From Source Credibility and Measurement Validity to the Effects of Information and the Limits of Political Interpretation
By Dr. Nabil Kukali
The postponement of the Palestinian elections was disappointing to many who had hoped that the opportunity was finally approaching to renew Palestinian institutions through free and credible elections. September 11, 2027, has now been set as the new date for the presidential and legislative elections. Yet elections are more than a date. They are closely connected to rebuilding public trust, accountability, and political legitimacy, and they require conditions and guarantees that allow Palestinians to exercise their democratic rights meaningfully.
Before the nomination process was completed and before the postponement was announced, public discussion about electoral lists and possible political choices had already begun, and polls may have been conducted to measure voter attitudes or test possible scenarios. This makes the current moment useful for a broader public-education discussion about how election polls should be read, rather than for judging any particular poll, party, list, or candidate. The more important question is not simply: Who came first? It is: How was the number produced, what does it actually measure, and what do the data allow us to conclude?
In elections, a percentage can be mathematically correct while the interpretation built around it goes beyond what the evidence can support. The number seen by the public is only the final stage of a longer process that begins with defining the target population, selecting the sample, wording the questions, determining the alternatives, choosing the interview method and timing, and then processing and interpreting the data. Poll quality therefore cannot be reduced to sample size or margin of error alone.
Before the Percentage: Is There a Real Study Behind the Number?
In today’s information environment, poll literacy may need to begin one step earlier. Before asking about the margin of error, we should ask: Is there an actual study behind the numbers? Who conducted it? Who funded it? Who were the respondents? How were they reached? And how were the data collected?
Transparency does not automatically make a study good, but it makes its quality possible to assess. Nor should a poll’s credibility depend on whether we like its findings or whether they agree with our political expectations. A credible poll is not one that tells us what we want to hear. It is one that shows us what people actually report, how the result was produced, and where the limits of interpretation lie.
The Question Is Part of the Meaning of the Number
Political affiliation is not the same as voting intention. Voting intention is not the same as willingness to consider a new option. Trust is not the same as competence. And none of these is identical to actual behavior on Election Day. A citizen may answer these questions differently without contradiction because each one measures something different.
This is where measurement validity becomes essential: Does the question actually measure what we claim it measures? Electoral choice may be influenced by a party, a political figure, a program, previous experience, economic conditions, or trust. If a poll presents the name of a person associated with a list, are we measuring support for the person or for the list? If it presents a program without names, we are measuring something different. The question is therefore not merely a wrapper around the number; it is part of its meaning.
We should also avoid turning several different movements in the data into one political story. One option may decline, another may remain stable, and a third may rise. Sometimes the closeness or fragmentation of support matters more than who happens to rank first by a small margin.
Timing Changes the Meaning of the Result
Fieldwork dates are not merely administrative details. A poll conducted while nominations are still open does not measure the same electoral environment as one conducted after candidate lists have been finalized.
Researchers may legitimately test possible political scenarios before nominations are complete, provided the findings are presented as scenarios. But if the purpose is to measure voting intention among actual electoral lists, early entrants should not be treated as though they represent the full set of options while other lists may still join. The researcher should either wait until the choices stabilize or make clear that the findings reflect the options available at the time, not a final simulation of the ballot.
When the Researcher Adds Information
Some studies provide respondents with information about a candidate or program and then measure their attitudes again. This is legitimate research, but it changes what is being measured. The first response reflects prior knowledge; the second reflects opinion after new information has been introduced.
If support changes, we can say that attitudes changed after the information was presented. We cannot automatically say which specific element caused the change unless the design can isolate that effect. Unequal information treatments do not, by themselves, prove bias. They may fit a particular research question. But the results should not then be treated as though all options were measured under identical conditions.
The Full Sample Is Not the Basis of Every Result
A poll may have a large national sample, but not every published figure necessarily rests on that full sample. Voting intention may be calculated among likely participants, then among those who have made up their minds, while youth or regional findings may be based on even smaller groups.
This is why the actual unweighted number of cases matters. Statistical weighting can correct some imbalances in sample composition, but it cannot create new respondents. Nor should the margin of error for the full sample be applied automatically to a much smaller subgroup.
The same caution applies to regional findings and youth results. A difference between two areas demonstrates a difference, but not its cause. A finding among young people may be an important signal, but it is not enough by itself to characterize an entire generation.
If different percentages appear for the same option, this does not automatically mean there is an error. One figure may refer to the full sample, another to likely voters, and another to those who have made up their minds. When several explanations are possible, asking for clarification is more scientific than jumping to accusation.
Change Does Not Prove Its Cause
A poll can identify a shift in public opinion after a political, economic, or security event, but it cannot always prove that the event caused the shift. Several factors may be operating at the same time, and some relevant variables may not have been measured at all.
We also record what respondents tell us; we do not read their minds. If we ask why they made a choice, we can analyze the answer. If we did not ask, we should not invent a motive and present it as fact.
Opinions Are Not Decisions
Even the best election poll remains a snapshot of a particular moment. Opinion is not a final decision, and voting intention is not the same as actual voting. A citizen may honestly say whom they intend to support and later change that choice because of new information, political developments, or a change in whether they decide to participate.
Poll estimates should therefore not be expected to match election results perfectly, especially when the race is close. A relatively small shift in turnout or among undecided voters can be enough to change the final outcome.
The quality of an estimate also depends not only on the research organization’s methodology and integrity, but on how citizens respond. Some voters are less strongly attached to political parties and may change their preferences more easily, while those with stronger partisan identification may be more stable. Response rates may also differ across political groups, making nonresponse an important part of evaluating poll quality.
For these reasons, the gap between what voters say in a poll and what they actually do on election day remains one of the hardest challenges in election forecasting. Forecasts should therefore be treated as probabilistic estimates, not certainties.
Even the best election poll is a picture of a particular moment. Opinion is not a final decision, and voting intention is not actual voting. A citizen may honestly tell a researcher today whom they intend to support and later change their choice because of new information, changing circumstances, or a different decision about whether to participate.
This gap between what voters say and what they later do is one of the hardest challenges in election forecasting. Forecasting combines scientific tools, including sampling, weighting, and estimating uncertainty, with analytical judgment about turnout, preference stability, undecided voters, and what may change before Election Day. Forecasts should therefore be treated as probabilistic estimates, not certainties.
One Poll Is Not the Full Picture
The problem is not only that a poll may be inaccurate. It is also how quickly one poll may be treated as though it represents the complete picture. A well-designed poll may still reflect a temporary event, uncertainty, or a political environment that has not yet stabilized.
Repeated trends over time are therefore stronger than a single snapshot, provided the measurements are genuinely comparable. If the question, data-collection method, or target population changes, part of the difference may come from the measurement itself. One poll cannot, on its own, establish a “historic” or “structural” transformation. The strength of the conclusion should match the strength of the evidence.
Why Do Opinion Polls Remain Important During Elections?
These methodological cautions do not reduce the importance of opinion polls. Rather, they help us use them more effectively. When polls are designed, conducted, and interpreted professionally, they help us understand citizens’ priorities, the stability of their choices, and areas of hesitation or possible change.
This is especially visible in Israel, where election polls are published frequently, allowing observers to track changes in voter attitudes, shifts in support between parties, and the effects of political alliances and developments, instead of relying on a single poll as if it were the full picture.
Frequent polling does not mean that every change represents a real political shift. But repeated measurement and comparability over time can help distinguish short-term movement from a more persistent trend. Polls do not replace elections, but they provide structured knowledge about what people are saying and help us understand the direction of public opinion more clearly
Transparency Begins With Us
Transparency does not mean that research is perfect. It means that others can evaluate it. If we expect others to disclose question wording, subgroup sizes, and limitations, we should apply the same standards to our own work.
A simple test of neutrality is to ask: If the result were completely reversed, would we apply the same methodological standards?
Errors are possible in any human endeavor. Correcting a number, table, or interpretation does not weaken credibility. Acknowledging and correcting an error is part of quality assurance.
The Number Is the Beginning of Responsibility
Public opinion polls remain among the most important tools for understanding societies during elections. They help us understand what matters to citizens, how stable their choices are, where uncertainty exists, and how attitudes may be changing.
But their importance does not remove their limits. Polls do not replace the ballot box. Voting intention is not actual voting. One poll does not represent society forever. Change does not prove cause. A subgroup is not the full sample, and weighting does not create new respondents.
Behind every number is a person. The researcher’s responsibility begins with selecting respondents properly, asking fair questions, protecting privacy and the right to express an opinion, and reporting answers faithfully without making the data say more than they can support. When the evidence cannot resolve an interpretation, “we do not know yet” may be more professional than an attractive answer that the data cannot prove.
The goal is neither to trust polls uncritically nor to reject them because we dislike their findings. It is to understand how to distinguish a good poll, how to read it in context, and how to use it without turning it into a substitute for the voter’s actual decision.
A good election poll does not merely tell us who may win. More importantly, it helps us understand the society that will go to the ballot box. The number should therefore be the beginning of analysis, not its end.
Dr. Nabil Kukali is a Palestinian public opinion research expert and the founder and president of the Palestinian Center for Public Opinion (PCPO). He has more than three decades of experience in survey design, analyzing public attitudes, and conducting social, political, and economic research in Palestine and the wider region. PCPO will be engaging in Diaspora Circles and connected activities with Generation1.ca.
