Every House is the Best House
A statistical comparison of Harvard houses
Don’t worry, freshmen! You’ll love whatever house you end up in. You’ve been told that before, but now we have statistical evidence to back it up.
Background
Before Housing Day 2013, I sent out a survey with a few questions about the Harvard houses. The goal was not to find the “best” or “worst” house on campus. Instead, I wanted to see how each house did on a few different scales, and I wanted to dispel the rumor that the envelope you receive on Housing Day will make or break your Harvard experience.
The idea was far more popular than I expected it to be. On the first day, the survey was sent out over all twelve house lists. At the time I compiled the data, less than two days after the study started, I had 1,025 responses.
That said, this is not a perfect survey. The respondents were far from a simple random sample of Harvard students, and the number of responses varied considerably from house to house. I welcome any suggestions as to how I could improve my methodology.
As this survey was conducted in 2013, some things may have changed. This survey was completed before Stone Hall opened and before Leverett began house renewal. Take this into account when looking at the results.
Surveys & Data
The initial survey consisted of three simple questions:
- Which house are you in?
- If you could go back in time to Housing Day and be placed in any house of your choosing, which house would you pick?
- If you could go back in time to Housing Day and be re-lotteried for a random house, would you do it?
A second survey which ran for a shorter period of time was directed at freshman, and consisted of just one simple question:
- If you could be placed in any house of your choosing this Thursday, which house would you pick?
The upperclassman survey got 819 responses, while the freshman survey got 206. Both sets of raw data are available upon request.
A first look at the data shows how happy students are with their placement. The table below shows what students from each house put as their ideal house. For example, the upper-right cell says that 14% of Winthrop students would go back in time and be placed in Adams.
The overwhelming takeaway here is that for each house, its residents’ most popular choice of ideal house is that house itself. No second choice is even close to half as popular!
But I promised you some statistics, so let’s do some statistics.
House Contentment
We’ll start with a simple measurement of house satisfaction: contentment. We define contentment to be the percentage of students in a house who would rather keep their house than go back in time and re-lottery.
We can’t know any house’s contentment without asking every single student in the house, so we use statistics. For each house, we can get a range* of plausible house contentments based on the information we have.
The first thing to notice is that every house has remarkably high contentment. We can state with near certainty that in no house do more than about one-third of the students think they could do better by reentering the lottery. That is, nearly everyone thinks they’re in a better-than-average house.
Followers of the Flyby rankings may be surprised to see Mather so high up and Pfoho so low (relatively speaking). Of course, this is just one measurement of house satisfaction, and the houses will rank differently by other measurements.
Finally, remember that since these represent ranges of possible actual house contentment, this ranking is not exact. For example, while we can say with 99.995% certainty that Kirkland has higher contentment than Pfoho, we’re only about 60.257% sure that Adams has higher contentment than Cabot.
Overall, 89.9% of respondents were content with their house. Adjusted for the size of each house, we can say that between 86.9% and 91.2% of Harvard students would not re-lottery.
House Supremacy
The next measurement we’ll look at is, for lack of a better word, supremacy. We define supremacy to be the percentage of students in a house who would not rather be in any other house. We should expect this number to be far lower than contentment—if everyone is objective, it will average about one-twelfth, or 8.3%.
Clearly, this is not the case. No house even has a supremacy lower than triple what the average supremacy should be. A wildly disproportionate number of Harvard students think no house is better than theirs.
You might have noticed that Eliot tops this list even though Quincy and Kirkland extend further to the right. The rankings here (and for every other measurement) are based on the middle of the plausible range, which is to say, the best guess at house supremacy. Eliot has a smaller range because it contributed more responses, so we’re more certain about where the true supremacy lands.
Overall, 70.2% of respondents chose to keep their assigned house. Adjusted for the size of each house, we can say that between 65.9% and 72.1% of Harvard students believe they’re in the best house on campus.
House Desirability
The third measurement we’ll consider is the first which takes into account the opinions of non-residents: desirability. We define desirability to be the percentage of students from other houses who consider this house to be the ideal house. We ignore the students form other houses who picked their own house as the best.
This ranking seems heavily influenced by Flyby rankings and similar understandings of what is a “good” or “bad” house (with a couple exceptions). Since the quality of a house is being judged here by people who have never lived in it, much of what they go off of is what they hear. So we have Adams, Eliot, and Quincy topping the list (#1, #2, and #4 of the 2013 Flyby rankings) with Kirkland (Flyby’s #3) not far behind.
The notable exceptions are Lowell and Currier, which rank above Kirkland by this measure despite being #7 and dead last, respectively, on the 2013 Flyby ranking. I might go as far as to credit Lowell’s desirability at the time of this survey to their 2013 housing video, but Currier remains a mystery to me. Looking at the data shows that a smattering of students from all different houses chose Currier as ideal. I’m sure some Currier residents will be more than happy to tell me why!
House Reputation
The fourth and final measure we’ll look at is the only one based on the freshman survey: reputation. We define reputation to be the percentage of freshmen who would like to be placed in this house.
This is only slightly different from the house desirability results, with each house’s placing in the ranking remaining about the same. (The notable exception is Currier, which upperclassmen seem to like a lot more than freshmen.) This might suggest that the only way to change your opinion of a house is to live in it, and a senior in Lowell knows about as much about Mather as a freshman does.
Closing Remarks
There are undeniably patterns in all rankings that show some houses coming out above other houses. And the non-resident rankings seem to somewhat accurately approximate the resident rankings (albeit with a few exceptions). However, we can’t lose sight of the fact that every house ranks exceedingly high on both the satisfaction rankings.
I’ll close with this pie chart.
* All confidence intervals are 95% adjusted Wald intervals. I could have adjusted for small population, since about one-sixth of the Harvard population filled out the survey, but I didn’t so as to simplify the automated calculations. The confidence intervals would be slightly tighter had I adjusted.
