 2019-20 Household Income and Expenditure Survey, with some fascinating statistics, which also included poverty and food poverty tables by ethnicity.
The next evening, the most powerful Minister, Aiyaz Sayed-Khaiyum, gave a long press conference attacking the FBS for contradicting the Bainimarama Government’s policy to not collect or publish data by ethnicity and religion.
But much worse, the minister slammed the statistical integrity of the household survey by the FBS, alleging that the ethnicity statistics were not reliable and they lacked “integrity”.
The minister ranted that the FBS could not make ethnic generalisations from a mere 3 per cent sample of 6000 households, supposedly representing more than 190 thousand households.
Soon after, the CEO of FBS (Kemueli Naiqama) was escorted out of his office and his employment terminated.
The Fiji public, of all ethnicities and political persuasions, need to register their public protest that an honest civil servant who was trying to do his job with integrity is being sacked for producing accurate official statistics for the benefit of all Fiji.
I won’t repeat here the arguments I have made in my previous articles why it is absolutely essential for development policies that statistics be gathered and publicised by ethnicity.
iTaukei and Indo-Fijians have many different development characteristics with respect to business involvement, fertility, health, aging, education, housing and other important things. Just as in Australia, the ABS recognizes that Aboriginals have different development needs from white and Asian Australians and in New Zealand NZ Stats recognises that Maori and Pacific Islanders have different development needs from Europeans and Asians.
In this article, I wish to rebut the baseless assertion by the Attorney-General that the 2019-20 HIES lacked statistical integrity because it was a “mere” 3 per cent sample, all because he does not agree ideologically with the release of statistics by ethnicity.
I call on the Fiji Public Commission and the Fiji Public Service Association to legally defend the rights of the civil servant concerned against this arbitrary and unjustifiable punitive action by the Bainimarama Government.
What 3 per cent sample?
When I first started working with the FBS household surveys, even I was skeptical whether such a small sample as 3 per cent could give accurate results.
But to my continued amazement, I found that the statistical results were astonishingly accurate – not the least because of the rigorous methodology of the statisticians in the Household Survey Unit of the FBS.
The Australian Bureau of Statistics has a wonderful webpage titled “Sample Design” which Fiji’s tertiary students and academics should read. Just google “ABS Sample Design”.
For numbers of households above 100,000, the ABS statisticians recommend that even a sample of 1000 households (1 per cent sample) can be more than enough.
Some might remember that I have been previously critical of the results of opinion polls like the Tebbutt Poll which questioned “only” 1000 persons in the Nausori-Suva- Lami-Sigatoka-Nadi-Lautoka-Ba-Rakiraki corridor.
But even then, the Tebbutt results were pretty accurate about what Fiji’s voters thought about Bainimarama and his political opponents.
But the FBS selects 6000 households, not just far more than necessary, but rigorously and proportionately selected from all over Fiji.
The FBS 3 per cent
Students, take the time to read this section (don’t be lazy).
The FBS statisticians first obtain a population “frame” of households which covers 100 per cent (yes, all) of all Fiji households, usually from their records from the previous Census.
Then, the FBS selects a smaller “list” of households (around a third) whose households contain all the different “strata” or “variables of interest” on which the survey wishes to obtain information: the four divisions (Central, Western, Northern, Eastern); and the urban and rural components within them (seven strata altogether because Eastern is just Rural).
Then from this “list” is proportionately and randomly drawn the smaller 3 per cent sample of 6000 households: wherever in Fiji they happen to fall.
By this time these 6000 households, out of the national 192,000 households, are also representative of the 14 provinces and the major ethnic groups (iTaukei, Indo-Fijians, and Others) and other variables analysed at length by the Bristol University group.
This is a multi-dimensional analysis of poverty; with respect to heads of households, aspects such as gender, marital status, number of children, etc. Or aspects of the house such as the number of rooms, toilets, access to running water, etc.
The hard slog by FBS staff
There then follows the incredibly hard slog by the FBS trained interviewers who physically go to every selected household wherever they are: in the towns, villages, hills, valleys, by foot, by horse, by foot.
They ensure that all the questionnaires are fully explained to the household head and filled at the beginning with critical data about the household: its physical characteristics, its occupants, its annual income and expenditure, etc. and more.
For the 2019-20 Household Income and Expenditure Survey, each household then also fills out a two-week diary of incomes (all types — cash, crops, livestock, fish, gifts, remittances etc.); and all expenditures (cash, “home consumption”, gifts) etc. These are all checked by the interviewer at the second visit.
Then all 6000 records are collated at FBS HQ, blanks or obvious errors clarified and corrected, and the data entered into computer programs.
This data is then “cleaned” and analysed by the FBS statisticians and for this particular survey by experts from the World Bank and the Bristol University group (and once upon a distant time, by a certain unnamed USP economist).
The A-G, in his virulent press conference diatribe (which a blogger called appropriately a “hissy fit”) gave no hint whatsoever that he understood an iota of the enormous technical effort of the FBS staff in conducting this year long HIES (from February 2019 to February 2020).
Nor did he give a slightest hint that he understood the enormous statistical rigour of this FBS exercise, which he contemptuously labelled as “lacking in statistical integrity”.
The population cross-check
From my personal experience in analysing at least five household sample surveys by the FBS (the HIESs of 2002-03 and 2008-09, and the Employment and Unemployment Surveys of 2004-05, 2010-11 and 2015-16) I have always been blown away by the incredible accuracy of the results for the national picture.
Before conducting the detailed analysis by different variables of interest, all analysts first check whether the national aggregates derived from the survey match the national totals from alternative sources.
The first is the total population estimates from the survey, and that from the census and the resulting population projections; and also components such as rural-urban and ethnic components.
In all the surveys whose raw data I have analysed, the differences have been less than1 per cent in the total population, or for rural- urban numbers or by ethnicity.
The components from the 2019-2 HIES estimates all match what the demographers have projected for 2019-20 for rural-urban (45 per cent and 55 per cent); iTaukei/Indo- Fijian (62 per cent and 34 per cent); as well as all the other variables such as divisions and provinces, and even the age group divisions (including those of voting age (18+).
The only difference has been that the household survey estimates do not include the non-household “institutional” populations of the army, police, prisons, school and college dormitories so the HIES estimate of Fiji’s population will usually be less than the estimates or projections derived from the census estimates.
The independent FNPF cross-check
A totally independent check that I suspect even the FBS is not aware of is the 2015- 16 Employment and Unemployment Survey (EUS) estimates of how much should have been paid to the Fiji National Provident Fund for 2015-16 (an average of $454 million).
The annual reports of FNPF for 2015 and 2016 given an actual average total FNPF contribution received of $456m, or less than 1 per cent different from the EUS estimates.
These FBS 3 per cent sample surveys can also give you accurate estimates of how much FNPF is NOT being paid (read my FWRM Report soon to be published, if you are interested in how much more FNPF ought to be receiving- it is not a small sum!). There are all kinds of other similar crosschecks possible from HIES data (and from
EUS data).
The HIES estimates are accurate
Contrary to the A-G’s ideological assertion, the estimates of the incidence of poverty are absolutely accurate, given the World Bank methodology the FBS has chosen to use for poverty analysis — and which the A-G allegedly agreed with in February of this year when the national estimate of 29 per cent as the incidence of poverty was released.
The A-G did not then raise a single boo about the statistical methodology of the FBS when he spoke at length on how this incidence of poverty could not be compared with previous estimates in 2008-09 or 2002- 03 because the criterion had changed from income to consumption expenditure (all agreed).
But the all-knowing A-G in his grand press conference speech pontificated that we should all realise that some households consumed bhindi or bhaji or cassava that they grew themselves, and this was all included in the consumption criterion but he alleged not in the income criterion (FT 16/9/2021).
Hilariously, the lawyer A-G demonstrated his statistical ignorance that even in the “income criterion” previously used by the World Bank or the FBS or me, there was always a component of income called “home consumption”.
This comprised precisely what the household grows (vegetables or fruit) or raises (cattle, goats or poultry) or catches (fish and marine foods) for its own consumption, in addition to what they buy using cash.
But suddenly in September 2021, the A-G wakes up to the FBS’s “lack of statistical integrity” when exactly the same HIES data (which gave the national estimates he approved in February 2021) are used to provide estimates of poverty by ethnicity (useful for policy in my opinion) and religion (useless in my personal opinion).
Remember Abraham Lincoln
I wonder what the real economist ministers in the Bainimarama Government, such as Dr Mahendra Reddy and Jone Usamate, are thinking about the A-G’s latest diversion from national discussions about the horrible impact of the COVID pandemic.
But they probably are chuckling inside, thinking about what Abraham Lincoln is thought to have said: “Better to remain silent and be thought a fool than to speak up and remove all doubt”.
- Prof WADAN NARSEY is an Adjunct Professor at James Cook University and a former Professor of Economics at the University of the South Pacific where he worked for more than 40 years. The views expressed are the author’s and not necessarily of this newspaper.
