Book Review: Strength in Numbers

In a new book, Strength in Numbers: How Polls Work and Why We Need Them, G. Elliott Morris sets out to provide an ‘insightful exploration of political polling and a bold defense of its crucial role in a modern democracy’. I am interested in political polling and, unsuprisingly, I was looking forward to reading this book. We need more literature on how polls work and why we need them.

The book in question is about how polls work and why we need them. But it is also a lot more than that. First and foremost, it is a cronological story about political polling in the United States, from the early days of capturing public opinion to political polling today. If you are not interested in the history of political polling in the United States, this book might not be for you.

In the first chapter, Democracy and the Public Will, the focus is not on political polling but on the nature of public opinion. This chapter deals with a series of thinkers and ideas, including Confucian philosophy, Greek philosophers, John Locke, Rousseau, Jeremy Bentham, James Madison, and Walter Lippmann. The main conclusion in this chapter is that the ‘better philosophy lies someplace between the extremes of past theories’, and this is symptomatic for most of the ideas and arguments presented throughout the book. That is, the book is – when taking a stand – mostly trying to find common ground between different perspectives. While it is interesting to introduce a series of thinkers and ideas, the main limitation is that there is a lot of name-dropping and if you are familiar with the core ideas of the names mentioned above, you can easily skim through this chapter or skip it altogether.

In the following chapters, we begin the historical introduction to political polling in the United States. In Chapter 2, Polling Comes of Age, we are introduced to the first straw polls, Literary Digest, George Gallup, ‘Dewey defeats Truman’, etc. Specifically, we are introduced to what can happen when sampling error is not random. In Chapter 3, Machine Politics, we read about what we can call scientific polling and the increased complexity in the collection of opinion polls. The focus here is on the early roots of data science (focusing on companies such as Simulmatics Corporation) and new concepts such as quota controls.

In Chapter 4, One Bad Apple Doesn’t Spoil the Bushel, we move beyond the United States and focus on extreme challenges with data collection and some of the worst problems we can deal with, namely data fabrication. Ironically, this is one of the best chapters in the book but also the one that is more of a detour in terms of the theme and narrative of the book. The chapter focuses on surveys of child mortality rates in Iraq, and while you will find a lot of great points in this chapter, it is not adding a lot to the history of political polling, nor ‘how polls work and why we need them’.

In the next two chapters, Chapter 5 and 6, we deal with the the most recent innovations in political polling, from poll of polls, probabilistic forecasts, and polling averages/aggregation to online surveys, big data and MRP. The focus here is on names such as Pollster, RealClearPolitics, FiveThirtyEight/Nate Silver and other actors that have played a crucial part in election polling in the most recent elections in the United States. In Chapter 7, Taking the Pulse of the Pulse of Democracy, we are introduced to some considerations on what went down in 2016 and 2020. While this is a fine chapter, there is a lot more recent literature I find more relevant on the topic. For example, I found these posts on Nate Silver and what went down in 2016 much more detailed and interesting.

If you are unfamiliar with the history of political polling in the United States and would like to read a brief introduction on the topic, this book is for you. If you are already familiar with the broad trends in political polling in the United States over time, you can consider skipping this book. This is a good book, but if you are already familiar with the topics and themes mentioned here, there is not much new under the sun.

Of course, additional points are made throughout the book, and a few of them are noteworthy here. First, the book emphasises that a core problem with opinion polls is not the polls, but what we expect of them. This is a point I have also made several of times (including in our book on how journalists cover opinion polls). This is an important point to make and I am happy to see this being discussed here as well.

Second, and related, the book serves as a corrective to the narrative on the failure of opinion polls. The go-to reference, that is also discussed here, is the great study by Will Jennings and Christopher Wlezien, showing that ‘the recent performance of polls has not been outside the ordinary’. In other words, we are not looking at a historical failure of political polling. In brief, opinion polls are not perfect and opinion polls are not bad. Opinion polls are good.

In the conclusion, the author proposes five reforms. First, that pollsters should abandon polls fielded entirely by phone. Second, we should acknowledge that the error is greater than what we expect from the margin of error and that individual polls should be treated with skepticism. Third, forecasters need to do better in terms of aggregating opinion polls to address biases (or at least be aware of the limitations). Fourth, the American Association for Public Opinion Research should do more to combat the influence of low-quality polls. Fifth, more political interest groups should advocate for the relevance of the public’s opinions. These proposals are great, sensible, and in some cases obvious.

My main (minor) disappointment with the book is what is not there. G. Elliott Morris was responsible for the 2020 election forecast model (for The Economist), but very little attention is given to the details of all this work. I would have liked to see much more insider knowledge on this work and more on the lessons and limitations from this work. Hopefully, this will be a topic for another book.

In sum, this is a good book. As opinion polls, the book is not perfect and not bad, but pretty good. It is definitely recommended for people not familiar with, but interested in, the history of political polling in the United States.