Malawi 19 May elections: An informed statistical prediction

Since the advent of multiparty democracy, we have noted the extent to which a candidate's home and to a lesser extent religion play a role in determining whether the candidate wins big or less. Of course the aforementioned factors do not predict the outcomes of elections with perfection but a study of 100 percent of the elections we have had so far confirms that these proxies are non-ignorable. The referendum race at the turn of the 1990s was probably the only exception when individuals voted for or against multiparty politics based on what they believed democracy had in store for them.  

However, an examination of data for the 1994 presidential and parliamentary elections does reveal that people had been quickly conditioned to vote based on seemingly less objective but unfortunately real and often important grounds. Most of these events are quite obvious and this posting will not waste your precious time narrating the palpable, rather effort is made to flag some connections between locations, altars and votes and the likelihood of such realities to drive this year's results. 

In 1994 people voted along regional lines generally and obviously given that we had three major parties (AFORD in the North, mighty MCP at the Centre and UDF in the south) and the party from the largest region, UDF, had to win. Regional affiliation had helped deliver the presidency to Dr Muluzi and multiparty politics set in together with politics of akwithu.

Indeed, AFORD had a majority of its MPs from the North; MCP dominated the centre while Muluzi dominated the south though we had a few exceptions to the rule in the central region where yellow flags flew in some areas. Whether the elections of 1999 were a real win for Muluzi or not (and I am not here to debate that), it seems that people thought of righting the wrongs of the 1994 elections such that the centre became increasingly Tambala wa kunda, the south continued to yellow while the North AFORDed even further. 

This posting spends more time on the more recent 2004 elections which showed the same pattern (notice that the fact that the North voted for Chakuwamba does not break the pattern because among all the strong parties that contested that year, it was only Gwanda's that had many Northern gurus and hence promissory to the North). The outcome of that election saw JZU coming victorious with the centre as the centre of his strength, Gwanda took the north and some parts of the south while the less known Bingu ridding on the shoulders of Dr Muluzi, took the Yao belt and scored well in Lomwe dominated southern districts. 

I believe that the past is very instrumental in predicting the future and that the past can be made more useful if data is analysed with rigor and insight. I have done this with considerable care based on the data files detailing the results of the 2004 elections circulating on Malawitalk, Malawiana and Freenyasanet forums. 

The table below shows snippets of the partial analysis of the data. It presents how the candidates fared. The results are presented in ascending order i.e. the candidates strongholds are presented first to aid interpretation. The figures shown are proportions of the vote going to the candidate in that district. 

DISTRICT	Chakuamba	DISTRICT	Mutharika	DISTRICT	Mpinganjira	DISTRICT	Malewezi	DISTRICT	J.Z.U. Tembo
Karonga	0.79	Machinga	0.88	Mulanje	0.57	Mwanza	0.36	Dowa	0.80
NkhataBay	0.77	Mangochi	0.84	Phalombe	0.37	Ntchisi	0.32	Dedza	0.75
Mzimba	0.75	Balaka	0.74	Likoma	0.20	Likoma	0.21	Lilongwe	0.73
Nsanje	0.73	Thyolo	0.58	Chiradzulu	0.18	NkhataBay	0.14	Mchinji	0.67
Rumphi	0.66	Chiradzulu	0.55	Thyolo	0.16	Ntcheu	0.03	Kasungu	0.63
Chitipa	0.60	Mulanje	0.51	Mwanza	0.16	Balaka	0.02	Salima	0.49
Chikwawa	0.58	Phalombe	0.50	Neno	0.15	Mangochi	0.02	Ntchisi	0.47
Blantyre	0.48	Ntcheu	0.49	Blantyre	0.11	Chiradzulu	0.02	Nkhotakot	0.36
Likoma	0.41	Nkhotakot	0.42	Balaka	0.11	Machinga	0.02	Ntcheu	0.26
Neno	0.39	Salima	0.38	Chitipa	0.11	Neno	0.02	Mzimba	0.05
Mwanza	0.36	Blantyre	0.38	Ntcheu	0.09	Mulanje	0.02	Neno	0.04
Thyolo	0.18	Neno	0.38	Zomba	0.09	Salima	0.02	Mangochi	0.03
Chiradzulu	0.18	Zomba	0.36	Machinga	0.08	Thyolo	0.02	Chitipa	0.03
Kasungu	0.14	Rumphi	0.27	Mangochi	0.06	Zomba	0.02	Balaka	0.03
Zomba	0.14	Nsanje	0.25	Kasungu	0.04	Phalombe	0.02	Blantyre	0.03
Nkhotakot	0.13	Mchinji	0.24	Nkhotakot	0.04	Nkhotakota	0.02	Chiradzulu	0.03
Lilongwe	0.11	Chitipa	0.24	NkhataBay	0.03	Chitipa	0.01	Mwanza	0.02
Balaka	0.11	Chikwawa	0.21	Salima	0.03	Blantyre	0.01	Machinga	0.02
Ntcheu	0.09	Dedza	0.20	Lilongwe	0.03	Dedza	0.01	Rumphi	0.02
Mangochi	0.06	Lilongwe	0.20	Mzimba	0.03	Lilongwe	0.01	Karonga	0.02
Machinga	0.06	Kasungu	0.17	Karonga	0.03	Kasungu	0.01	Mulanje	0.01
Mulanje	0.05	Likoma	0.17	Chikwawa	0.02	Mchinji	0.01	Zomba	0.01
Salima	0.04	Karonga	0.17	Mchinji	0.02	Dowa	0.01	Phalombe	0.01
Dowa	0.03	Mzimba	0.15	Rumphi	0.02	Mzimba	0.01	Thyolo	0.01
Mchinji	0.03	Dowa	0.15	Dedza	0.02	Karonga	0.01	Chikwawa	0.01
Ntchisi	0.02	Ntchisi	0.14	Dowa	0.01	Nsanje	0.01	Nsanje	0.01
Phalombe	0.02	Mwanza	0.07	Ntchisi	0.01	Chikwawa	0.01	Likoma	0.01
Dedza	0.02	NkhataBa	0.01	Nsanje	0.01	Rumphi	0.01	NkhataBay	0.01

 
 

Gwanda fared well in the North and in Sena areas of the South. BJ fared well in Lomwe areas of the South while Malewezi did well in Ntchisi his home town and the Nyau areas of the South. Bingu, riding on the back of Muluzi, did well in all Yao areas and in the Lomwe belt but was a mediocre in the North and centre.  

The election next month will not be any different because although some people could argue that Muluzi is not popular any more, his people of the Yao belt will follow him. Also notice that Muluzi and JZU fared very poorly in the North, but together they still had some small numbers worth counting. Those are unchangeable and reliable figures because anyone who voted for them in 2004 was likely someone who genuinely wanted them to rule. So a proxy of the Alliance vote from the North is found simply by {summing up Muluzi's, JZUs, Mpinganjiras and Malaeweze votes in the north}. We cannot use Gwandas because whoever voted for Gwanda did so because of the nature of his party at the time. His party was similar to today's DPP and hence most of them will be DPP. 

When we go to the south however, Gwanda's votes in his home town count but not in the other areas considering other issues which are not the purpose of this article. Similarly, all the votes from Yao land that went to Bingu in 2004 will be for the alliance but Bingu will score more in the Lomwe belt though his scores will be checked by his tribesman BJ. I predict that Bingu will get 52 % from the Lomwe belt while Mpinganjira will collect the rest for the alliance. I have given Blantyre to Bingu because I want the analysis to be very parsimonious and conservative so that my predictions are robust even if we allow for some rigging by the incumbent.  

In  case you did not know, the North accounts for about 16 % of the total vote; the Yaowland even including Zomba, plus Nsanje, and Mulanje (leaving the rest including the controversial Blantyre for DPP)  accounts for 22 % of the total vote; the rest of the southern region accounts for  circa 21 % of the total vote . The centre however accounts for 42 % of the total vote which should not be surprising. 

My calculations do show that though the aggregate regional voting power correlates well with actual voting patterns, it is not perfect. In fact if this was to be taken as a predictor, Bingu would lose slightly more than he actually loses under the more parsimonious scenario given below. A more precise predictor of voting patterns is found by making use of previous actual voting patterns to get proportions that can be used to simulate future voting. If this is done, the predictors fare well in forecasting previous election results. With this in mind, the predictors are used to predict the 19 May elections and the results show the following: 
 

•  
  •  
    •  

      	Jzu	Bingu
      North	0.040	0.114
      Centre	0.308	0.118
      Yao+MJ+Sena	0.168	0.040
      Lomweland	0.040	0.160
      Total	0.556	0.431

 

So if the assumptions espoused herein are realistic (and there is no reason why they shouldn't), unless the government rigs the elections, or something too Herculean  takes place between now and the elections day, JZU and the alliance will emerge victorious with at least 52% of the vote. Statistical methods show that this is true even at 1 percent level of significance, in other words the standard error associated with prediction is so small that it almost true. You can start betting, you have my word.  

This does not have anything to do with what I want, it is the reality. It will take a unifier to reverse this trend. To be a unifier one has to look beyond advice from greedy advisors whose advice often diverts from reality, is based on academic and not real facts, or is based on anger and the need to revenge or unfairly promote disgruntled minority standings etc.  Once a president forgets this tenet, he and his government are doomed and must Mugabelize or Kibakilize elections to the detriment of their legacies and to the nation. Prepare for a JZU presidency my friends and countrymen.   

*Chitsamba Undichangamire Mwachajesotafuna (Osaka, Japan) -- Computable Fluid Dynamics PhD student

http://www.nyasatimes.com/opinion/3325.html