About

About the Paper I Analyzed

Have you ever picked up your phone to check something quickly and somehow ended up scrolling for an hour? You’re not alone — and a group of economists decided to study exactly that. In 2020, researchers Hunt Allcott, Matthew Gentzkow, and Lena Song recruited about 2,000 American adults with Android smartphones and asked them to install an app called Phone Dashboard, which silently and objectively recorded how much time they spent on their phones each day. The study focused on six specific apps that participants said felt the most addictive and hardest to put down: Facebook, Instagram, Twitter, Snapchat, web Browsers, and YouTube — nicknamed “FITSBY” (one letter from each app name). Participants filled out surveys at four points over roughly 12 weeks, answering questions about their phone habits, how they felt about their usage, and how much they thought they’d use their phones in the weeks ahead. The results, published in one of the world’s top economics journals (the American Economic Review) in 2022, paint a striking picture of what the researchers call “digital addiction.” That’s what these visualizations explore.

You can find the full paper here: https://www.aeaweb.org/articles?id=10.1257/aer.20210867

About my Analysis

What I did in this analysis: The other day at a friends dinner I noticed everyone was in their phone at some point or another. I decided to take this one paper and visualize the data in it to explore the idea of how visualizations of data can help us understand psychology/psychiatry topics better. This is a purely experimental analysis based on a public data set and as such should be for educational purposes only. I’m looking to craft my skills in presenting this data so more public health infromation can be readily accessed by a larger percentage of the population. I have not verified if all the numbers are 100% accurate as of yet. - 3/24/2026

More context on my Analysis

There is a a bit more context you need to know so you can understand my visualizations. In the study there were The three groups: Participants were randomly sorted into one of three conditions. The Control group received nothing — no tools, no incentives, just the Phone Dashboard app recording their usage in the background. The Limit group was given a feature inside the app that let them set a hard daily time cap on any of the six FITSBY apps — for example, they could decide “I’m only allowing myself 30 minutes of Facebook per day.” Once they hit their limit, the app would lock them out for the rest of the day. Crucially, these limits were hard to override immediately, which is what made them effective as a commitment device. The Bonus group was offered a financial incentive: they could earn up to $50 extra if they reduced their FITSBY usage during a specific three-week window (weeks 7–9 of the study), paid out at a rate of $2.50 per hour of reduction below their personal baseline. Some participants were in both the limit and bonus groups simultaneously.

The big questions the study caused me to ask was: Do people use their phones more than they actually want to? Do they know it? And can anything — a hard app limit, or a cash reward — actually help them cut back? The results, published in one of the world’s top economics journals (the American Economic Review) in 2022, paint a striking picture of what the researchers call “digital addiction.” That’s what these visualizations explore.

Setup - The Packages Used

In order to visualize the data in R, I had to download a few packages to be able to work with the data. This is needed for data wrangling and plotting, set up a shared visual style, and load the actual study dataset. I also label each participant by which group they were randomly assigned to: the control group (no intervention), the limit group (could set screen time caps), the bonus group (paid to reduce usage), or both.

The Packages I used include: “ggplot2”, “haven”, “dplyr”, “tidyr”, “scales”

To confirm the data I’m using was loading properly I also ran a quick check:

## Dataset loaded: 1933 participants, 882 variables
## Group sizes:
## 
##    Control Limit only Bonus only       Both 
##        589        865        194        285

What this tells us: The dataset has 1,933 participants and 879 variables. Most participants were in the limit group (865 people), followed by the control group (589), with smaller groups for bonus-only (194) and both interventions (285).

Static Data Visualizations

Figure 1 — How Long Do People Spend on Each App?

What I’m looking at: I calculated the average number of minutes per day each participant spent on each of the six FITSBY apps during the baseline period — before any intervention began. This gives us a simple picture of where people’s phone time actually goes.

What the results show: Facebook is by far the biggest time sink at 69 minutes per day — more than an hour, every single day. Web browsing comes in second at 44 minutes. Instagram (24 min), YouTube (23 min), Snapchat and Twitter (15 min each) trail behind. Combined, these six apps consume about 2.5 hours of people’s day on average.

Figure 2 — Where Does Total Screen Time Go?

What I’m looking at: I looked at the full picture of phone usage — not just the six FITSBY apps, but all apps combined. We split total screen time into two slices: the FITSBY apps that the study focuses on, and everything else (messaging, email, maps, games, etc.).

What the results show: The average participant spent 333 minutes — over 5.5 hours — on their phone every day. Of that, 46% (153 minutes) went to just the six FITSBY apps. The other 54% went to everything else. To put it simply: nearly half of all phone time is going to a handful of social media and entertainment apps.

Figure 3 — Do People Feel They Use Their Phones Too Much?

What I’m looking at: At the very start of the study, before any intervention, participants were asked a simple question: do you feel you use your phone too much, too little, or just the right amount? This chunk counts up the responses and plots them as a bar chart. It’s a direct measure of whether people themselves feel their phone use is a problem.

What the results show: 57% of participants — more than half — said they use their phone too much. Only 42% said their usage felt just right, and almost nobody (less than 1%) said they wanted to use it more. This is striking: the majority of people already know they have a problem, even before the study tries to help them.

Figure 4 — Distribution of FITSBY Usage (How Spread Out Is It?)

What I’m looking at: The average can be misleading — it hides a lot of variation between people. This histogram shows the full spread of daily FITSBY usage across all participants. Each bar represents a group of people who use their phone a similar amount per day. The blue dashed line marks the average.

What the results show: The distribution is heavily skewed to the right — most people cluster around 1–2 hours per day, but a significant tail of heavy users goes well beyond that, with some people spending 5–6+ hours daily just on these six apps. The average of ~153 minutes is pulled up by those extreme users. This tells us that phone overuse isn’t a uniform problem — it hits some people much harder than others.

Figure 5 — Predicted vs. Actual Usage (Do People Know Their Own Habits?)

What I’m looking at: At multiple points during the study, participants in the control group were asked to predict how much time they thought they’d spend on their phones in the coming weeks. We then compare those predictions to what they actually did. This tests a key idea in the paper: are people accurate about their own phone use, or do they consistently fool themselves?

What the results show: The gap is small in Period 2 (people predicted 140 min, used 141 — pretty accurate). But by Period 3, the gap widens considerably: people predicted they’d use about 130 minutes per day, but actually used 142 — a 12-minute underestimate. This pattern of underestimation happening repeatedly even after people had already seen their own data is telling. It suggests people aren’t just uninformed about their habits — they keep believing they’ll do better next time, even when they don’t.

Figure 6 — How Did Each Intervention Reduce Phone Use?

What I’m looking at: This is the core result of the study. We compare three groups over time — the control group (no help), the limit group (could set daily app time caps), and the bonus group (paid cash to reduce their usage). All three groups started with similar phone usage. We track what happened to each group across the study periods to see which intervention worked better and whether the effects lasted.

What the results show: All three groups started at roughly the same usage (~154 min/day). By the end of the study (Period 5), the control group had drifted down only slightly to 135 min/day. The bonus group — who were paid to reduce usage — dropped to 125 min/day, a meaningful reduction, but the cash incentive was only active in Period 3, so some of that gain faded. The limit group had the most dramatic and sustained drop, falling all the way to 114 min/day by the end — a reduction of about 40 minutes per day compared to where they started. The key takeaway: giving people a tool to restrict their own behavior (limits) worked better and lasted longer than paying them money to behave better (bonuses).

Figure 7 — The Headline Finding: How Much Phone Use Is Unwanted?

What I’m looking at: The researchers didn’t just measure usage — they built a mathematical model to figure out how much of people’s phone use is actually what they want, versus how much is driven by addiction-like self-control problems. This final chart shows that headline estimate: out of all social media use, what fraction do people wish they could take back?

What the results show: The researchers estimate that 31% of social media use — nearly one third — is not something people actually want. It’s use driven by habit and the inability to stop, not by genuine enjoyment or intention. To put that in concrete terms: if the average person spends 153 minutes on FITSBY apps per day, roughly 47 of those minutes are happening against their own better judgment. This is the core finding of the paper: digital addiction is real, it’s measurable, and it’s responsible for a substantial portion of the time people spend on their phones.

Interactive Visualizations

The charts below go a step further — they’re fully interactive. You can hover over data points for exact numbers, zoom in, click groups on/off in the legend, and in the case of the animation, watch how behavior changed week by week across the study. These are built using plotly (for hover/zoom interactivity) and gganimate (for frame-by-frame animation).

Interactive Figure 1 — Week-by-Week: Watching the Groups Diverge (Animation)

What I’m looking at: The study collected phone usage data every single week for 15 weeks. This animation plays through those 15 weeks one at a time, showing how the average FITSBY usage changed for each group. The key moments to watch: week 4, when screen time limits become available to the limit group; and week 7, when cash bonuses kick in for the bonus group.

What the results show: The animation makes the story unmissable. All three groups start together around 150–155 minutes per day — they’re identical at the baseline, which confirms the random assignment worked. At week 4, the limit group begins to peel away cleanly as participants start setting their daily caps. The most dramatic moment is week 7: the bonus group, now being paid to cut back, plummets from ~140 minutes all the way down to ~80 minutes — nearly cutting their usage in half in a single week. But watch week 10, when the cash stops: the bonus group climbs right back up. The money worked, but only while it was there. The limit group, by contrast, holds steady all the way to week 15 without any ongoing incentive. The animation captures in real time the central lesson of the study: a hard structural constraint on your behavior outlasts a financial reward for good behavior.

Interactive Figure 2 — Did Everyone Improve? One Dot Per Person

What I’m looking at: Instead of looking at group averages, this chart zooms in to the individual level. Each dot represents one of the 1,933 participants. Their position on the horizontal axis shows how much they used FITSBY apps at the start of the study; their position on the vertical axis shows how much they used at the end. The diagonal line is the “no change” line — dots below it mean the person reduced their usage, dots above it mean they used more by the end. Hover over any dot to see the details.

What the results show: The majority of dots sit below the diagonal line, meaning most participants ended the study using their phones less than they started — even in the control group, which received no intervention at all (likely because just being observed and surveyed made people more aware of their habits). The limit group (red dots) clusters noticeably further below the line, especially among heavier users on the right side of the chart — the people who had the most to cut back did so the most. What’s also striking is the spread: at any given baseline level, some people barely changed while others dramatically reduced their usage. This is a reminder that group averages hide a lot of individual variation — the interventions worked on average, but they didn’t work the same way for everyone.

Interactive Figure 3 — Which Apps Do Different Age Groups Use?

What I’m looking at: Not everyone uses their phone the same way. This chart breaks down FITSBY app usage by age group, letting you see which apps dominate for each generation. You can hover over each bar for exact numbers, and click any app name in the legend to show or hide it.

What the results show: The generational divide couldn’t be clearer. Participants aged 18–22 spread their time across YouTube, Snapchat, Instagram, and Browser roughly evenly — they’re platform-hoppers. But as age increases, Facebook becomes more and more dominant, eventually crowding out almost everything else for participants in their 40s and 50s, where it accounts for roughly 90+ minutes of their daily FITSBY time alone. Snapchat is almost entirely a Gen Z app — usage essentially hits zero by the time you reach the 31–40 age group. Instagram declines sharply with age too. Try clicking Snapchat and Instagram off in the legend to isolate Facebook across age groups — the growth curve with age is stark. The practical implication: if you’re designing a digital wellness tool, what you’re targeting needs to look very different depending on who you’re trying to help.

Interactive Figure 4 — The Treatment Effect, Week by Week (Interactive)

What I’m looking at: This is the same story as the animation above, but now fully interactive — you can hover over any point to get the exact usage figure, zoom into any part of the timeline, and click groups on and off in the legend to compare them directly. The shaded bands highlight the three distinct phases of the experiment.

What the results show: Hover over any point to see the exact numbers for each group each week. A few things jump out. First, the control group barely moves — it drifts gently from around 155 minutes in week 1 to about 131 by week 15, but there’s no dramatic change. Without any tool or incentive, people don’t naturally reduce their phone use much on their own. Second, the limit group’s drop at week 4 is immediate and clean — they go from ~151 minutes to ~118 in a single week, then hold that level steadily for the rest of the study. Third, try hovering over the bonus group at weeks 7, 8, and 9 — the numbers are dramatic (around 80–85 minutes), nearly half of where they started. Then hover over week 10 and watch the rebound. By week 15, the final numbers tell the whole story: Control = 131 min, Bonus = 121 min, Limits = 111 min. The limit group wins — by holding a 20-minute-per-day advantage over the control group, with no ongoing incentive required to maintain it.