I don’t know where the phrase came from, but as a long-time data person, I’ve heard it roughly one million times: “data speaks for itself.”

Perhaps I’m assigning my current beliefs onto my past self, but I’ve never really understood the phrase. As someone who’s worked with real-world quantitative datasets continuously since I was an undergraduate, it seems like I’ve always had to do a lot of work to get the data to say anything, and that those answers are often very confusing. Numbers, I know, are convincing, but typically those numbers that convince people are themselves not the actual data, but analysis of the data. That is, analysis embedded with human decision-making, choice, and values.  

Class Conversations

Occasionally I have conversations about beliefs, values, and perceptions of data with my students, though some contexts lend themselves better to this conversation than others. This past semester I taught my upper level research methods course for Economics students, Causal Inference. This course focuses on the distinction between correlation and causation—why it’s hard to disentangle these claims, and how economists go about trying. I used the last day of class to review a chapter from their textbook, Thinking Clearly with Data (Bueno de Mesquita and Fowler 2021), that explored the ways in which values and assumptions become embedded within the research process without us even realizing it.  

This particular group of students, while great in small groups, often held their tongues during full group discussions. On this last day, the questions came flying. This idea of where bias makes its way in seemed to fascinate them and get them thinking. Their questions were insightful and complex, and just kept going.  

I tried to be honest in answering their questions that when we get into the realm where values and bias come into the data analysis and research process, there isn’t a checklist to go down—and there isn’t universal agreement amongst scholars and analysts, not even within individual disciplines. I gave my honest perspectives on their questions about conflicts of interest, measurement issues, and much more, but made clear that if they asked each of us in the Economics Department, we’d all probably provide somewhat different answers.  

For example, one student essentially wanted to know whether there could ever be a truly perfect dataset. My first reaction was “no,” but the question challenged me to get more specific. I pulled on language I’ve learned from being in community with folks across different disciplines, like my colleague Amanda Kleintop in History who gave me the language “what is this a reliable source for?” (Kleintop and Wigger 2025) and in critical spaces like with the Quant4What Collective, as well as from Data Equity Training with We All Count. After considering the question more carefully, I responded that it depends on the question we’re trying to answer. Some datasets, some variables, or some samples are better at answering some questions than other questions. Even if those questions are related, data does not just exist in a vacuum—we use it to tell us something about the thing we want to understand better.  

What Do Students Think About “Data Speaks for Itself”?

Shortly before we got into this discussion, I had them complete a survey as part of a cross-disciplinary study on how undergraduate students view, understand, and evaluate data. The survey included a series of statements and asked them to rate the extent to which they agreed or disagreed. Over the past year, four of us (myself and my CEL collaborator, Amanda Kleintop, along with Dr. Jessica Carew and Dr. Olivia Healy), representing three departments, distributed the survey to students across five different courses and received over 150 responses.  

Yes, one of those statements was “Data speaks for itself.” 

This particular statement was one I was perhaps most interested in. On average, students rated their agreement with this statement as 3.25 out of 5, but there was quite a bit of variability. Even if we simplify the levels of agreement, there was no majority: 42% rated it as a 4 or 5 (indicating some level of agreement), 21% as a 1 or 2 (indicating some level of disagreement) and 37% rated it as a 3.  

Given the ubiquity of this phrase, perhaps I expected more students to agree strongly with this statement more than they did. Only 11.5% of students rated this statement as a 5 out of 5, and even though the results suggest that many students lean towards agreement with this statement, there was a large chunk that stayed relatively neutral, and a small but meaningful chunk that outright disagreed.  

But what really struck me was how those Causal Inference students responded. This was a small group (fifteen responded to the survey), but they leaned towards disagreement with the statement, with an average rating of 2.5 out of 5. Forty percent of students rated the statement as a 1 or 2, 47% as a 3, and only 13% gave it a 4. No student rated this statement as a 5 out of 5 (strongly agree). These were the lowest ratings across all of the courses that we distributed the survey in, and also the course with the most advanced data analysis topics.  

I’m still figuring out what to make of this. I don’t know if this was a fluke, if this just happened to be a set of more skeptical students, or the extent to which my own opinions influenced theirs. But ultimately my guess is that the more these students learned about the difficulty of data analysis—the ways in which not just data but results required careful interpretation and assumptions—that perhaps this statement to them also began to make less sense.  

Implications for Teaching Data Literacy

All of this is getting me reflecting on how to integrate conversations about values, beliefs, and philosophies about data into the spectrum of data literacy education, not just at the capstone of their college careers. My own values and beliefs about data have changed as I’ve learned more over my formal and informal education, and I’ve benefitted enormously by having colleagues and friends both within and across my discipline and industry to discuss these ideas with. There is not a right or wrong answer, strictly speaking, about whether or not “data speaks for itself,” so this isn’t something we can just teach once and be done. Perhaps there is a benefit to giving students multiple chances over their college career to reflect and ask questions about the ways we think about data. 

References

Bueno de Mesquita, Ethan, and Anthony Fowler. 2021. Thinking Clearly with Data: A Guide to Quantitative Reasoning and Analysis. Princeton University Press.

Kleintop, Amanda and Cora Wigger. 2025. “Data Literacy in Engaged Learning: Understanding Bias.” Center for Engaged Learning (blog), Elon University. July 22, 2025. https://www.centerforengagedlearning.org/data-literacy-in-engaged-learning-understanding-bias.

Quant4What Collective. n.d. Quant4What. Accessed August 5, 2026. https://www.quant4whatcollective.org/

We All Count. n.d. Foundations of Data Equity. Accessed August 5, 2026. https://weallcount.com/courses/foundations-of-data-equity/

About the Author

Cora Wigger is an assistant professor of economics and a 2025–2027 CEL Scholar. Her research focuses on the intersections of education and housing policy, with an emphasis on racial inequality and desegregation. At Elon, she teaches statistics and data-driven courses and contributes to equity-centered initiatives like the “Quant4What? Collective” and the Data Nexus Faculty Advisory Committee. 

How to Cite This Post 

Wigger, Cora. 2026. “What Does Data Say?” Center for Engaged Learning (blog). Elon University. August 11, 2026. https://www.centerforengagedlearning.org/what-does-data-say.