Search :
         
 
Election Analysis
 
   
 
Proposal Summary: Poll-Level Electoral Return Map

Owen Lippert, Resident Director, NDI
A Poll-Level electoral return map of Bangladesh is needed in order to more effectively deploy election observers, identify instances of potential voter fraud and use polling data to forecast election results. The latest use would be in any redistricting of seats.

A poll-level electoral return map of Bangladesh is needed in order to more effectively deploy election observers, identify instances of potential voter fraud and use polling data to forecast election results. Another use would be in any redistricting of seats.

Background:

In developed Westminster Parliamentary democracies, political parties base their electoral strategy on concentrating their resources on swing voters in swing constituencies. They try to avoid expending resources reaching voters in constituencies where they have more than enough support to win or less than enough to be competitive.

A standard political strategy builds on three research activities:

• Identifying swing voters in swing constituencies through electoral analysis,
• Modeling their social characteristics through census data analysis, and
• Discerning top-of-mind vote-determinant issues through polling and focus groups.

 
 


The most reliable way to identify swing voters in swing constituencies is through an analysis of election returns at the smallest possible aggregation which is usually the individual polls. The goal lies in finding the polls where a small shift in the vote will -- on the margin – tip the constituency-wide result one way or the other.

The traditional next step in the process of building an electoral strategy consists of identifying the social characteristics of the individuals in the swing polls. Generally the residents of individual poll share economic, social or ethnic traits. The premise is that individuals with shared traits tend to have similar voting intentions. Census data typically provides this information.

The third stage consists of finding the issues that have the potential to shift the swing voters’ voting intention. Modern public opinion polling, using sophisticated statistical techniques in respondent selection, has proven highly effective in identifying “hot button” issues. Focus groups are then used in order to test out communications strategies that can drive issue concern into voting preference.

The end result is that political parties can forecast how a shift in swing voters’ voting preference will translate into won or lost electoral seats.

It also provides the opportunity for parties to shift their strategies and communications should their present efforts appear to fail to move the voting intentions of crucial swing voters. Too often electoral strategies consolidate the core vote at the expense of broadening the coalition.

Situation in Bangladesh

The country has 300 constituencies with approximately 30,000 polling centers.

101 constituencies in Bangladesh would qualify as swing seats as defined by a margin of victory of 20 percent or less. That is a ten percent increase in the vote of the next closest party would lead to the loss of the seat for the party currently holding it. The overall vote in the 2001 election shifted five percent in favor of the BNP and against the AL despite the latter actually increasing its total vote. The critical factor was the alliance of the BNP with the Jamaat-e-Islami party and the decline in the Jatiya Party vote.

With only three national elections, Bangladesh has a limited, but still potentially useful, data set of electoral returns.

Initial inquires with the former Chief Election Commissioner, Mr. M. A. Syed, did not prove fruitful as to the electronic record of poll level election data. One Commission researcher said that the data was held by the District Electoral Commissioners. This has since been confirmed. There are 64 districts (Zillas) in Bangladesh.

This is not to suggest that the various parties do not know the swing polls and swing voters, but that knowledge remains anecdotal with local leaders and not available for systematic analysis. In some cases, individual candidates do have the information.

Initial inquiries with the political parties suggest they are interested in using organized records of poll data. An Awami League representative said that they had started collecting this data once but that the computers had been seized in a police raid. Moreover, the decision of the coalition partners, BNP and Jamaat-e-Islami, as to where to run candidates suggests some knowledge of the poll data.

An inquiry to Professor Ataur Rahman, President of the Bangladesh Political Science Association, did not reveal any academics working with this data, though he said he would do a broader canvas of the members.

There remains, however, room for optimism. Further research, particularly with the local media, should uncover at least enough poll-level electoral data to provide a valid statistical sample if not a comprehensive record.

What is needed is a solid canvass of the District Electoral Commissioners.

Benefits of a poll-level electoral map

1. The most immediate benefit to democracy and governance agencies – and the parties themselves -- lies in a more systematic method for choosing where to send election observers. It makes sense to send foreign and domestic election observers to the polls that will determine the outcome of the election.

2. A knowledge of the past results of key polls provides a bell weather indicator of any possible electoral irregularities. If a result from a swing poll shows a dramatic shift in turnout or in voting preference that presents a prima facie case for investigating possible election manipulation.

3. If there is to be a redistribution of seats in Bangladesh, the poll-level data is essential to ensuring that the re-drawn constituency boundaries provide for fair and competitive political landscape.

4. Public opinion polling data does not provide an adequate guide to the electoral outcome of an election in the absence of poll-level data and even rudimentary analysis of census data.

The above statement holds particularly true in Bangladesh. A cursory look at the election returns of 1996 and 2001 show that a relatively small increase in the BNP’s vote (8 percent, perhaps half of which was the transfer of Jamaat support) led to a 44 percent increase in seats elected. The AL, despite increasing its vote (3 percent), witnessed a 57 percent decline in seats elected. A small shift in voters produced a disproportionate outcome.

This is, of course, a vagary of the first-past-post voting system, but one which makes understanding the swing polls and the swing voter all that more important for purposes of forecasting electoral results.

At present, some modern public opinion polling is being conducted in Bangladesh that deals with political issues. However, its utility is hampered by the lack of the underlying electoral and census data. One might say the polling horse is front of electoral and census data cart.

The greater the accuracy in polling and electoral analysis then the more likely that incidences of irregularities may be detected and the more likely that the parties will focus on the constituents and their issues upon whom electoral victory hinge.