Translation of excerpts from Robert Heinlein’s book, Take Your Government Back - part 24

Analysis of the pre-election situation in the district

And now Mr. Chestnyagi’s campaign to nominate a party candidate has reached the stage of the last election month. You have done a good job, and now you would like to know what results your work has brought. Do you need to work hard now, or is the game already lost, and is it time for you to turn down the shop?

Most likely, you cannot afford to hire a professional analytical agency, and your volunteers do not have time to do analytics: they spend all their free time on the campaign. And, although at the weekly meetings of the Voter Surveyors Club, each of them shares the results of their own polling at their polling station in their reports, you would like to be confident in the accuracy of the figures reported by volunteers, because their reviews often represent the situation in an overly rosy light.

There is a technique for obtaining a reliable picture of voters' opinions, which requires experience, but which you can start practicing as soon as you come into politics, so by the time you need to practice it, you will already be experienced enough. This technique consists in trying to predict the results of voting on the candidates, and the questions put to the vote - first before, and then after a selective survey of the opinions of voters, conducted by you personally. Record all your predictions, then compare them with the election results.

From experience, you will understand whether your forecasts are too optimistic, or, on the contrary, they are too pessimistic. Over time, your analytical skills will improve, and the accuracy of forecasts will improve to such an extent that the time will come when you can come to the district and literally feel the victory or defeat in the air in the elections - many weeks before the elections. And when you learn this, you can turn defeats into victories. Review your predictions regularly - from submitting an application for elections to the night before the election. Record your predictions to use them after the election in the “debriefing”. This lesson attracts more than solving crossword puzzles, and some prefer it to reading detectives and playing on the tote.
')

Forecasts by statistical methods

Even if you can afford to hire professional pollsters and statisticians to study the pre-election situation, it is better to spend this money on campaigning. Does this mean that you have to move blindly, working like a damn when things have been lost for a long time, or lose because of a small lack of votes, although a small final spurt would lead you to victory, but you didn’t guess about it? No - it’s not worth working without imagining the situation in the district. There is a fairly simple and inexpensive way to conduct a poll in a district of any size, even the largest of them. And this method will give statistics, fairly reliable, in order to make decisions on how well your campaign is progressing, and adjust your plans accordingly.

The secret of reliable forecasts based on a study of a random sample of voter opinions lies, firstly, in the correctness of the survey method, and secondly, in the right attitude to the survey results, without trying to derive more information from them than they contain.
Mathematical theory of probability is complex and difficult to understand. Therefore, instead of trying to teach you, I will give you a simple rule with instructions on how to use it. And I will talk a little about what mathematical methods are incorporated in it. But in order to use my rule, one need only remember the rule itself. The explanation of the principles laid down in it is for those of you who understand mathematics, and will want to verify for themselves how this rule is derived. After that, it is possible that he will enter into polemics with me, both about the theory I have used, and I propose another rule that is more suitable for our purposes.

This is the rule: polling randomly selected voters in the district, until you get 50 clear and unambiguous answers - it does not matter in your favor or not. Do not take into account those who refuse to answer or find it difficult to answer. Now take the number of answers in favor of your candidate and double it. Subtract 8 from the result. What happened is the percentage of votes in your favor. With a probability of four chances to one, your candidate will not receive a percentage of votes lower than that calculated by the rule - provided that the elections are held right away. And you can almost be sure that the percentage will not be much lower than the figure you received. So, rely on this estimate as if it were reliable: in fact, it gives a cautious, even a little underestimate. After all, it is better to err on the smaller side than on the big one, losing the election because of this.

For example: you interviewed 93 randomly selected voters, until you received 50 answers: 14 respondents refused to answer, 29 - found it difficult to answer. Of the 50 responses received, 28 turned out to be for Mr. Chestnyagu, 22 - for Mr. Hope. We double 28 - we get 56, we subtract 8 - we get 48. Thus, if elections are held in the near future, Mr. Chestnyag can expect to receive at least 48% of the votes. We can get the upper estimate of the percentage of votes cast if we add 8 to 56, so the optimistic estimate will be 64% of the votes. But you don’t want to use this figure, because it’s more useful to know how many votes you need to catch up in order to guarantee the winning elections, so your score is 48% of the votes. This is obviously not enough for your candidate: only 2% of the votes can separate him from the election victory. For an expected total of 25,000 voters, this will be 500 votes. So you need to collect at least 500 votes, in addition to what you have already collected. But it is better to plan with a guarantee, and set your task to collect 3 times more - 1500 votes. To do this, you collect an urgent meeting of agitators, show them the 48% you received, and explain that for a guaranteed victory, each of them, in addition to the votes they have already collected, must collect another 6 new votes, visiting those voters who did not vote on the primaries, and therefore not listed in your lists. Let the agitators work until they find 6 voters whom they can bring to vote for the primaries. Let your candidate do the same thing, now 60 hours a week, and you yourself, spending four half-days a week to bypass the voters, instead of the previous two, even if after that you have to finish your work in the office until midnight.

Such a final breakthrough should go on for three weeks in a row, and it will lead you to victory, which you might have missed because of such a meager margin of votes. You will probably win with a considerable margin, and perhaps the final breakthrough was not needed - you will not know for sure. But you don’t need it, because you won!

Suppose polls showed that a pessimistic estimate shows a more than 50% advantage in your favor. Then you have the right to continue your campaign as usual, without a final breakthrough, but without slowing down the pace of work.

Now, suppose that the polls showed a reverse picture - 22 answers for Chestnyagu, 28 - for Hope. Then your score gives only 36% of the vote for you. Does this mean you have to give up and give up? No, because at the same time the percentage of votes in favor of your opponent is less than 50%, which means some opportunity for you to win - if you work hard, although not for sure. Accept the probability of defeat, and continue to fight.

Suppose Mr. Nadezhda received 30 votes in the poll, that is, the preponderance of votes in his favor is at least 52%, and at most 68%. In such a situation, would you rather surrender and quit the game?

At least not on their own initiative. I would advise you to discuss this issue with the candidate, then convene for a closed meeting all your employees and sponsors of the campaign fund, tell them this sad news, and get their opinion on how to act in such a situation. Based on a cold-pragmatic point of view, it is best for you to minimize your losses by leaving the race, but I foresee that your colleagues will vote to fight to the very end, and your meeting will turn into a pre-election rally. You can even manage to win: politics is not a game of dice, and not the theory of probability, a fighting spirit here can work wonders, knocking over all scientific calculations. So, if the meeting decides to campaign until the end, submit to their opinion, and get busy. At the very least, you will gain invaluable experience; many of the lost campaigns were the germs of future stunning political victories.

Now let's talk about what we mean by “randomly selected” voters: this is a sample of voters representing the current situation in the district as objectively as possible, and it is easiest to get it, eliminating all the factors of personal predilections. This is done in the following way: suppose you need to select 100 people, out of 200 voter lists you have. Take from each second list the name of the first voter of your party from the end of the list, from the second column of the list. Or invent your own rule, without making a decision on your part, to evenly and mechanically select people from the existing lists, without taking into account the distribution of voters over the territory.

Never take voters from just one polling station, or any part of the district. If you are going to conduct a survey by phone, keep in mind that some voters will not have a phone. Do not replace a selected voter who does not have a telephone next to him in the voter list with a telephone, interview the selected voters by visiting their home. Otherwise, you will distort the results of the poll, eliminating voters by the level of their economic well-being.

It is better to conduct a poll of voters by telephone in the evening, so that the whole family is at home. Do not rely on the answer of the spouse of the voter, instead of him, it will also distort your results. Usually women vote more conservatively and differently than men.

When conducting a survey, do not formulate the questions so that they suggest the desired answer. Here is an example of suitable wording: “Good evening, is this Mrs. Mable Smith? Mrs. Smith, you're worried about the civic awareness bureau. We are conducting a survey. Tell me, please, did you decide for which candidate you would vote in the primary elections of congressmen next Tuesday? ”

Practice shows that one person can prepare a list of polled voters in one evening, and in no more than three evenings - get fifty answers. Preparing a survey using postcards with a paid answer will take about the same amount of time, and such a survey is no worse than a telephone one, but getting its results takes more time, and you need to send at least 250 postcards. However, this method may be cheaper than a telephone survey, especially if you have to make a long-distance call. (Postcards with a paid answer, 2 cents each, is generally an irreplaceable thing in a low-budget policy).

Do not combine polls with voters bypass, and do not interrogate everyone you meet on the street: the list of respondents must be compiled by a mechanical method, eliminating any personal influence.

Now let's talk about the mathematical basis of the Rule of Eight (if you are not interested in mathematics, feel free to skip it). To begin with, when analyzing a random sample, the larger the sample size, the smaller the statistical errors, unless these are systemic errors related to how the sample was taken. In my opinion, in any political survey conducted without the help of an expert, system errors are so great that it makes no sense to make a sample of more than 100 voters. On the other hand, with a sample of less than 50 voters, the statistical errors determined by the laws of probability theory overlap all other patterns. Therefore, for a reasonable waste of time and money, we need to get a sample of the minimum volume, in which we can trace the laws we are looking for. Thus, to assess the predicted ratio of votes, it is enough for us to get 50 responses. In addition, out of 50 answers, you can immediately get the predicted percentage of votes by simply doubling them (some of us do not know how to calculate percentages).

To obtain the “Rule of Eight”, the Bessel formula was used to estimate the probable error, subject to a statistically independent sample of events. The statistical independence of the sample can be challenged, but the errors introduced by its irregularity can be neglected: we are only interested in the current political situation, and the preponderance of it in one direction or another will be clearly seen against the background of other errors. Just keep in mind that the “Rule of Eight” is not a probable measurement error applied by engineers, and not the standard deviation used by statisticians. The first is too difficult for an ordinary agitator, and the second requires a much larger sample size than volunteer policies can afford. The error rate of 8% is chosen to obtain a probability of 4 versus 1 chance that the outcome of the elections will be higher than our pessimistic estimate, which is excellent for our purposes, and more reliable ways of making important decisions such as choosing a wife, for example.

If you can afford to do a wider sample, conduct a survey among 100 voters, and do the calculations described above, substituting the number 5 instead of 8. A reader who understands mathematics can conduct his own statistical analysis using the Bessel, Peters, or other formulas.

Forecast "Feels Like"

In addition to surveys and the development of forecasting skills, try another way, and over time you will learn how to skillfully use it. To do this, wander around the polling station, go to the store to buy Coca-Cola, and talk with the seller. Stopping at a gas station - chat with the staff. Ask a passerby to smoke, and gossip a little with him. Cut your hair and talk to a hairdresser. Ask the counter road and talk to him. When you do this, your subconscious will relate the impressions you just received to the impressions of the voter visits (which you, as an election campaign manager, are scattered around the district), and you will have a curious, difficult to explain expression deep inside. Pull that feeling out of the mind, examine it carefully, and decide whether it tells you that your candidate will win the election, or whether it believes the candidate will lose them.

A trained human mind is able to produce solutions to complex problems with many unknowns, faster than any mechanism invented by mankind does. Over time, you will develop your intuition, and will be able to predict not only the election results, but also the ratio of votes and the number of voters. Mastering this talent is completely painless, almost without effort on your part. In the meantime, you are just learning, do not panic, even if it seems to you that you are losing with a crash.

Beginning politicians are inclined to believe that their energetic efforts will instantly create a whole tidal wave of electoral votes, and they are very disappointed when they come out to survey the polling station, revealing only light ripples instead of waves. This is normal: the primaries barely awakens the public from civic sleep. You need just such ripples - the right size and in the right place. And you know for sure that this polling station is the right place, once you have found it necessary to visit the voters living here. All you have to do is find out what size of ripples you need. You interview only adults, 200,000 people live in the district, Mr. Chestnyage needs 15,000 votes. If one of the four passers-by you have heard something about your candidate, it means that your candidate is at hand to win. If it’s already the last month of the campaign, and only one out of ten people you’ve met knows about your candidate, then you need to roll up your sleeves and get to work. And do not forget to ensure that on election day your election organization ensures that all guaranteed votes are received and all holders of probable votes visit the polling stations, otherwise you will suffer a crushing defeat! But even in such a situation, you can still squeeze out the winning number of votes - hard work in the last week before the elections. Although, it will not be easy - and no matter how bright the situation will be shown by telephone polls of voters.

→ Part 1, where there are links to all other parts