Wednesday, April 20, 2011

Very Different Survey Results: Telephone vs Personal In-Home

This post also available in Macedonian:
македонски @ Блогерај: Анкети со мноогу различни резултати: телефонски наспроти теренски
македонски @ Блогспот: Анкети со мноогу различни резултати: телефонски наспроти теренски


A current discussion on Twiter is based on a question by @banekoma:
#howcome [mk] 2 surveys with completely different results :)
1. Survey: VMRO-DPMNE persuasivelly ahead of SDSM [mk, 28% : 12%, Kurir]
2. Survey: VMRO-DPMNE 22% : SDSM 20% [Press 24]
#onlyinmk
[mk]
Assuming that both surveys were honest, they still can yield different results because the first was a telephone survey, and the second, as far as I know, was a personal in home survey.

Telephone surveys are conducted via fixed telephone network(s). In Macedonia, maintaining a fixed phone line is a luxury for some social strata. After some relatively inexpensive prepaid mobile options became available, poorer citizens renounced their fixed phone numbers on a massive scale. According to the Enlargement Countries Monitoring Report by Cullen International (Dec 2010), published on the website of the Agency of Electronic Communications, fixed telephony in Macedonia has a negative "growth" rate. In January 2005 there were 29.21 fixed phone lines per 100 inhabitants, which dropped to 20.56 in January 2010 - almost by a third.

Allegedly, this trend is especially present in more rural areas. Therefore, a survey based on calling people via fixed telephone would not accurately reflect the general opinions, unless it contains additional mechanisms to provide a representative sample of the whole population. Just doing a telephone survey with the same number of people as an in-house is simply not enough, because it would omit representatives of people who (no longer) own a fixed telephone. If the survey designers did not take this into consideration, their results will deviate from reality. During the 2009 Presidential Elections there were instances when some politicians who were oriented toward the rural population "surprised" with a better vote count than the surveys predicted

Also, there's a difference in the level of ease of the surveyed people depending on the method. Probably there's a precise psychological formula on how to relax the subject, but I think that when someone gets a telephone call he's more aware that someone could relate his personal data (telephone number leading to name and address) with their political orientation and would respond more carefully. Due to politicized nature of our society and the real influences on personal and family well-being, probably a portion of the future voters would decline to respond, or would say they are undecided, or would say they'd vote for an option they think would look best if checked by someone who they fear.

Unfortunatelly the website of the Institute for Sociological, Political and Juridical Research presents no data on this survey with explanations of the methodology. It seems that the quoted Institute Dimitrija Chupovski has no website at all. There's a Facebook page [mk] with that name which has a free Gmail as e-mail contact address, which does not contribute to building trust. Anyone can open such a page in literally 4 minutes, which does not help its credibility.

During the same discussion, pointed [mk] an example of a famous U.S.A. survey which failed due to non-representative sample, even though it covered an enormous percent of the total population.