<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Harvard Open Data Project on Medium]]></title>
        <description><![CDATA[Stories by Harvard Open Data Project on Medium]]></description>
        <link>https://medium.com/@harvardopendataproject?source=rss-b312fc49ff05------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*fpW6uneTHnatvmtf.jpg</url>
            <title>Stories by Harvard Open Data Project on Medium</title>
            <link>https://medium.com/@harvardopendataproject?source=rss-b312fc49ff05------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sun, 22 Nov 2026 17:26:57 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@harvardopendataproject/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Open Data on Open Learning]]></title>
            <link>https://medium.com/harvard-open-data-project/open-data-on-open-learning-9b044ba131b1?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/9b044ba131b1</guid>
            <category><![CDATA[university]]></category>
            <category><![CDATA[students]]></category>
            <category><![CDATA[harvard]]></category>
            <category><![CDATA[covid19]]></category>
            <category><![CDATA[online-learning]]></category>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Thu, 30 Jul 2020 16:19:09 GMT</pubDate>
            <atom:updated>2020-07-30T16:19:09.093Z</atom:updated>
            <content:encoded><![CDATA[<p>An Analysis of HarvardX Course Enrollment</p><p>By: Karina Halevy, Robert McKenzie, and Neha Gupta</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YjIuc2GS-uUrWKrE" /><figcaption>HarvardX <em>(</em><a href="https://www.thecrimson.com/article/2016/8/10/new-online-harvardxplus/"><em>Source</em></a><em>)</em></figcaption></figure><p>Online learning has become the default form of education in light of the COVID-19 pandemic. Many higher education institutions are not only converting their curricula to be delivered digitally but also releasing their course materials to the public through several open learning platforms. Unless students are seeking a verified certificate, all of these courses are free. Harvard is one of the institutions that has long been publishing educational content online.</p><blockquote><strong><em>In light of the pandemic, we thought it would be pertinent to analyze some of the trends in Harvard’s online course enrollments.</em></strong></blockquote><p>In this project, we aim to analyze some of the driving factors behind enrollment in <a href="http://learn.edx.org/partner-harvard/?g_acctid=926-195-8061&amp;g_campaign=new-gs-us-brand-partner-harvard&amp;g_campaignid=1661780386&amp;g_adgroupid=62996190414&amp;g_adid=320507630783&amp;g_keyword=%2Bedx%20%2Bharvard&amp;g_keywordid=kwd-350135246388&amp;g_network=g?utm_source=adwords&amp;utm_campaign=1661780386&amp;utm_medium=62996190414&amp;utm_term=+edx%20+harvard&amp;gclid=Cj0KCQjwy6T1BRDXARIsAIqCTXo1XnfTKaF3ermXvNFl57LtWqBJastPLs_hZqtxsgMZkNPoTt2zOMkaAvHPEALw_wcB">online courses offered by Harvard</a> through the <a href="https://www.edx.org/">EdX</a> platform.</p><h3>Methodology</h3><p>At the time of our analysis (4/28/20), HarvardX offered 109 courses, with 12 session-based courses starting soon. As the EdX course page turned out to be somewhat unintuitive to navigate, we manually scraped data and recorded each course’s Enrollment, Subject, Course Length, Weekly Workload (Effort), Price for Certificate, and Level of Difficulty. Enrollment is a raw number representing the number of user accounts registered for the course. Subject is a categorical variable with values including Archaeology, Art and Culture, Biology and Life Sciences, Business and Management, Chemistry, Computer Science, Data Analysis and Statistics, Education, Environmental Studies, Health and Safety, History, Humanities, Literature, Math, Medicine, Music, Physics, Science, and Social Sciences. Course length is a discrete integer-valued variable in units of weeks. Effort is a range in units of hours per week from low to high, both of which are integer values. Price is a continuous variable measured in US dollars. Finally, Level of Difficulty is a categorical variable with values of Introductory, Intermediate, or Advanced.</p><p>We split the weekly workload variable, which was originally a range, into a lower limit (Effort Minimum), upper limit (Effort Maximum), and average (Effort Mean). To minimize the effect of potential outliers in the data, we also removed two courses which had very large enrollments and were the only course offered in their field: Leaders of Learning (categorized as Communication) and The Architectural Imagination (categorized as Archaeology), since those were both offered by Harvard Graduate Schools and were heavily marketed by edX. We also removed two additional courses with very high enrollments: CS50 and Mechanical Ventilation for COVID-19. CS50, one of the first open courses on the platform, has more than double the enrollment of the next most popular course, and the COVID course was a substantial outlier in its length, field, objective, and price.</p><p>We performed a few regressions on the data to try and understand some of the enrollment trends. The model we ended up building was only able to predict 22% of the variance in enrollment numbers, but it did reveal some predictors as significant, namely Course Length, Field of Study, and the Price for Certificate.</p><h3>Results and Discussion</h3><p>A Few Explanatory Variables</p><p>First, we examined enrollment in relation to course length, which seemed to be one of the more explanatory variables.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*LxnCb7OPUriv0Niw" /></figure><p>From the graph, we see that there is a weak positive correlation between enrollment and length.</p><blockquote><strong><em>This may be because people see longer courses as more valuable and are more likely to at least start the courses by enrolling. It may also be the case that longer courses offer more solid training in marketable and employable skills, thus driving enrollment numbers up.</em></strong></blockquote><p>Next, we examined enrollment by subject, which also seemed to be one of the more influential variables.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*gQjmF5vO_hViYRnC" /></figure><p>The most popular categories are Business and Management, Computer Science, Biology and Life Sciences, Chemistry, and Data Analysis and Statistics.</p><blockquote><strong><em>This may be explained by the fact that skills in these categories are in high demand (or are at least marketed to job-seekers as high-demand skills). It has been floated that </em></strong><a href="https://hbr.org/2012/10/data-scientist-the-sexiest-job-of-the-21st-century"><strong><em>data science is the hottest job of the 21st century</em></strong></a><strong><em>, so that may be driving enrollment in data-related courses and courses in closely tied fields.</em></strong></blockquote><p>In contrast, the least popular categories seem to be music, art and culture, and some of the more theoretical natural sciences such as physics. This may also be explained by the fact that skills in these categories are not only less practically marketable but also more likely to be appreciated through formats that are not massive open online courses.</p><p>Next, we looked at the price of the courses’ verified certificates.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*fXWjfYvB3luU1dyJ" /></figure><p>The graph shows a weak positive correlation between price and enrollment.</p><blockquote><strong><em>This might be because people might perceive more expensive courses as more valuable and hence enroll in greater numbers.</em></strong></blockquote><p>Going in the opposite direction, more valuable courses (e.g. those that teach more marketable skills) may also be priced at higher levels.</p><p>We also decided to explore some variables that didn’t show up in linear modelling but might offer some insight into the shape of the data.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*HPTfs39CjQCpunai" /></figure><p>We looked at enrollment by course level, and found that there seems to be the most enrollment in intermediate level courses. While there wasn’t enough Intermediate or Advanced data to confirm this relationship was significant, the “hottest” skills and courses were usually classified as intermediate.</p><p>Finally, we looked at the maximum effort listed on each course description.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*580BWAlW9SFzKuS3" /></figure><p>Between zero and nine hours per week, there seems to be an exponential positive correlation between enrollment and maximum effort. This might be because people see the courses that involve more effort as more valuable to their careers. However, beyond ten hours, there is a significant drop in enrollment. While this could possibly just be because of a few outliers, another explanation might be that online learners simply do not have that much time on a weekly basis to dedicate to their courses.</p><h3>Conclusion</h3><p>There is still much to be explored about what makes people enroll in online courses — from factors such as enrollment trends over time that we have not accounted for in this analysis to breakdowns of how many of these enrollees actually follow through with the courses and how many of those get verified certificates. However, this preliminary analysis does show that popular online courses don’t typically focus on traditionally academic subjects such math, science, or history.</p><blockquote><strong><em>Popular courses help thousands of learners get job-ready skills. With unemployment rates skyrocketing during the pandemic, these professionally oriented online courses can be even more important in helping learners get hired.</em></strong></blockquote><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. To see our raw data,</em><a href="https://docs.google.com/spreadsheets/d/178cykkOAZ7PvGmXtMy_usQZ3yrjbwXAB0pv2IeQM0bo/edit#gid=1451488234"><em> </em></a><em>look </em><a href="https://github.com/ENSCMA2/hodp-open-learning"><strong><em>here</em></strong></a>.</p><p><em>Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join us!</em></a></p><p><em>Thanks to Karina Halevy, Robert McKenzie, Neha Gupta, Alison Hu, and Henry Austin for help with this article.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9b044ba131b1" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/open-data-on-open-learning-9b044ba131b1">Open Data on Open Learning</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What will Harvard students do this Fall?]]></title>
            <link>https://medium.com/harvard-open-data-project/what-will-harvard-students-do-this-fall-cdccb698eb61?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/cdccb698eb61</guid>
            <category><![CDATA[students]]></category>
            <category><![CDATA[college]]></category>
            <category><![CDATA[university]]></category>
            <category><![CDATA[covid19]]></category>
            <category><![CDATA[harvard]]></category>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Mon, 13 Jul 2020 00:52:53 GMT</pubDate>
            <atom:updated>2020-07-13T01:01:35.092Z</atom:updated>
            <content:encoded><![CDATA[<p><em>A data-driven look at how Harvard students are responding to the university’s policy for the Fall 2020 semester.</em></p><p>By Kevin Bi and Sahana Srinivasan</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*gr7Obg4yE0pXJqLDXGsJuA.png" /><figcaption>Graphic by Leo Saenger and Isabelle Zheng</figcaption></figure><h3>At a glance</h3><ul><li>70 percent of students are at least somewhat likely to enroll, and 15 percent are extremely unlikely</li><li>First-years approve of Harvard policy more than upper-class students</li><li>40 percent of off-campus enrollees will try to live with Harvard friends</li><li>Majority of students are unhappy with the plan overall (average rating of 5.1/10)</li><li>Students who predicted more people in their year would enroll are likelier to enroll</li><li>Most students want more assistance for students on financial aid</li></ul><p>On July 6th, Harvard students woke up to an all too familiar sight: an email from the administration with a major announcement about Harvard policies regarding the COVID-19 pandemic. The administration announced that the Fall 2020 semester would be taught entirely online, and that only incoming first-years and a subset of approved upper-level students would be allowed to return to campus. Everyone else would spend another semester learning remotely, with seniors prioritized to return in the Spring.</p><p>We at the Harvard College Open Data Project wanted to see how students reacted to the new policy. We sent out a survey to the student body on the evening of Tuesday, July 7th, one day after the announcement, that ran until the evening of July 11th. In total, we received <strong>1766 </strong>responses, which represents just above one quarter of the undergraduate student body, including incoming first-years. Here is what we found.</p><h3>Who is coming back?</h3><p>With the majority of students not allowed to return to campus, and with heavy restrictions in place for those who are, many students are considering taking the Fall semester off and returning when things are more normal.</p><blockquote><strong><em>The majority of students, 74 percent, told us that they were at least somewhat likely to enroll for the Fall, while a sizable minority, 15 percent told us that they were extremely unlikely to return.</em></strong></blockquote><p>The results also heavily differed by class year, with first-years significantly more likely to enroll than upper level students. (The high enrollment of first-years could be because they are the only class cohort allowed on campus in the Fall, because incoming first-years have already had a chance to defer, or some combination of both.)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*P4tnlef0aNZ8Ef4P" /></figure><p>Our sample had considerably more incoming first-year respondents than respondents from any other class year, so we re-weighted the responses based on how many students are actually in each class year.</p><blockquote><strong><em>After this adjustment, around 70.6 percent of the student body is extremely or somewhat likely to return.</em></strong></blockquote><p>We also examined enrollment likelihood among international students and based on financial aid status. The same day that Harvard announced its policy, Immigration and Customs Enforcement (ICE) announced that international students would not be allowed to remain in the United States if courses at their school were entirely online.</p><blockquote><strong><em>Predictably, international students were less likely than the general population to favor enrolling, given that they would not be allowed in the U.S. and would potentially be forced to live in a very different timezone. Conversely, students on financial aid (53 percent of our sample) were more likely to enroll than the student body at large.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*zC__IYUwgvpI2Eli" /></figure><p>To get a slightly better estimate of fall enrollees, we conducted simulations assuming that <strong>99</strong> percent of “Extremely likely,” <strong>75 </strong>percent of “Somewhat likely,” <strong>25 </strong>percent of “Somewhat unlikely,” and <strong>1</strong> percent of “Extremely unlikely” respondents would enroll in the Fall. Our simulations predicted that around <strong>68 </strong>percent of students would enroll.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*HxhMYpsrVu-Hgrbs" /></figure><p>Another way of trying to gauge the percentage of students who will enroll is to use the wisdom of the crowds, so we asked students what they predicted the overall enrollment would be. The average prediction was around <strong>70</strong> percent, which is exactly the reported proportion of students likely to enroll after re-weighting for class year. The distribution of predictions is below, with the dashed line representing the (unweighted) proportion of respondents who said they were extremely or somewhat likely to enroll:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*JZZownt4_T_usmhj" /></figure><p>We also asked students how many students in their class year they thought would enroll, and once again, the results are more or less consistent with what students were reporting about their own enrollment:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*mSSClNisL9qFSojE" /></figure><p>One interesting point that we found in an <a href="https://medium.com/harvard-open-data-project/harvard-in-the-age-of-covid-19-1b06e6501335">earlier analysis</a> was that students were most likely to enroll if they thought their friends would enroll. We tested that prediction by comparing a respondent’s individual likelihood of enrolling to the proportion of students they predicted would enroll in the fall. We found that respondents who thought a greater proportion of students would enroll were more likely to enroll themselves:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*g7fTS3si25aj-rIf" /></figure><p>The relationship was even stronger when comparing individuals’ likelihood of enrolling with the share of students they predicted would enroll within their class year. This makes sense because most students likely want to stay with friends in the same year, so students who think most of their friends are not coming back would also be less likely to enroll themselves:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*x3g96-aziSirc5Rl" /></figure><p>With many students considering taking time off, we wanted to understand if they would only take one semester off, or an entire year. For the most part, students who were considering taking time off said they would take the Fall off, and then wait and see for the Spring. However, around a third of students reported that if they took time off, they would plan to take the entire year off rather than just one semester:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*6sNG-DxmaCx4oRlK" /></figure><p>We also wanted to see what restless Harvard students might do if they took time off.</p><blockquote><strong><em>65 percent of students reported that they would try to find an internship or some type of industry work if they took time off.</em></strong></blockquote><p>Only <strong>3 </strong>percent of students said they would “Chill at home” without seeking other opportunities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*L1n5nu6ScpZHp1PP" /></figure><h3>On campus or at home?</h3><p>While only incoming first-years were guaranteed to return to campus, Harvard is allowing upper-level students to petition to return to campus under certain circumstances.<strong> 19 </strong>percent<strong> </strong>of upper-class students reported that they were planning to petition to return.</p><p>Under Harvard’s petitioning process, students must indicate which criteria for needing to be on campus apply to them. We posted these same criteria in our survey, and found that the most common criteria satisfied by students had to do with mental health or difficulties working at home:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*_u3K4nENkSz0KHUQ" /></figure><p><strong>42</strong> percent of students indicated that they would not satisfy any of the criteria, with an average of <strong>1.26 </strong>criteria satisfied per student.</p><p>Students were generally pessimistic about their chances of returning. Among upper-class students planning to enroll, upwards of<strong> 60</strong> percent thought that they would be extremely unlikely to be eligible to return.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*L7RMyIU9cQNpgXkh" /></figure><p>Most students however, would be happy to return if given the chance. Over <strong>80</strong> percent of all respondents said that, if eligible, they would be extremely or somewhat likely to go on-campus in the Fall.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*pIrcV9Zsfl3vR1kA" /></figure><p>To gauge whether students were more likely to return if they thought the risk of infection was lower, we asked students to predict what proportion of students on campus they believed would be infected by COVID-19 during the semester. We found that students who were more likely to return if allowed had had a lower perception of the risk of infection from COVID:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*eu3xAIhu0QuNHzg3" /></figure><p>We also asked students to predict how many students would actually be on campus among those enrolled. On average, students predicted that 40 percent of enrollees would be on campus, compared to 43 percent of students who reported that they would be enrolling and likely on campus in our survey. This was about on par with the administration’s stated intention to bring back 40 percent of the student body on campus:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*dxZetBJYt0TYyVij" /></figure><p>Finally, we were curious about where students would take classes if they were not allowed to return. Many students had discussed renting apartments or AirBnbs to live with friends during the semester. Though 85% of students in an <a href="https://medium.com/harvard-open-data-project/harvard-in-the-age-of-covid-19-1b06e6501335">earlier survey</a> reported living in their family home after leaving campus in March, we found that only half of students planned to take courses from a family home in the Fall.</p><blockquote><strong><em>A considerable portion (40 percent) of likely off-campus enrollees indicated that they would try to live with Harvard friends.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*szgj44DJl7hGQP1w" /></figure><h3>Overall reactions</h3><p>This survey was by far the most popular survey that HODP has ever sent out, likely due to the fact that students have strong opinions about Harvard’s policies.</p><blockquote><strong><em>We asked students to evaluate the policy overall, and it was met with a generally lukewarm response, with an average rating of 5.15 on a scale of 1 to 10:</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*b769VdDaItgEielx" /></figure><p>Unsurprisingly, first-years, who are allowed on campus in the Fall, rated the Fall plan a full point higher than the next cohort. Seniors, who are slated to be allowed back in the Spring, were the next most approving.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*S1c0yCLhbQ_yuQRr" /></figure><p>Most respondents felt that too few students were allowed to return to campus, although disaggregating results by year shows a number of upper-level students agree with the number of students allowed to return, but thought a different cohort should have been chosen in the Fall.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*MMGyurNjRGeVdZXT" /></figure><p>Students were more satisfied with the number of precautionary measures taken to protect on-campus students, which includes testing every three days and heavily restricted movement. On-campus students will have swipe access only to Harvard University Health Services, their dorm, and one assigned dining hall. No guests (including other Harvard students) are allowed in dorms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*Gl1P8P1forLY_1ye" /></figure><p>We can gauge how effective students think these precautions will be based on how many students they think will be infected during the Fall semester. The distribution was right-skewed, so the mean prediction of <strong>22</strong> percent was considerably higher than the median prediction of <strong>15 </strong>percent of the on-campus students being infected by COVID-19.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*tu5mm_Tc9kAn2ASX" /></figure><p>In addition to announcing the policy on returning to campus, Harvard also announced that they would take measures to address a variety of concerns students had with another virtual semester. For the most part, respondents were most unsatisfied with how Harvard addressed maintaining social interactions both on and off campus, but thought the administration did a better job addressing equity of access to school resources.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*DOYMGi7SDzGx5yGu" /></figure><p>We were also curious about how international students or students on financial aid rated Harvard’s policies. We did not survey students on the ICE order or how it affects Harvard’s plan since the order was released after Harvard announced their plans for the Fall (Harvard and MIT have since <a href="https://www.thecrimson.com/article/2020/7/8/harvard-mit-sue-immigration-authorities/">sued</a> ICE, a move supported by many universities nationwide). However, we did want to check if international students’ opinions of the plans were notably different from their peers’.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*IyFA4Aq3vE_kALHy" /></figure><p>We also wanted to gauge reactions to the plan’s accommodations for students on financial aid. Harvard announced that enrolled, off-campus students on financial aid will receive $5,000 a semester (in lieu of traditionally receiving aid that covers dorm costs) for their living expenses.</p><blockquote><strong><em>43 percent of students believe Harvard isn’t providing enough money to students on financial aid, and a little over half believe the policy should have been individualized to student needs. 32 percent took no issue with the policy.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*fEYAaAm1Vq8B-VeV" /></figure><p>Note that these percentages don’t add up to 100, since students could select multiple concerns (for example, that the money given should be increased and individualized).</p><p><strong>53</strong> percent of our respondents reported being on some form of financial aid, matching almost exactly the <a href="https://college.harvard.edu/financial-aid">55 percent of students</a> at the College on some aid. Financial aid policy opinions among students on financial aid matched the opinions of the overall sample.</p><p>In our <a href="https://medium.com/harvard-open-data-project/harvard-in-the-age-of-covid-19">last analysis</a> of COVID-19 at Harvard, a majority of students favored a fully on-campus semester, but a plurality of respondents also said that if they were administrators, they would only allow some students on campus, which is what administrators actually did.</p><p>Now that Harvard and many of its peer colleges have concretized plans, we wanted to see which fully fleshed-out plan students preferred. Many other Boston-area and Ivy League institutions are allowing more students on campus at a time and are allowing each class year back at some point.</p><blockquote><strong><em>Yale and Cornell stood out as having Harvard students’ preferred plans.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1022/0*4NZ8GTkNA1wLgXO9" /></figure><p>Yale announced a hybrid remote and in-person instructional model, with almost all undergraduate classes to be remote. First-years, juniors, and seniors will be allowed on campus in the Fall, and sophomores, juniors, and seniors in the Spring, maintaining a 60 percent campus density compared to Harvard’s 40 percent. Cornell will welcome its entire undergraduate body back to campus in the Fall, also with a hybrid learning model. Both universities will impose social distancing rules on residential students.</p><p>Finally, we asked students for their qualitative thoughts on the policy and gave them a long-form answer box to reply.</p><blockquote><strong><em>Most students mentioned wanting to be on-campus, while many also pointed out issues with tuition (which Harvard has left unchanged), social interactions, mental health, and online classes.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ZfnfWPHiTJaezmFZ" /><figcaption>A cloud of the most frequent words in students’ free-text responses. Common words such as “the” and “and” were removed, as were “students” and “Harvard.”</figcaption></figure><p>We performed a sentiment analysis of the comments using a <a href="https://github.com/huggingface/transformers">PyTorch Transformer</a> to gauge whether they were positive or negative. Three quarters of responses to Harvard’s policy were classified as negative, with first-years generally more positive than upper-class students. However, it is important to note only around half of respondents left a comment, and that those leaving a comment likely also had the strongest feelings about the policy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*hlulJRPLvI0OT6nD" /><figcaption>Sentiment analysis by Asher Noel</figcaption></figure><h3>Sample Representativeness</h3><p>We want to be transparent about the representativeness of our survey sample: We emailed the survey to undergraduate students in the classes of 2021, 2022, 2023, and 2024. We had <strong>1766 </strong>respondents. <strong>54<em> </em></strong>percent of respondents reported being on financial aid, and <strong>13</strong> percent were international students, compared to 55 percent of students and 12 percent of students in the College overall, respectively.</p><p>The racial and ethnic demographics of our respondents roughly match that of the College, though Black students were underrepresented by about <strong>4</strong> percentage points (<strong>10</strong> percent of our sample compared to <strong>14</strong> percent of the College).</p><p>The most significant skew in our data was an overrepresentation of incoming first-years. Members of the class of 2024 comprised about <strong>38</strong> percent of our sample while comprising around a quarter of the student body overall. The proportions of sophomores, juniors, and seniors were roughly equal. To account for this, we re-weighted our data for visualizations in which we saw differences by year, such as enrollment, so that first-years accounted for a quarter.</p><h3>Final thoughts</h3><p>It is fair to say that Harvard’s policy is not what most students were hoping for, but it’s also fair to say that with COVID-19, the semester could never be normal. Students are in for another semester of major disruptions to their learning, friendships, and college experiences. We are facing an unprecedented number of students planning to take time off to wait out the worst of the pandemic’s effects.</p><p>If students cannot return to campus for the Spring semester, many will have lost more than a quarter of their college career to COVID-19. And even if students are allowed back in the Spring, it will be a very different campus than the one they left in March 2020.</p><p>We are extremely grateful to the many students who filled out our survey, without whom this analysis would not have been possible. For students who are considering petitioning to live on campus, the deadline is tomorrow, Monday, July 13th. The non-binding deadline to decide to enroll or take a leave for the Fall semester is July 24th.</p><p>We hope that our analysis here has helped inform the Harvard community and will assist them as they continue to plan for the Fall semester.</p><p><em>This article is an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. Check out the code and anonymized data for our analysis </em><a href="https://github.com/TheKeveloper/HODP-Fall-2020-Analysis"><em>here</em></a><em>.</em></p><p><em>We’d like to thank Thuan Tran for building and sending the survey with us, and Asher Noel for doing the sentiment analysis. Edited by Asher Noel, cover image by Leo Saenger, and graph style by Seth Billiau.</em></p><p><em>Want to work on more projects like this? </em><a href="http://hodp.org/"><em>Join us!</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cdccb698eb61" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/what-will-harvard-students-do-this-fall-cdccb698eb61">What will Harvard students do this Fall?</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Predicting the Straus Cup Champion]]></title>
            <link>https://medium.com/harvard-open-data-project/predicting-the-straus-cup-champion-c1a23dd5a988?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/c1a23dd5a988</guid>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Sat, 11 Jul 2020 21:44:12 GMT</pubDate>
            <atom:updated>2020-07-11T21:44:12.515Z</atom:updated>
            <content:encoded><![CDATA[<h3>Predicting the 2020 Straus Cup Champion</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*3fbxV19fshr522-pz6_sOw.png" /></figure><p>By Samuel Thau</p><h3>Introduction</h3><p>One of the many things we missed out on during the Spring was the conclusion of the intramural season. Here is what the rankings looked like when we left campus:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*feKYdjBWJiXKr3uWA7IIGw.png" /><figcaption>(<a href="https://intramural-manager.com/league/harvard-intramurals-19-20/scoreboard/">Source</a>)</figcaption></figure><p>Mather was in the lead, and was therefore awarded the coveted Straus Cup for the second year in a row. But what would have happened if we had been able to play out the spring? Would Mather have held on? Could Kirkland, Currier or Lowell stage a comeback?</p><h3>The Data</h3><p>To tackle the problem, I started examining IM data from past years. Thanks to the current leaders of the intramural referees, and through the current <a href="https://intramural-manager.com/league/harvard-intramurals-19-20/">IM website</a>, I got access to the points awarded from every sport during every season, ranging from the fall of 2012 to the winter of 2020. From 2012 to 2018, the data was given directly to me in the form of Google Sheets, and for 2018 onward I copied the data from the IM website.</p><p>Taking a first glance at the data, some complicating factors jump out. First, the scoring systems for sports vary greatly from sport to sport. In addition, the rules have changed for how certain sports are scored over time, making cross year comparison more difficult. Secondly, which sports are played changes annually: for example, broomball replaced hockey starting in 2016–2017, and badminton was not played during the 2016–2017 or 2017–2018 seasons.</p><p>Understanding the rules of how IMs are scored is vital to understanding the IM data. Broadly, there are four categories of ways to score points for IMs:</p><ol><li>Sports played for wins and losses</li><li>Tournaments or one shots</li><li>Participation</li><li>Other</li></ol><p>Sports played for wins and losses are sports like soccer, basketball, and tennis. Tournaments or one shots are sports like spike ball, climbing, and crew that are only played once. Also included in this category are soccer, frisbee, and flag football in the spring.</p><p>Participation currently includes B volleyball (since 2018–2019), Harvard on the Move, and River Run (literally running, not the pre-housing day tradition). While there are some points allotted for placing, the vast majority are based on participation. For volleyball, there are two “versions” of the sport that are played: A, which is generally more competitive, and B, which tends to be people playing for fun. Harvard on the Move and River Run are both runs, scored on participation; River Run happens twice a year, and Harvard on the Move happens weekly. Finally, points in the other category are things like meeting attendance. Going forward, I will refer to the first three categories as the event categories.</p><h3>First Impressions</h3><p>To predict the winner of the Straus Cup for the 2019–2020 season, I started with examining approximately how much information we lost out on from spring 2020. On average, from 2012–2013 through 2018–2019, houses earned 45.77% of their points in the spring, the largest percentage. Fall makes up 33.99%, and winter makes up only 20.26% of the points earned. This effect seems to be driven by the number of sports in each season; in general, there are fewer sports in the winter, and there are often more tournaments or one shots in the spring than fall meaning more points are up for grabs.</p><p>A second natural question to ask is to see if houses that do better in the standings pick up more of their points during the spring, or fewer. To measure this, examine the average percentage of points earned in the spring by rank, average from the 2012–2013 through 2018–2019 years:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/808/1*1E5SfDuGdD4E0dExHwkdHQ.png" /></figure><p>This table shows the average percentage of points by final ranking in the Straus Cup standings. The only outlier is that houses that finish last have a larger proportion of points from the spring on average, but otherwise, the percentages are relatively consistent across rankings.</p><p>One other baseline to look at is breaking down scoring by category. One could hypothesize that Straus Cup winners dominate the tournaments and one shots, while only being above average for events played for wins and losses, because people are more willing to show up to one time events. To examine this possibility, I converted the raw point totals into z-scores, by year and sport. For example, the distribution of fall soccer points in 2015–2016 was rescaled to have mean 0 and standard deviation 1, in order to make comparison easier. Sports are scored differently in each category, and have very different variances, so standardizing the data makes it easier to compare and interpret. First, I averaged the z-scores across categories, and then across rankings as before.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*oXwYJwgjvNmThTjLKzmxrg.png" /></figure><blockquote><strong><em>The first takeaway is that houses that win the Straus Cup are the best at everything in the event categories. Winners average 1.71 standard deviations above average for total points, and also average more than 1.5 standard deviations above average for sports in the event categories.</em></strong></blockquote><p>Secondly, there is some clustering: the top two houses tend to be significantly better than the rest of the field across every event category. After the top two, there is a gradual decline, and then a steep drop to last place.</p><h3>Data Cleaning</h3><p>Before we continue, I need to make three notes on the scoring systems and related data cleaning. It used to be that A, B, and C basketball were scored independently, until the 2014–2015 year. However, since then, only the positive points for the largest two divisions were counted, but negative points from forfeits are counted from all three divisions. I know this sounds confusing, which is why I combined all the points from basketball, both positive and negative, into one scoring category called “totalBB.” While this may not be an actual scoring category, it captures more information and is easier to understand when being used in regressions.</p><p>Second, because the scoring systems changed across years, there needs to be some adjustment of the raw scores. To do this, I subtracted the mean point value of each sport, by year (i.e., the mean of flag football in 2014–2015 was normalized to 0). There are three reasons for this: it standardizes the data across overall scoring systems, preserves variation in scoring among sports, and makes it easier to account for missing data. I wanted the regression to take into account that some sports could have more value than others, because there could be value in excelling in particular sports. For example, it could be that performance in soccer is indicative of having better athletes, which would give an advantage throughout the year.</p><p>Missing data turns out to be an issue because of how often sports rotate in and out of being played. Roughly a third of the sports played in the 2019–2020 fall and winter have not been played every year in the dataset. To remedy this, I set those missing to values to 0; because the mean of each sport is normalized to 0, this means that for years where a sport was not played, every house is given the score of being exactly average.</p><h3>Regressions</h3><p>Now, we come to the exciting part: predicting the Straus Cup winner from this past year. There are many ways to approach this problem, but I decided on a relatively straightforward method: linear regression.</p><h4>Regression #1</h4><p>We begin with a simple regression: let’s examine how well the rankings after the winter season (based on points in the fall and winter) predict the full year rankings. Then, we use the regression to predict the 2019–2020 results.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xVPje5SJmOe4wyq-rq_8KQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*H9uTS3x7vT7xoG22rwv6Jg.png" /></figure><p>There are a few important features to note about this regression: first, a line fits the data quite well, and secondly, the coefficient on the ranking after winter is positive and statistically significant. The goodness of fit by the line is demonstrated by the R squared of 0.913, which indicates that 91.3% of the variance in the data is accounted for by the rank after winter. Secondly, the coefficient on rank after winter is positive and close to one, meaning that there is close to one to one correlation between ranking after winter and the final rankings. This result fits with the basic hypothesis that houses that do better in the fall and winter can carry forward that performance in the spring.</p><blockquote><strong><em>For this regression, we predict that Mather would have won the Straus Cup, with Kirkland in second place.</em></strong></blockquote><p>The predicted rankings are not whole numbers, but we can still compare the relative ranks to crown a theoretical champion.</p><h4>Regression #2</h4><p>A second, and slightly more sophisticated, approach to the linear regression model is to calculate a regression of the end of year rank on all of the sports that were played in the fall and winter of the 2019–2020 year.</p><p>Two things jump out immediately: the magnitude of the coefficients and the relatively small number of statistically significant coefficients. The small magnitudes (outside of the intercept), makes sense intuitively: racking up a huge number of points in soccer can only help your ranking so much, because soccer is only one of many sports. Second, most of the coefficients are not statistically significant, and because of the small magnitudes, saying they have an effect on the rankings may not be true.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-mesXz4U5CHMFxnijqt4-g.png" /></figure><blockquote><strong><em>In addition, there are some plainly weird implications from the coefficients.</em></strong></blockquote><p>Negative coefficients make intuitive sense: as you score more points, your ranking goes down, which is the goal. That means that any positive coefficients, like for B volleyball and spike ball, suggest that doing well at those sports actually <em>hurts</em> your chances of winning the Straus Cup. Given that none of the positive coefficients are statistically significant, it’s hard to take the correlation seriously, but the idea is amusing nonetheless. Despite the strange results, let’s examine the predicted rankings from this year:</p><p>Again, Mather is predicted to come out on top, with Kirkland in second place. This regression has Kirkland slightly farther behind than the first one. Winthrop has the largest improvement, by almost 4 spots, and Adams has the worst drop off ordinally though the predicted rank stays relatively close (8.91 vs 9.31).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hqQPvxXAPdclpBYmeKoDuA.png" /></figure><h3>Confounding Factors</h3><p>It’s important to remember that the points from just the scores themselves may not contain all the relevant information, and there could be confounding factors that systematically skew the results. Across years, there could be house effects, meaning that certain houses consistently over or underperform their predictions. Within a year, it could be that a particular class is extremely athletic or enthusiastic in one particular house, so they carry that house to victory. To look for house effects, I began by plotting the residuals, meaning the actual rank minus the predicted rank, differentiated by house.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/0*4d_yrCj06DMXAIkE" /></figure><p>This graph is noisy, and it is hard to see consistent trends, which is actually a good thing as it suggests there are no consistent house effects. Take a closer look at the top ranked houses from the 2019–2020 season:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/0*swLva_31SlxdheLS" /></figure><p>Again, no consistent patterns. All four bounce around, meaning that it is hard to say there are consistent house effects. Based on that, it’s hard to draw strong support for including house effects as a control in the regression.</p><p>Next, to address having a particularly enthusiastic or athletic class, I incorporated the rankings from the previous two years into the left hand side of the regression.</p><blockquote><strong><em>The hope is to take advantage of continuity across years in the house community: because classes stay in a house for three years (ignoring transfers), past rankings could serve as a proxy for some outside force, like enthusiasm or athletic ability.</em></strong></blockquote><h4>Regression #3</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*juLfKNSOnx0it_BakzbVig.png" /></figure><p>Similar to Regression #2, the coefficients are relatively small, and there are relatively few statistically significant coefficients, meaning that it is hard to take this regression particularly seriously. Still, examining the coefficients on the previous years’ rankings yields some interesting hypotheses (though the hypotheses themselves have no evidence from this data). The coefficient on rankings from one year prior is negative, suggesting a small positive correlation between previous years ranking and the current ranking. This impact could be the athletic ability or enthusiasm carryover from year to year. Ranking from two years prior has a much stronger negative impact, which could indicate some sort of return to the middle of the pack. However, neither is statistically significant at any level, meaning that it is hard to take either impact particularly seriously. Finally, look at the new rankings generated by this regression, mostly for fun:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/810/0*qzjCSLyGr8vwnmE-" /></figure><p>Again, Mather is predicted to take the top spot over Kirkland, this time by a slightly larger margin. Given that the Mather was the defending Straus Cup holder, and came in second two years ago, this result is not particularly surprising.</p><h3>Prediction Intervals</h3><p>One final piece of evaluating these regression models is to look at the prediction intervals for the predicted standings. Prediction intervals are similar to confidence intervals, but are meant for single point estimates like the predicted values from a regression. These prediction intervals give a 95% interval for the range of the predictions, similar to a confidence interval.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*a-qkFK3cM1X3vITP_g0w-g.png" /></figure><p>In all three cases, the intervals overlap by a substantial degree, showing the weakness of these regression models.</p><blockquote><strong><em>It’s hard to say that Mather would definitively beat Kirkland when the prediction intervals overlap this much, meaning that we should not put too much stock in the predictions made by regressions.</em></strong></blockquote><h3>Further Questions</h3><p>There are still many interesting questions to explore in the area of IM sports that could yield interesting follow ups or extensions to this article. One extension would be to try to come up with predictions that have more promise using more sophisticated methods. One definite possibility could be K nearest neighbors regression, given the relatively limited data. There are also probability questions to explore, like figuring out the probability that Mather actually ends up ahead of Kirkland based on the regression (or other methods) predictions.</p><p>There are also other interesting questions related to the enthusiasm question raised earlier. Do houses that perform well early in the fall inspire more participation, improving year long performance? Are there correlations between certain sports and houses (due to distance from fields or courts)? Can we isolate the impact of a particular class of students in a house?</p><h3>Conclusion</h3><p>After all that, it seems that confidently predicting the winner of the Straus Cup will take some more effort, meaning that for now, we have to be content with the results from the partial year and from our preliminary regressions. Congratulations to Mather on the back to back Straus Cup victories!</p><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. The code and data for this analysis can be found here: </em><a href="https://github.com/samthau/HODP_IM_Sports"><em>https://github.com/samthau/HODP_IM_Sports</em></a><em>.</em></p><p><em>Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join us!</em></a></p><p><em>Thanks to Leo Saenger for advice on this article, and to Nicole Kim and Suriya Kandaswamy with the Harvard Referees for help getting access to the data. Cover image by Leo Saenger. Edited by Alison Hu and Sahana Srinivasan.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c1a23dd5a988" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/predicting-the-straus-cup-champion-c1a23dd5a988">Predicting the Straus Cup Champion</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Harvard in the Age of COVID-19]]></title>
            <link>https://medium.com/harvard-open-data-project/harvard-in-the-age-of-covid-19-1b06e6501335?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/1b06e6501335</guid>
            <category><![CDATA[coronavirus]]></category>
            <category><![CDATA[harvard]]></category>
            <category><![CDATA[covid19]]></category>
            <category><![CDATA[surveys]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Wed, 01 Jul 2020 06:07:35 GMT</pubDate>
            <atom:updated>2020-07-02T01:59:12.546Z</atom:updated>
            <content:encoded><![CDATA[<p>By Lucy Li, Melissa Kwan, and Kevin Bi</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*MO7I5Rwhr0YvGaEv" /></figure><p>On March 10th, Harvard College students awoke to a new reality. President Lawrence Bacow announced in an email that students had to leave campus by the end of the week in light of the threat posed by COVID-19. What followed was perhaps the most unusual semester in Harvard history.</p><p>We at the Harvard Open Data Project wanted to investigate how students responded to the pandemic and the disrupted semester. We conducted three surveys over the course of the semester to capture both students’ immediate responses and their retrospective accounts: we sent out the first survey in March, right after students were starting online classes (437 responses). The second survey was conducted in April, as the University announced that all undergraduate courses would be graded with an Emergency SAT/UNSAT system (190 responses). And the final survey took place at the end of the semester, with students looking ahead to an uncertain future (179 responses).</p><p>In this article, we use our survey results to analyze how students reacted to leaving campus and taking classes at home, as well as what they think will happen looking forward. We hope this is able to better inform the Harvard community about how the pandemic has affected students, and to hopefully guide decisions moving forward.</p><h3>Leaving campus</h3><p>With only five days’ notice, undergraduates had to quickly make plans to leave campus. Students had to figure out where they would live, how they would get there, and where they would store their items on campus. The College and the Undergraduate Council announced that they would provide stipends to assist with moving, but many students nonetheless had to pay out of pocket to cover the expenses. The median student spent $250 moving off campus. 19% of students reported no out of pocket expenses from the moveout, while 10% of students reported spending in excess of $1000.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*WdDg4pC6Y-3bXIio" /></figure><p>The vast majority (85%) of students returned to a family home, while only 3% of students reported staying on campus. The remaining students either stayed with a friend (6%), had their own apartment (3%), or had some other living arrangement (2%). Three quarters of students lived with a time-difference of less than 3 hours from East Coast time, with 61% of students living on the East Coast.</p><p>While it is not known how many students were infected with coronavirus, Harvard University Health Services did announce that Harvard affiliates had tested positive for COVID-19 prior to students leaving campus. It is unknown whether those infected were students.</p><blockquote><strong><em>However, nearly 30 percent of students reported experiencing some flu-like symptoms in the weeks after leaving campus.</em></strong></blockquote><p>While we do not know whether or not these were related to COVID-19, it is nonetheless indicative of the potential danger of sending possibly infected students off-campus and across the country.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*3nOg2V8EIDMYW-6j" /><figcaption>Asked in March 2020, two weeks after leaving campus</figcaption></figure><p>At the time, it was difficult to get tested for the virus, and only 2 percent of respondents reported receiving COVID-19 test. However, the uncertainty over exposure on campus or on flights home led many students to self-quarantine after returning home, with 22 percent of respondents isolating even from family members.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*H5KRt9moEQO4Eb2W" /><figcaption>Whether students self-quarantined in the days after returning home</figcaption></figure><h3>Living at home</h3><p>Students returned home under the shadow of the pandemic. Social distancing became the new norm as governments shut down public areas and strangers became potential sources of infection. 44 percent of respondents to our end-of-semester survey reported having had no in-person contact with people outside of their immediate household since leaving campus.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*XaU80P8Jw-C6Scb6" /><figcaption>Asked in June 2020, three months after leaving campus</figcaption></figure><p>Despite the decline in face-to-face interactions, students found ways to keep in touch with friends from school. The vast majority of survey respondents reported contacting their friends from school at least once a week, with 57% indicating that they communicated with their friends every day. However, much of this contact occurred through text messages or group chats, which are an accessible mode of communication but may not replace the richness of in-person interactions. When asked how often they contacted their friends through a medium other than texting/messaging, most respondents indicated that these interactions took place less frequently (once every 2–3 days or once a week).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*_TfekTA79yQfGZa8" /></figure><p>Many students also left behind the comparatively even playing field of a Harvard dorm. 30 percent of respondents reported not having access to a quiet workspace, while 14 percent of students reported a lack of reliable internet access. These figures are also likely underestimated, as students without reliable internet would be unlikely to fill out our online survey.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*TeJ1cGIqxtijVR4R" /></figure><h3>Academics</h3><p>This inequality of working conditions also portended inequality in academics. Immediately after leaving campus, the College extended deadlines to switch classes to pass/fail and to drop classes. With students adjusting to online courses and facing the fear of infection, over 15% reported changing their course enrollment in some way.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*Y56x3-JsVgI_uE9g" /></figure><p>This clearly shows that some students found the prospect of online classes a serious challenge. However, there were concerns that even a voluntary system would be insufficient. Disadvantaged students may have been pressured to maintain their courses or avoid going pass-fail for fear of judgement by admissions committees for graduate schools.</p><p>With rumors swirling about a potential grading change, we asked students in late March what their preferred grading system would be. The most popular was a Universal A/A- system that also received support from the Undergraduate Council at the time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*7GLrOi5tCI3MHuvh" /></figure><p>The University announced on March 27th that Harvard would adopt an Emergency Satisfactory/Unsatisfactory grading system to address potential inequities arising from distinct home environments. Notably, this was the least popular of our three listed grading choices in March.</p><p>We asked how students felt about the new mandatory SAT/UNSAT policy compared to the optional pass/fail option. Students seemed mostly ambivalent about the mandatory SAT/UNSAT grading system compared to optional pass/fail policy, likely because SAT/UNSAT prevents students from raising their GPAs.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*G1myJUBQqbIMN6tZ5AAg2A.png" /></figure><p>We were also curious about how well different departments transitioned to online teaching. We asked students to name their most effectively taught online course and their least effectively taught online course. Out of the most frequently mentioned departments, government classes had the highest ratio of positive to negative mentions and Expository Writing classes had the lowest ratio. The most controversial course in our survey was COMPSCI 51: Abstraction and Design in Computation, which five respondents rated as their most effective course and four respondents rated least effective. Reviews of computer science, economics, and general education courses were mixed, but math earned a largely positive rating, bolstered by MATH 21B: Linear Algebra and Differential Equations.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*vVKv1zwbd0KWKpyo" /></figure><h3>Rating the response</h3><p>While Harvard’s response may have seemed sudden at the time, similar actions were soon taken by universities across the country. We were curious how students judged Harvard’s policies. In both the March and end-of-semester surveys, we asked students to rate Harvard’s response to the pandemic on a 1 to 10 scale. Despite the rushed move out and transition to online classes, students gave Harvard a median rating of 6 in March and a median rating of 7 at the end of the semester.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*WctnupSkX5vCnUuB" /></figure><p>By comparison, students were much less happy with the government’s response to the pandemic in our March survey, which received a median rating of 3.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*cxSX9IRdGjxuJMYf" /></figure><h3>Summer disrupted</h3><p>In addition to disrupting coursework and public health, the pandemic has also had a marked effect on students’ summer and post-grad plans. Many students’ plans became remote or were cancelled outright, and companies implemented hiring freezes that affected students who were recruiting during the outbreak. As seen in the plot below, many students who were planning on working in industry or completing an internship before the outbreak were forced to find other types of work. Students who had planned to do research or enter academia appeared to be less affected, while all respondents who had originally intended to study abroad or travel were forced to change their plans.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*MxZ5GD5iqD68gVFC" /></figure><p>To more precisely capture the effect of the outbreak on students’ summer and post-graduate plans, we asked students to report all the ways in which their plans were impacted.</p><blockquote><strong><em>The majority of respondents (54.9%) reported that their summer or work experiences had become virtual instead of in-person, and a significant portion of respondents (37.8%) said that their plans had been cancelled entirely.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*CGQ2LeCI7FHAFUia" /><figcaption>Note: the percentages in this plot do not add up to 100 because some students’ plans were affected in multiple ways.</figcaption></figure><h3>Looking forward</h3><p>One area of particular interest to students is how the Fall 2020 semester will be handled. Given that the pandemic is unlikely to subside before September, it appears improbable that the fall semester will proceed as normal, and Harvard is expected to announce its plans for the semester in the next week. Many respondents to our June survey were pessimistic about their chances of returning to campus in the fall, with a plurality (31.7%) indicating that they thought this scenario was somewhat unlikely.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*eejkmcHc2K9L0U3A" /></figure><p>In contrast to the perceived unlikelihood of being brought back to campus in the fall, many students appeared to be willing to return if allowed to decide for themselves. The majority of respondents (66%) stated that they were either very comfortable or somewhat comfortable with the idea of returning to campus, with only 23% of students reporting that they were somewhat or very uncomfortable.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*m0Yzs2ABY-bYwUk6" /></figure><p>Consistent with their high levels of comfort with returning to campus, most respondents expressed that their preferred policy for the fall was for students to be fully on-campus. The next most popular option out of those listed in our survey was to cancel/delay the fall semester to the spring, although this option does not <a href="https://www.thecrimson.com/article/2020/4/28/harvard-coronavirus-garber-fall-2020-email/">appear</a> to be on the table for the administration. Less popular were the options of only allowing some students on campus, conducting a fully virtual semester, and starting virtual but migrating to campus during the semester.</p><p>In contrast, when asked which policies they would implement for the fall if they were administrators at Harvard, far fewer students (26%) said that they would bring the whole student body back to campus. However, this option was still one of the most frequently chosen, along with allowing only some students on campus and starting virtual but going back to campus midway through the fall. The differences between respondents’ answers to this question and the previous question indicate that while students largely prefer to be on campus as usual, they understand that allowing all students to return may not be the wisest decision for the administration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*EjkiMYxIuo5F1pIg" /></figure><p>If students are not allowed back on campus in the fall, many are considering taking the semester off. However, because Harvard is unlikely to be able to accommodate large portions of its student body taking leaves of absence and returning at the same time, it is unclear what the potential consequences would be for taking time off. Students may be forced to extend their leaves of absence so that their returns to campus are staggered, or on-campus housing may not be guaranteed upon their return. In addition, an important consideration for many students is whether their friends would join them in taking a gap semester.</p><blockquote><strong><em>Out of our survey respondents, 44% stated that they would take Fall 2020 off if there were no consequences for doing so, and a majority (51%) expressed that they would take the semester off if most of their friends were doing the same.</em></strong></blockquote><p>Far fewer students indicated that they would take a gap semester if the consequences were more severe, or if most of their friends elected to remain enrolled for the fall.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*pAU6ieYbS-0BipFM" /></figure><p>Beyond the Fall semester, the COVID-19 pandemic may also redefine “normal” life for some time to come. Given the pandemic’s far-reaching effects, we wanted to see how long students estimated it would be before the world was able to return to normal. Because students were isolated from each other when making these predictions, this may represent a set of independent estimates that could be used to harness the wisdom of the crowds.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*0bRgYnmAN6LWQJds" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1000/0*iyJRk0-UEDtPHEI0" /></figure><p>In both our March and end-of-semester surveys, we asked students to predict how long it would be before things returned to normal. In March, a plurality of students (46.7%) predicted that things would return to normal within 3–6 months, while at the end of the spring semester, the plurality (38.4%) was more pessimistic in its prediction that it would take another 6–12 months to return to normal.</p><h3>Conclusion</h3><p>It has been around three months since the March survey was conducted and more than a month since the spring semester ended, and states have begun to reopen to a limited extent. However, most aspects of American life are far from normal, and cases are again <a href="https://apnews.com/3a3642cb4b7244665dda34c437c87988">surging</a> in many states. The future course of the pandemic as well as the academic future of students are far from certain.</p><p>What is certain is that students are now facing a changed world, not just for the Spring 2020 semester, but for the foreseeable future. Nonetheless, we are learning to adapt. While a digital experience may never replace an in-person one, we are discovering new ways to learn and to connect, no matter what happens in the fall.</p><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. Check out the code </em><a href="https://github.com/TheKeveloper/HODP-COVID"><em>here</em></a><em>. Editors: Sahana Srinivasan, Asher Noel, and Henry Austin. Cover image by Melissa Kwan.</em></p><p><em>Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join us!</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1b06e6501335" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/harvard-in-the-age-of-covid-19-1b06e6501335">Harvard in the Age of COVID-19</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Do Police Protests Work?]]></title>
            <link>https://medium.com/harvard-open-data-project/do-police-protests-work-718a8f91dffd?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/718a8f91dffd</guid>
            <category><![CDATA[protest]]></category>
            <category><![CDATA[harvard]]></category>
            <category><![CDATA[blacklivesmatter]]></category>
            <category><![CDATA[police-brutality]]></category>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Fri, 12 Jun 2020 23:16:57 GMT</pubDate>
            <atom:updated>2020-06-12T23:16:57.036Z</atom:updated>
            <content:encoded><![CDATA[<p>by Kevin Bi</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/960/1*IEb22Z_7tN0fsqu8sJBBFg.png" /></figure><p>Another unarmed Black man has been <a href="https://www.nytimes.com/2020/05/31/us/george-floyd-investigation.html">killed</a> by the police. His name was George Floyd. The video of a police officer’s brazen cruelty and his utter disregard for the words “I can’t breathe” was shocking, but not new. In March, Breonna Taylor was <a href="https://www.nytimes.com/article/breonna-taylor-police.html">shot</a> by police in the middle of the night after they broke into her apartment. In February, Ahmaud Arbery was chased and <a href="https://www.nytimes.com/article/ahmaud-arbery-shooting-georgia.html">gunned down</a> by two white men. Unfortunately, such tragedies are common: Black people have been killed and abused in this country for centuries.</p><blockquote><strong><em>What set Floyd’s murder apart was the sheer intensity and scale of the response it provoked. As of June 5, more than 450 protests have </em></strong><a href="https://www.nbcnews.com/news/us-news/map-protests-rallies-george-floyd-spread-across-country-n1220976"><strong><em>erupted</em></strong></a><strong><em> across the United States.</em></strong></blockquote><p>Buildings in Minneapolis, Minnesota, where Floyd was killed, have <a href="https://fox8.com/news/live-minneapolis-still-burning-after-police-station-torched-other-buildings-burn-amid-george-floyd-protest/">burned</a> to the ground. Curfews have been <a href="https://www.nbcnews.com/news/us-news/curfews-go-effect-cities-around-country-george-floyd-protests-continue-n1219801">declared</a> across the United States, and massive police responses have been <a href="https://www.voanews.com/usa/national-guard-called-11-states-handle-protests">mobilized</a>. The protests across the country were some of the largest in nearly half a century, but demonstrations against police brutality are nothing new.</p><p>The #BlackLivesMatter movement was officially <a href="https://blacklivesmatter.com/herstory/">launched</a> in 2013 by Black organizers Alicia Garza, Patrisse Cullors, and Opal Tometi when neighborhood watch coordinator George Zimmerman was acquitted of all charges in the killing Trayvon Martin, a Black teen from Florida. BLM also <a href="https://journals.sagepub.com/doi/full/10.1177/1532708616634770">led</a> protests across the country in 2014, after police officers killed Michael Brown and Eric Garner, and has since become a national movement, spurring widespread protests against police brutality and racial inequality in criminal justice over the past seven years.</p><p>We at the <a href="http://hodp.org/">Harvard College Open Data Project</a> were interested in studying the effects of these protests.</p><blockquote><strong><em>While the recent protests have once again raised the issue of police reform, and the Minneapolis City Council even </em></strong><a href="https://kstp.com/news/majority-of-minneapolis-city-council-announces-support-for-dismantling-police-department/5753304/"><strong><em>voted</em></strong></a><strong><em> to abolish the local police department, the question of whether the protests actually lead to a reduction in violence by police remains.</em></strong></blockquote><p>HODP conducts an empirical analysis to answer that question here.</p><h3>The Data</h3><p>We relied on existing datasets to measure both the number of killings by police and the number of protests. The protest data comes from a <a href="https://dataverse.harvard.edu/dataset.xhtml?persistentId=doi:10.7910/DVN/L2GSK6">dataset</a> <a href="https://www.cambridge.org/core/journals/perspectives-on-politics/article/black-lives-matter-evidence-that-policecaused-deaths-predict-protest-activity/BFA2E74F4BCB25C3C222807E8B1111D4">used</a> in a paper by Williamson, Trump, and Einstein that examines which factors contributed to the occurrence of BLM protests in 2014 and 2015. The data includes protests sourced from news articles written at the time, totaling 780 BLM protests during the two-year period.</p><p>Data on killings by police comes from the <a href="https://mappingpoliceviolence.org/">Mapping Police Violence</a> project, which includes data from 2013 to 2019. The data is sourced from media reports of police violence, and includes demographic information on the victim, the location of the killing, and whether the victim was armed.</p><p>Some preliminary analysis of the data gives us a picture of police violence trends over time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*X-ODVOi_ohxqL1jG" /></figure><p>We can also get a sense of the number of killings where the victim is reported to be unarmed.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PjorG8Z9D3e6XTZH" /></figure><p>While it seems like the number of reported unarmed killings per year has declined since 2015, the overall number of killings has remained largely the same. However, these statistics are all on a national scale. In order to really understand the effect of protests, we need to perform a deeper analysis to see how protests affect local police killings.</p><h3>Analysis</h3><p>To gauge the local effect of protests on police killings, we decided to examine them at the city level. Both the protest dataset and the killings dataset labeled the cities in which they occurred. To make sure our analysis was not overly affected by outlier cases (e.g., where a city normally has no killings, but then has two killings in one year), we decided to limit our analysis only to cities that had at least 10 police killings between 2013 and 2019, which left us with 96 cities. We then regressed the change in the number of killings from 2013–15 to 2016–18, both three-year periods, against the number of BLM protests that happened in that city in 2014–15.</p><blockquote><strong><em>We found a statistically significant negative relationship between the number of protests and the change in the number of police killings. In other words, the more protests in the city, the fewer police killings there were in the next three years.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*4DDT3RIEEXi-a8-0" /></figure><p>However, the above analysis is a bit problematic because we do not account for how many police killings were in a city in the first place. This leads to cities that already had a lot of police killings exerting far more influence on the final result than might be desired.</p><p>To account for this issue, we normalized the change in the number of killings by dividing by the total number of police killings in the city between 2013–18. Once again, we find a statistically significant negative relationship. The y-axis can roughly be interpreted as the percentage change in the number of killings between the two three-year time periods. So, we find that on average, each additional protest in a city between 2014–15 reduced the number of police killings in the following three years by about 1%.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NrACsCDwGp8pBJSw" /></figure><p>Of course, the BLM protests are not just about police violence in general, but about the targeting of Black Americans by the police. We performed a similar analysis on the number of Black individuals killed by police. In this case however, the number of cities analyzed was smaller because we restricted once again to the cities where at least 10 Black individuals were killed by police between 2013 and 2019, and there were only 32 cities where this was the case.</p><blockquote><strong><em>We once again find a statistically significant negative relationship between the number of BLM protests and change in the number of Black people killed by police.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Ctz6LkLuWpmvb4PZ" /></figure><p>And we find the same negative relationship after normalizing for the total number of Black people killed by police during the time period.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NvA__FZOEd85Ltu_" /></figure><p>We also considered that maybe what matters is not how many protests there are, but whether there were any<em> </em>protests in a city. That is, it may be that each additional protest does little, but just having any protests at all is what matters. We compared the average change in the number of total killings in cities that had at least one protest against cities that had no protests. While there was a large difference in average change, the difference was not statistically significant at the 5% level.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*43WMk0adm-Da2NFq" /></figure><p>An analysis with the normalized changes in each city shows similar but slightly more statistically significant results. This finding seems to suggest that it is not only the existence of at least one protest that matters, but that each additional protest continues to make a difference in the number of people killed by police for a given city.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*DeejU-u7MszM5v8I" /></figure><blockquote><strong><em>This offers compelling evidence that BLM protests decrease police violence against Black individuals as well as the population at large. Per our analysis, every additional protest decreases police killings by approximately 1% on average in the city where it is held.</em></strong></blockquote><p>It is important to note, however, that our analysis does not tell us the mechanism leading to this reduction in violence. It could be due to institutional police reforms in response to the protests, or it could be that police are simply less willing to use lethal force for fear of prompting more protests. In either case, though, this result is a promising sign that today’s protests can make a difference.</p><h3>A Few Words of Caution</h3><p>While we believe our analysis offers some fairly solid evidence for the effectiveness of protests, there are a few reasons for caution.</p><p><strong>Mean Reversion</strong></p><p>One issue that we were concerned about in this analysis was whether the decrease detected from 2016–18 was due to an unusually high number of killings by police in 2013–2015, which would have prompted protests, and that the subsequent reduction in police killings was simply a reversion to the mean. We attempted to correct for this by examining a three-year window rather than a single year, but it would have been preferable to compare to a longer-term trend prior to 2013. Unfortunately, we were unable to obtain such data.</p><p><strong>Outliers</strong></p><p>New York City had 43 BLM protests in 2014–15 and also had a sharp decrease in the number of police killings. While we do believe that the New York data point contributes nontrivially to the strength of our results, we found that the same negative correlation held after omitting the New York data point, albeit with less statistical significance.</p><p><strong>Unreliability of Data</strong></p><p>Unfortunately, data on both protests and police killings can be highly flawed. Both of our data sources compiled the killings and protests based on news articles. In particular, this is because previous self-reports of killings by police departments have <a href="https://www.washingtonpost.com/news/post-nation/wp/2014/09/08/how-many-police-shootings-a-year-no-one-knows/">proven</a> unreliable. However, we do not believe that there is any systematic bias in the data that would weaken our results.</p><p><strong>Correlation vs. Causation</strong></p><p>This analysis is correlational, which of course does not necessarily imply causality. While reverse causality seems unlikely in this case (future decreases in police killings would not increase the number of protests), there is possibly some omitted, confounding variable. For example, it could be that communities that protest more are more civically engaged, and therefore also more capable of pushing for reductions in police violence.</p><p>To control for the possible confounder of civic engagement, we ran an additional regression controlling for the 2012 presidential turnout of the state (turnout data was unavailable at the city level) using data from the <a href="http://www.electproject.org/2012g">United States Election Project</a>. We used 2012 because it predates the BLM protests, enabling us to capture the preexisting civic engagement. Not only did we obtain largely the same results for both normalized and unnormalized regressions, but they became even more<em> </em>statistically significant.</p><p>Below, we have plotted the voter turnout residual regression of killings on protests. The x-axis is the difference between the actual number of protests and the expected number of protests based on voter turnout, and the y-axis shows the difference between the actual change in killings for a city and the expected change given voter turnout.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*9bsOMcG2kHVWbGih" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*jVTBCSKPp_p26rFx" /></figure><p>It is important to note, however, that this finding does not rule out all other possible confounding variables, and so some caution should be exercised in causal claims of protests decreasing police violence.</p><h3>Related Work</h3><p>While we are unaware of any previous work that has examined the effectiveness of the BLM protests in reducing police violence, there has been notable work on BLM and protests generally. <a href="https://www.cambridge.org/core/journals/perspectives-on-politics/article/black-lives-matter-evidence-that-policecaused-deaths-predict-protest-activity/BFA2E74F4BCB25C3C222807E8B1111D4">Williamson, Trump, and Einstein</a>, from which we obtained our protest data, finds that BLM protests are more likely to occur in areas with a history of police violence against the Black community.</p><p>On the issue of the effectiveness of protests, <a href="https://dash.harvard.edu/bitstream/handle/1/13457753/TeaParty_Protests.pdf">Madestam et al.</a> find that protests by the Tea Party strengthened Tea Party support among the general public, increased the number of votes received by Tea Party candidates, and led to incumbent legislators voting more conservatively. In other words, protests can and do affect legislation.</p><p>Finally, one potential area of concern comes from <a href="https://scholarship.law.upenn.edu/cgi/viewcontent.cgi?article=3111&amp;context=faculty_scholarship">Ba and Rivera</a>, which finds that police may also become less active in response to civilian complaints leading to a rise in crime rates. This result may indicate that the reduction in killings by police following protests is not a result of reform but rather of police no longer responding to crime reports as vigorously. We were unable to quantify police reforms in order to correlate with protests, but that may be a viable area of research if the data is made available.</p><h3>Why it Matters for Harvard</h3><p>Most HODP analyses include a Harvard-specific dataset, but unfortunately, we were unable to develop one for this article. We reached out to the Harvard University Police Department for data on the number of complaints against officers, but we were denied access. We will continue trying to obtain data from the administration and searching for ways to analyze police, racism, and protests at Harvard through publicly available data.</p><p>While there is no Harvard-specific application for this analysis, we believe that this article is important for everyone, including the Harvard community. Numerous Harvard organizations, including <a href="https://docs.google.com/document/d/1Dtf8LlJC-nzpKNsS4D0dW0bb_LAtZqfYah_PSJMZHwE/edit">HODP</a>, have <a href="https://www.thecrimson.com/article/2020/6/1/college-reactions-police-brutality/">taken action</a> to support and raise money for the nationwide movement against racial inequality, racism, and police violence. Many students are protesting, raising awareness, and donating to the cause. We hope that this analysis serves as a reminder that actions can and do create change in our communities.</p><p><em>The author extends his thanks to Sahana Srinivasan, Leo Saenger, Seth Billiau, and Lucy Li for their helpful feedback and suggestions on this piece. The code and data for this analysis can be found here: </em><a href="https://github.com/TheKeveloper/Police-Protests"><em>https://github.com/TheKeveloper/Police-Protests</em></a><strong><em>.</em></strong></p><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join </em><strong><em>us</em></strong><em>!</em></a></p><p><em>This article was co-published and co-edited with the </em><a href="https://harvardpolitics.com/protests-for-racial-justice/do-police-protests-work/"><em>Harvard Political Review</em></a><em>.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=718a8f91dffd" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/do-police-protests-work-718a8f91dffd">Do Police Protests Work?</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[HODP x OCS Career Fairs]]></title>
            <link>https://medium.com/harvard-open-data-project/hodp-x-ocs-career-fairs-7e27bf8ea5c6?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/7e27bf8ea5c6</guid>
            <category><![CDATA[internships]]></category>
            <category><![CDATA[career-fair]]></category>
            <category><![CDATA[data-journalism]]></category>
            <category><![CDATA[harvard]]></category>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Thu, 28 May 2020 05:58:46 GMT</pubDate>
            <atom:updated>2020-05-28T05:58:46.668Z</atom:updated>
            <content:encoded><![CDATA[<h3>HODP x OCS Career Fairs: Fun for Freshmen or Solely Seniors?</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XOIic_DiQSdKivpE7co16g.png" /></figure><p>by Daniela Shuman, Moritz Pail, and Robin Robinson</p><h3>Introduction</h3><p>You’re a sophomore interested in consulting. Is it worth taking a couple hours out of your busy day to travel all the way to the SOCH and network with consulting firms at the OCS Career Fair? Judging by the data from the 2019 Consulting Career Fair, it doesn’t look like you should go.</p><p>What else does the data tell us? We analyzed the data from 10 of Harvard’s Career Fairs: Biotech, Business and Tech, Consulting, Data and Tech, Diversity Opportunities, Engineering, Fintech, Media and Marketing, Non-profit, and Start-ups. The Office of Career Services website offers data on which companies attended the career fairs and to which class years they offered opportunities. We weren’t able to analyze all career fairs Harvard’s hosted due to a lack of data. One caveat to this analysis is that there is overlap between career fairs, since one opportunity could have been presented at multiple fairs.</p><p>Here is what we found out about summer opportunities for Harvard undergraduate students.</p><h3>Analysis</h3><p>Freshmen are likely to find opportunities available in the engineering/tech fields. The Business and Tech Fair offered a whopping 23 opportunities for freshmen with Engineering closely behind at 18. Data and Tech companies similarly offered 16 opportunities.</p><blockquote><strong><em>It is important to note that these tech career fairs generally have more companies, and yet they have proportionally more opportunities for freshmen.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/533/0*dY-7Mvsey9-fAk6x" /><figcaption>Figure 1: The number of opportunities open to freshmen per career fair in 2019.</figcaption></figure><p>The Non-profit Fair, however, led in the opportunities offered for freshman with 25 opportunities and Start-ups in a close 2nd with 23 opportunities. We reason that start-ups do not have a preference for what classes were taken, nor do they have nearly as strong of a goal in recruitment. Rather, their priorities lie in getting work done cheaply and quickly. This is why any intern is valuable for a start-up and likely why they offer so many opportunities for younger students.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/724/0*A6geyEd6oOUeOgpa" /><figcaption>Figure 2: The number of opportunities open to sophomores per career fair in 2019.</figcaption></figure><p>Our data suggests a similar pattern for sophomores. However, there are generally more opportunities at each fair than for freshmen. With only 3 internship opportunities, Consulting again remains in last place. The Biotech, Fintech, and Media and Marketing Fairs are also on the lower end of the ranking. Data and Tech, Diversity Opportunities, and Engineering are somewhere in the middle with 23, 28, and 25 opportunities respectively. Just like for freshmen, the most opportunities for sophomores were from the Business and Tech, Non-profit, and Start-ups Fairs. This indicates that freshmen and sophomores are seen similar in the eyes of employers.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/724/0*pl9VeZOPr__3AokK" /><figcaption>Figure 3: The number of opportunities open to juniors per career fair in 2019.</figcaption></figure><p>For juniors we see a major increase across the board in all industries.</p><blockquote><strong><em>Juniors are especially valuable for recruitment, along with seniors, not only because of their experience, but also because they are more likely to join the company after graduation.</em></strong></blockquote><p>We see this reflected in the huge jump in companies that offer opportunities in consulting especially, considering how scarce opportunities were for underclassmen. However, there wasn’t a big increase in Media and Marketing.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/724/0*DF3oBATf-nXs1NEY" /><figcaption>Figure 4: The number of opportunities open to seniors per career fair in 2019.</figcaption></figure><p>Seniors had a relatively similar pattern to juniors. Every industry offered all their opportunities to seniors, as we expected. Although it is important to note, as mentioned above, that business and tech career fairs offer many more opportunities total, which skews the proportions for the underclassmen.</p><p>Here are the final proportions for your reference:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Cewk0_yZAusGtY3793Td9A.png" /><figcaption>Figure 5: The number of career fair opportunities for each class in 2019.</figcaption></figure><h3>Conclusion</h3><p>The purpose of this project was to give students some insight on how helpful career fairs would be for their particular industry and year.</p><blockquote><strong><em>We noticed that tech and engineering career fairs (Business and Tech, Data and Tech, and Engineering) offered more opportunities in general to Harvard students.</em></strong></blockquote><p>And because of this underclassmen have some benefit if they are looking for tech opportunities. In other fields, underclassmen aren’t provided with nearly as many opportunities. If you’re a freshman, it is definitely helpful to go to career fairs for software jobs, particularly if you’re interested in startups, since they seem to be more open to younger students and you’re less likely to hear of them if you don’t attend. But maybe look elsewhere for consulting or fintech opportunities.</p><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. Check out the data we used </em><a href="https://drive.google.com/file/d/1aH4Bl9B3YXxDu0URz_TgULvzyAtSmt4W/view?usp=sharing"><em>here</em></a><em>. Article was edited by Sahana and Terry. Cover image by Melissa.</em></p><p><em>Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join </em><strong><em>us</em></strong><em>!</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=7e27bf8ea5c6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/hodp-x-ocs-career-fairs-7e27bf8ea5c6">HODP x OCS Career Fairs</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Harvard University: Hefty Tuition with Healthy Financial Aid?]]></title>
            <link>https://medium.com/harvard-open-data-project/harvard-university-hefty-tuition-with-healthy-financial-aid-2f3482a1c2ee?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/2f3482a1c2ee</guid>
            <category><![CDATA[college-tuition]]></category>
            <category><![CDATA[harvard]]></category>
            <category><![CDATA[college-admissions]]></category>
            <category><![CDATA[tuition]]></category>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Sun, 24 May 2020 00:32:41 GMT</pubDate>
            <atom:updated>2020-05-25T21:42:29.740Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hPKQsvdTdFnyAn_pzCluEw.png" /></figure><p>by Anna Li and CC Darms</p><h3>Introduction</h3><p>People, including HODP, often <a href="https://medium.com/harvard-open-data-project/of-varying-value-harvards-tuition-over-time-cd7711184589">scrutinize tuition</a> at top universities. But it’s important to dovetail any discussions of tuition with analysis of financial aid to see to see if Harvard really lives up to its promise to “bring the best students to Harvard, <a href="https://college.harvard.edu/financial-aid">regardless of their ability to pay</a>.”</p><p>Harvard’s tuition ranks as one of the highest in the U.S. The 2019–2020 <a href="https://handbook.fas.harvard.edu/book/tuition-and-fees">Harvard College tuition</a> was $47,730. Tuition, room, board, and fees combined were $68,401.</p><p>However, fifty-five percent of full-time undergraduates receive some kind of need-based financial aid, and the average need-based scholarship or grant award is $55,678.</p><p>Given these contrasting realities, we wanted to further investigate how Harvard’s financial aid actually works. In particular, we took a look at which schools within Harvard had the most financial aid, where exactly this aid was coming from, and how aid per student changed over time.</p><h3>Gathering our Data</h3><p>We used the <a href="http://hodp.org/catalog/">HODP catalog</a>, which contains data on <a href="https://oir.harvard.edu/fact-book/degree-student-financial-assistance">Degree Student Financial Assistance</a> from the Office of Institutional Research. We used a PDF converter to transfer the data to Excel, and we used Jupyter Notebook and Excel to create our visualizations.</p><h3>Aid by School</h3><p>First, we broke down 2018 financial aid data by the school that awarded it, among the College and Harvard’s graduate schools. As pictured in Figure 1, the Graduate School of Arts and Sciences (GSAS) receives the greatest amount of total financial aid, followed closely by Harvard College, and then the Law School. The first graph depicts the total sum of money received by all students, broken down into loans, grants, and job income.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/0*CeofqzlS77cShY2V" /><figcaption>Figure 1: Total amount of financial assistance by type and by Harvard school.</figcaption></figure><p>Figure 2 uses the same data but depicts the proportion of each type of financial assistance by school. We can see that for most of the schools, almost all of the financial aid comes in the form of loans and grants, with few exceptions for jobs, mostly among GSAS students. Financial aid eligibility is determined by the information provided in the student’s Free Application for Federal Student Aid (FAFSA). Grant money never has to be repaid, loans can be either student or parental, and job financial aid comes in the form of student employment by Harvard.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/720/0*1FiAgizP1hpKdOjd" /><figcaption>Figure 1: Proportions of financial assistance by type and by Harvard school.</figcaption></figure><p>Something that stood out to us in these graphs is the amount of financial aid coming from the Law School in the form of loans. This is particularly interesting because the enrollment isn’t substantially higher in comparison to the other larger graduate schools, and neither is the cost of attendance, yet the amount of loans is substantially greater. The Dental and Extension Schools also have high proportions of loans but lower total loan amounts.</p><p>The Business School has the next-largest amount of loans. Job financial aid seems to be a relatively small portion of financial aid aside from at GSAS, where job income accounts for 20% of student assistance. In summary, grants and loans play a very large role in Harvard’s financial aid, with the former being very significant in the College and GSAS and the latter playing a substantial role in all the graduate schools other than GSAS.</p><h3>Types of Grants</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/0*8HW1prp-r-90iIS4" /><figcaption>Figure 3: Sources of aid by school.</figcaption></figure><p>In each of the three categories of jobs, loans, and grants, aid can be broken down further into three sources: Institution, Federal, and Outside. Institutional financial aid comes directly from Harvard, and federal aid may come in the form of a Federal Pell Grant of a Supplemental Educational Opportunity Grant (SEOG). Outside financial aid may include scholarship funds from sources such as secondary schools, civic organizations, students’ parental employers, corporations, the National Merit Scholarship Programs, the G.I. Bill, and the ROTC. Figure 3 above shows the sizes of various sources of grant aid by school.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/0*PPUHFIMmFvadiRXL" /><figcaption>Figure 4: Sizes and sources of loans by Harvard school.</figcaption></figure><p>Figure 4 above shows us the sizes and sources of loans specifically. We can see that most loans to Harvard students are federally funded, and that a plurality are given to Law School students. Business school students receive more outside loans than do most other students. Finally, as seen in Figure 5 below, job financial aid seems to be the most spread out between the three sources of institution, federal, and outside job financial aid is most relevant in GSAS and the College.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/0*Z2ixRs-Z8zxuko_B" /><figcaption>Figure 5: Sources of job financial assistance by school.</figcaption></figure><h3>Trends Over Time</h3><p>The Harvard OIR keeps public records of financial aid by school going back to 2009, so we visualized their historical data to look for trends over time. It is important to recognize that this data is not inflation-adjusted. That being said, Figure 6 tracks changes to the total sum of financial aid given to each school. We then found historical data for enrollments and tuition (not including room and board) at Harvard’s various schools.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*sR1TWdm0jOHFmR6h" /><figcaption>Figure 6: Total financial aid over time by school (not inflation-adjusted).</figcaption></figure><p>Figure 7 depicts changes over time to the average financial aid as a percentage of total tuition. This graph gives a better sense of the “real financial aid” being given by comparing it to the cost of attendance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*b-UY1w3h7jVe9obH" /><figcaption>Figure 7: Financial aid as a percentage of total tuition over time, by school.</figcaption></figure><p>Notably, some of the percentages are larger than 100% because financial aid also contributes to room and board and other fees.</p><blockquote><strong><em>We saw that, while total financial aid increased for all schools, and increased most significantly for GSAS, aid as a fraction of total tuition decreased for all schools in the past ten years.</em></strong></blockquote><p>The Crimson reported that GSAS’s professional stipends have <a href="https://www.thecrimson.com/article/2016/3/22/gsas-phd-stipend-increase/">increased by 3% every year for the last six years</a>, which puts Harvard on par with peer institutions for financial aid to grad students. However, our data suggests that, in comparison with tuition, GSAS aid fell by 8% from 2009 to 2017, as rents in Cambridge rose. Meanwhile, the decrease was least severe at the College; it fell precipitously (by more than 30%) at some of the smaller professional schools, including Divinity, Design, and Public Health.</p><h3>Concluding Remarks</h3><p>Our goal was to explore Harvard’s financial aid data to find interesting trends. We were interested in the breakdown of financial aid into grants, loans, and jobs, and we also looked at the further breakdown into institution, federal, and outside.</p><blockquote><strong><em>We found that out of the Harvard schools, GSAS receives the greatest amount of total financial aid, mostly job-based financial assistance, followed by the College and the Law School.</em></strong></blockquote><p>In addition, GSAS and law school students receive the most financial assistance per student. Something that we’d like to explore further is why such a large proportion of financial aid for the Law School comes from loans.</p><p>We also observed that most grant financial aid comes from the institution, a large majority of loan financial aid is federal, and job financial aid is a much more heterogeneous mix of institution, federal, and outside sources.</p><p>Next, we looked at how financial aid data evolves over time. An important key takeaway is that many of the total sums of financial aid given to each school is increasing, while the aid as a percentage of total tuition is decreasing across almost all the schools. We wanted to take a look at this data because we found it personally relevant, particularly in light of last semester’s graduate student union strike, and because we wanted to see how much Harvard delivers on its promises of financial aid.</p><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. Check out the data we used </em><a href="https://oir.harvard.edu/fact-book/degree-student-financial-assistance"><em>here</em></a><em>. Article was edited by Sahana and Terry. Cover image by Melissa.</em></p><p><em>Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join </em><strong><em>us</em></strong><em>!</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=2f3482a1c2ee" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/harvard-university-hefty-tuition-with-healthy-financial-aid-2f3482a1c2ee">Harvard University: Hefty Tuition with Healthy Financial Aid?</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Of Varying Value: Harvard’s Tuition over Time]]></title>
            <link>https://medium.com/harvard-open-data-project/of-varying-value-harvards-tuition-over-time-cd7711184589?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/cd7711184589</guid>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Mon, 20 Apr 2020 00:12:55 GMT</pubDate>
            <atom:updated>2020-04-20T00:12:55.202Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DsQRpV7V63pj8JvyDEU6pg.png" /></figure><p>by Benji Kan and Richard Luo</p><h3>Introduction</h3><p>It’s no secret that attending private post-secondary institutions like Harvard is <a href="https://www.nytimes.com/2015/09/13/magazine/is-college-tuition-too-high.html">expensive</a>. For the 2019–2020 academic year, Harvard College’s <a href="https://handbook.fas.harvard.edu/book/tuition-and-fees">total cost of attendance, not including optional fees, was a hefty $69,607</a>. (Assuming an average of 12.5 hours spent in class per week, tuition averages out to $159.10 <em>per hour</em> of class). In this article, we present the results of our analysis of Harvard’s tuition over time.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*aYo2CohjR9RjSi1B" /></figure><h3>Why Tuition Matters</h3><p>(For this analysis, we consider tuition as the total cost of attendance. In a general sense, one might consider Harvard as an education inclusive of experiences outside of the classroom. Then, it makes most sense to analyze the data on what we pay for the <strong>entire </strong>experience.)</p><p>There are a number of ways we can understand what the tuition is. Primarily, we think of education as a service: students choose to go on to higher education to learn skills and acquire knowledge that will be useful in their future professional or personal lives. Then, the tuition is simply how much that education is worth to students.</p><blockquote><strong><em>Or, at least, tuition is what the University thinks their education is worth: the students have to decide whether or not they agree.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/518/0*6hYE4kBHuyMmjjE0" /><figcaption>Breakdown of Harvard University’s Revenue Sources (<a href="https://finance.harvard.edu/files/fad/files/fy19_harvard_financial_report.pdf">FY19 Harvard Financial Report</a>)</figcaption></figure><p>Tuition is also a significant source of the University’s revenue.</p><p>Here, we can see that the total tuition paid by students makes up a solid 22% of Harvard University’s annual income. Changing the cost to students can significantly alter the total revenue for the University.</p><p>However, tuition can also be a significant barrier to entry. Yes, financial aid exists, and <a href="https://college.harvard.edu/financial-aid">55% of Harvard College students receive some form of need-based financial aid</a>, but the sticker price is still what many students have to pay in some amount to attend. Due to high tuition, there are many qualified students who simply can’t afford a Harvard education, which we will discuss in more detail later.</p><blockquote><strong><em>This article will show that Harvard tuition is high, and it’s only increasing from here on out.</em></strong></blockquote><p>These are two seemingly obvious results, but they beg some more interesting questions: why does tuition continue to increase every year? Why is the cost so high to begin with? Which components of the tuition increase the most, and what are the causes of these increases? Does the value of the Harvard experience actually improve at a rate commensurate with the increase in costs?</p><h3>Data Collection</h3><p>We analyzed Harvard College’s annual Cost of Attendance over the 35-year period from 1985 to 2020. We acquired the bulk of data (1985–2016) from the Harvard Open Data Project’s own visualization of <a href="https://hodp.org/visual/?id=tuition2018">Harvard College Tuition, Fees, Room, and Board</a>, and supplemented the remaining years to complete our analysis from <a href="https://oir.harvard.edu/common-data-set">Harvard’s Office of Institutional Research’s published Common Data Sets</a>.</p><h3>Results</h3><p>We divide our results in two parts: overall trends in the total cost over the time period, and trends in specific subcomponents of the total cost.</p><h3>Trends in Overall Costs</h3><p>First, we provide a visualization of the increase in total cost over time:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*rHfFkccW7prDhfB3" /></figure><p>We see that not only does the tuition rise each year, but it actually rises at an increasing rate over time. The average percentage increase in total cost over the past ten years has been 3.6%.</p><h3>Inflation</h3><p>However, the raw increase in cost does not tell the full story, as the value of the dollar changes and fluctuates across time. Thus, using year-over-year inflation data acquired from the <a href="https://www.bls.gov/cpi/">U.S. Bureau of Labor Statistics</a>, we compared the percentage growth in the cost to the year-over-year inflation:</p><p>With one exception in the year 2000, the growth of the total cost has <strong>always </strong>exceeded the inflation rate. Indeed, the average difference between how much the growth in tuition has exceeded inflation over the past ten years has been 1.9%. In other words, the <strong>real</strong> cost of attendance, adjusted for inflation, has increased by an average of 1.9% over the past decade.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*sJbL7m4iy04MLQgO" /></figure><p>Even after adjusting the value of tuition to 2019 dollars, the total costs appear to be consistently rising over time. Over the last 35 years, the average real income of all but the top quintile of US households has grown relatively little.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/910/0*qD0rInxs-GLd5A84" /></figure><p>For relatively constant incomes and increasing costs of education, the proportion of a household’s income spent on education increases. Then, it is interesting to consider: does this mean that Harvard considers the actual value of its education to be improving each year? Do Harvard students and faculty feel like the college experience improves this much year after year?</p><h3>Median Income</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/528/0*u7XFnu_hCgC_NEMB" /></figure><p>According to data collected on annual salaries, the cost of attendance only began to exceed the median income of workers in the U.S. around the year 2010. Interestingly enough, Harvard College <a href="https://news.harvard.edu/gazette/story/2011/09/harvards-record-166-million-financial-aid-program-will-increase-aid-to-low-income-students-and-provide-a-new-financial-aid-calculator-for-students-and-families/">expanded its investment in undergraduate financial aid by $10 million in 2011</a>, providing students with $166 million in need-based scholarships. Beginning in fall 2012, financial aid was also increased for low-income students by raising the income level under which parents pay nothing towards the cost of attendance to $65,000.</p><blockquote><strong><em>So this begs the question: what will happen if the median income of U.S. workers drops again sometime in the near future, further increasing the gap between the cost of attendance and how much families can afford?</em></strong></blockquote><h3>Trends in Specific Costs</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/416/0*Rs2z0WW6BywRcmT8" /></figure><p>Here, we have plotted data on additional student fees, which include the annual student activity fee and health fee. There’s been a steady rise in the amount each student pays in additional fees, except for significant drops in 2008–09 and 2012–13; we believe these outliers may have been due to changes to the national health care policy under President Obama and his administration.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/391/0*F17oGJx2R5xraT86" /></figure><p>As a percentage of the total cost of attendance, however, these additional fees have continued to steadily decrease over the last 35 years. Between the years 2015 and 2019, this percentage has sat relatively close to 6%.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/441/0*HTAr_v6HobG7mpCR" /></figure><p>There’s also been a steady increase in base tuition since the year 1985 (not including room, board, additional fees and other personal expenses).</p><blockquote><strong><em>As the data reveals, even inflation alone is not sufficient to justify this base tuition climbing upward year after year, which suggests that Harvard is indeed getting more and more expensive.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/396/0*Fkn3-nOuzldg_8o3" /></figure><p>When looking at base tuition as a percentage of the total cost of attendance, we observe that it’s stayed between 65 to 70%. Though there was a noticeable increase in this percentage coming into the 21st century, it has been going back down since the year 2006 and is currently at about 68.5%.</p><h3>Future Analyses</h3><p>While the current analysis has generated some interesting results and points of discussion, there are still points left to be analyzed, among which we’d recommend:</p><p>Analyze costs for <strong>other divisions of Harvard University</strong>. Harvard is not limited to the College for undergraduates, but has a total of <a href="https://www.harvard.edu/schools">twelve degree-granting schools</a>. How does tuition for Harvard College compare to that of other arms of the University?</p><p>Consider the cost of <strong>room and board in the context of the housing market</strong>. <a href="https://www.forbes.com/sites/stephenmcbride1/2019/11/20/how-to-make-a-killing-during-the-biggest-housing-boom-in-history/">The US real estate market has seen abrupt and significant contractions and expansions in recent history</a>, and it could be interesting to see what the data says about Harvard’s price of room and board compared to fluctuations in the larger market.</p><p>Compare Harvard’s tuition over time to <strong>admissions data</strong>. There may be interesting trends in the relation between Harvard College’s rising costs and <a href="https://qz.com/1584304/acceptances-rates-at-top-us-colleges-dropped-further-for-the-class-of-2023/">dropping admission rates</a> over time.</p><p>Similarly, we could compare the costs of Harvard College <strong>with peer institutions</strong>, such as the other schools in the Ivy League as well as post-secondary institutions in the Boston/New England area. It may be interesting to note differing changes in tuition over time, and this analysis would be useful if combined with a concurrent investigation of how much financial aid is being awarded to students at these other schools.</p><h3>Conclusion</h3><p>Harvard College is expensive, even after we’ve adjusted for annual inflation rates, and especially compared to the median U.S. household income. Of course, Harvard’s tuition is also responsive to local and domestic events, from policy decisions at the University level to larger societal currents happening on a global scale. Finally, since the College’s real costs are consistently increasing over time relative to the economy, Harvard should carefully consider the growing difference between its tuition increases and the incomes of most American families.</p><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. Check out the data we used </em><a href="https://drive.google.com/file/d/1DO9BygyrlwIEeuHZPIJ5Xj8ixGpb5Go4/view?usp=sharing"><em>here</em></a><em>.</em></p><p><em>Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join </em><strong><em>us</em></strong><em>!</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=cd7711184589" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/of-varying-value-harvards-tuition-over-time-cd7711184589">Of Varying Value: Harvard’s Tuition over Time</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Mapping The Hometowns of Harvard’s Student-Athletes]]></title>
            <link>https://medium.com/harvard-open-data-project/mapping-the-hometowns-of-harvards-student-athletes-daf82fea501f?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/daf82fea501f</guid>
            <category><![CDATA[student-athlete]]></category>
            <category><![CDATA[harvard]]></category>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Mon, 13 Apr 2020 02:35:10 GMT</pubDate>
            <atom:updated>2020-04-13T02:36:35.109Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*OqsD9IKxnYSBeqtk08KXuA.png" /></figure><p>by Alison Hu, Jordan Lee, and Richard Zhu</p><h3>Introduction</h3><p>While Harvard is more commonly known for its academic rigor, the university’s athletic program is remarkably impressive as well. Harvard offers a Division-I leading 42 intercollegiate sports teams who have won hundreds of <a href="https://college.harvard.edu/life-at-harvard/athletics">Ivy League and NCAA titles</a>. Roughly 20% of Harvard’s Class of 2019 played a varsity sport, with 12% coming to Harvard as recruited athletes according to an annual <a href="https://features.thecrimson.com/2015/freshman-survey/makeup/">Crimson survey</a>. And with the Harvard admissions trial offering new data and information to the public, there has been a recent increase in <a href="https://www.theatlantic.com/ideas/archive/2019/10/harvard-university-and-scandal-sports-recruitment/599248/">national media scrutiny</a> on Harvard’s athletic recruitment and its impact on admissions.</p><p>Given the integral role of athletics on campus life, we wanted to investigate Harvard’s sports teams in terms of where athletes are from. Our main question was:</p><blockquote><strong><em>Where do Harvard athletes come from (both in terms of their hometowns and high schools)?</em></strong></blockquote><p>How does this differ when isolating factors of different sports, gender and class?<strong> </strong>Beyond geography, we also wanted to see if there were significant differences in the average wealth of athlete hometowns controlling for different factors.</p><h3>Data and Methods</h3><p>The data used for this analysis was collected from the <a href="https://gocrimson.com/">Harvard Athletics</a> website, where each athlete’s hometown and high school are published. Because the hometowns were not formatted uniformly across all athletes, e.g. Massachusetts abbreviated as MA or Mass., we cleaned the data using the <a href="https://pypi.org/project/geopy/">GeoPy package</a> in Python to ensure that all of the locations were in the same format. The cleaned data then allowed us to plot the locations on a map using QGIS. In addition, we used the hometowns from the cleaned data set to collect median household income data of each town from <a href="https://datausa.io/">Data USA</a>. One important note is that we excluded athletes who listed international hometowns — our aggregation only included athletes who listed hometowns in the United States in line with our Data USA source.</p><p>Using our clean data with each athlete’s name, sport, hometown, and high school, we used QGIS to plot our geolocation data onto a US map.</p><h3>Results</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qnIyc_hsKgaYC7AI_f7wfg.png" /><figcaption>Figure 1: A map of student-athletes’ hometowns, with the number of athletes from each town corresponding to the numbers in the legend.</figcaption></figure><h3>Geography</h3><p>The above graph plots the hometowns of each athlete, with the color of the dots indicating the number of athletes from that location. As can be seen, Harvard recruits athletes from all over the United States — almost if not every state in the US is represented by Harvard’s student-athletes.</p><blockquote><em>However, there is a clear difference in density between the two halves of the country — </em><strong><em>predictably, many more athletes come from the eastern half of the United States, with a stark difference in density between the east and west sides of the map.</em></strong></blockquote><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wrF-rs7qi3gVAVWo1DwuLw.png" /><figcaption>Figure 2: States with the Most Harvard Athletes. As can be seen, states with larger populations and states closer to the East Coast display higher levels of athletes.</figcaption></figure><p>Figure 2 corroborates this spatial data; with the exception of California at the top of the list, states closer to the East Coast tend to have higher numbers of athletes. One explanation for this contrast is targeted recruitment: because Harvard and other schools in the eastern US compete with schools in the western US to recruit athletes, Harvard may either a) initially target more East Coast high school athletes due to their proximity, allowing for an easier scouting and recruitment process or b) have a higher yield for athletes who come from the East Coast in the first place.</p><h3>Median Hometown Household Income</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*-W4HK0nYdRO6nvA3" /><figcaption>Figure 3: Median household income of U.S.-based student-athletes from various Harvard teams.</figcaption></figure><p>The median household income by sport was calculated by taking the median of the median incomes of the hometowns of all U.S. athletes in each sport. Sports with both male and female teams generally display similar median household incomes. In particular, Men’s and Women’s Skiing represent two of the three lowest incomes, Men’s and Women’s Track and Field are both in the lower 25th percentile, Men’s and Women’s Water Polo both have the same median income, and Men’s and Women’s Golf are both in the top 25th percentile.</p><blockquote><strong><em>In general, it seems that athletes from sports involving minimal equipment, like Track and Field and Cross Country, come from hometowns with lower median household incomes, whereas sports that require more specialized facilities, training, and equipment, such as Lacrosse, Crew, Golf, and Fencing, come from wealthier hometowns.</em></strong></blockquote><p>However, it’s important to note that this data is limited. First, it only features athletes from U.S. hometowns — for example, the Women’s Squash team has only four U.S. athletes while the Men’s Squash team has six U.S. athletes, and thus the median household incomes for those teams may not be representative given the small sample size.</p><p>In addition, it’s possible that the median household income of each individuals’ hometown doesn’t necessarily indicate the athlete’s household wealth. Information regarding athletes’ family incomes compared to the overall Harvard student body can be found from the <a href="https://features.thecrimson.com/2015/freshman-survey/makeup/">Crimson’s annual survey.</a> However, the average wealth of athletes’ hometowns nonetheless gives a good indicator of whether certain sports have athletes from wealthier areas.</p><h3>Conclusion</h3><p>In conclusion, we sought to find out where Harvard’s athletes come from around the country and world, and whether there were differences in their hometowns by sport or other factors. We found that in general, Harvard’s athletes come from all over the United States, although much more come from the eastern half (and specifically, the Northeast US) than from the western US. In addition, we found that there were noticeable differences in the average household income of hometowns between different sports. Specifically, more training, equipment, and facilities-intensive sports such as Lacrosse and Golf demonstrated higher median household incomes while less equipment-intensive sports such as Track and Field and Cross Country demonstrated lower median household incomes. Overall, looking at Harvard athletes’ hometowns before college provides interesting insights into Harvard’s athletic recruitment process and the backgrounds of Harvard’s student-athletes.</p><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. You can check out the data we used </em><a href="https://docs.google.com/spreadsheets/d/1JY8q51WJbEagKy96n3bMdiKgvgNGu17JkEX8qQ5kawU/edit#gid=557129802"><em>here</em></a><em>.</em></p><p><em>Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join </em><strong><em>us</em></strong><em>!</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=daf82fea501f" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/mapping-the-hometowns-of-harvards-student-athletes-daf82fea501f">Mapping The Hometowns of Harvard’s Student-Athletes</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Harvard’s Most (and Least) Desired Houses Updated: Housing Day 2020]]></title>
            <link>https://medium.com/harvard-open-data-project/harvards-most-and-least-desired-houses-updated-housing-day-2020-836f5003d36c?source=rss-b312fc49ff05------2</link>
            <guid isPermaLink="false">https://medium.com/p/836f5003d36c</guid>
            <dc:creator><![CDATA[Harvard Open Data Project]]></dc:creator>
            <pubDate>Tue, 07 Apr 2020 20:29:17 GMT</pubDate>
            <atom:updated>2020-04-07T22:16:53.225Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/960/0*SK6CJx8y0QIckPLJ" /></figure><p>by Alex Cheng, Kevin Huang, Ethan Lee</p><h3>Introduction</h3><p>Every March, freshmen wait in high anticipation as they look forward to learning which House they will live in for the rest of their Harvard careers. But this year, unusual circumstances caused by the pandemic outbreak of COVID-19 have forced them to wait almost a month longer. So, whether you’re a freshman sitting on the edge of your seat anxiously waiting to receive their housing results or an upperclassman excited to learn about the new members of your House community, read on to find a data-driven analysis of housing preferences to hype up the wait.</p><p>The Harvard Open Data Project collected data from roughly 275 freshmen through a survey sent over various email lists and compiled all things housing-related. Students ranked all 12 upperclassman houses from 1 (most desirable) to 12 (least desirable), reported their blocking group size, and recorded other miscellaneous data.</p><h3><strong>Lowell Reigns, Adams Gains</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*HpCaUZWjH1Ijlp4C" /></figure><p>We observed the overall sentiment towards different houses by finding the average ranking given to each one by our survey respondents. In the plot below, a taller bar signifies a greater dissatisfaction with receiving a house. Lowell, Winthrop, and Adams took the gold, silver, and bronze, while the three Quad dorms were (unsurprisingly) given the lowest rankings.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*deVAWTz1-f3vkSei" /></figure><p>The results this year see a big leap in preference for Adams compared to <a href="https://medium.com/harvard-open-data-project/harvards-most-and-least-desired-houses-updated-housing-day-2019-521e6db3e854">last year</a> and seems to be on the rise, jumping from 6th to 3rd this year and from 8th to 6th last year. This is likely attributed to the fact that Adams is one of the closest houses to the yard, and because residents will have the chance to live in swing housing and the Inn, due to the renovation that is slated to be finished in August 2023. <a href="https://medium.com/harvard-open-data-project/harvards-most-and-least-desired-houses-updated-housing-day-2019-521e6db3e854">Previous</a> <a href="https://medium.com/harvard-open-data-project/where-do-harvard-students-want-to-live-bd19c9be3dc7">years’</a> <a href="https://medium.com/harvard-open-data-project/housingday-4b0bd5e33b3e">data</a> has shown that renovations for houses generally associated with an increase in preference for that house. Quincy also saw a gain in preference (5th to 4th) while Dunster and Eliot seemed less appealing to the class of 2023 (3rd to 5th and 4th to 6th respectively).</p><p>In the graph below, you can see the change in rankings for houses every year that HODP has collected data on housing preferences. Looking at the trends, we saw the Quad dorms being consistently ranked in the bottom three while Winthrop was consistently in the top two.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/763/1*rxC9bEeKw8brY9GVO6cLJA.png" /></figure><h3><strong>Crazy Eights, Floaters’ Fate</strong></h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Oz3gmZQ0Htz9yjLv" /></figure><p>Similar to last year, we once again saw that the most popular blocking group size among respondents was eight, the maximum size for a blocking group. To determine this, we took the total number of blocking groups of each size and divided that number by the blocking group size. For example, with our total of 92 respondents reporting a group of 8, we divided 92 by 8 to get ~11.5% as the total number of blocking groups consisting of 8 people. The next most popular was a floater, or a single person, comprising 14.4% of blocking groups.</p><p>This is because smaller blocking groups are likely underrepresented in this survey, since those in larger blocking groups make up a larger proportion of the respondents. For example, if a person from one blocking group responded to this survey, they are more likely to share the survey with their blockmates and have them respond to the survey, which would happen more frequently in larger blocking groups. Although this method cannot account for all the confounding variables regarding blocking size, it provides a rough estimate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*vFq8DddhPUf_7vyA" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*QfA6JHjmXjJTkxIA" /></figure><p>Additionally, smaller and larger blocking groups seemed to have almost identical preferences with regards to housing, as seen by the graphs above.</p><h3><strong>Wins Delivered, Athletes Choose the River</strong></h3><blockquote><strong><em>When it came to housing preferences, varsity athletes diverged from the rest of the freshman class. Particularly, athletes tended to rank the quad houses less preferably.</em></strong></blockquote><p>Varsity athletes gave Currier House an average ranking of 10.79 in comparison to the non-athlete ranking of 9.17. This difference is similarly large for the two other quad houses: Pfoho and Cabot. The explanations for the differences in ranking become less apparent when referring to the other dorms. Dunster has extended meal times to accommodate the demanding schedules of many varsity athletes. Elliot and Kirkland are particularly close to the athletic facilities. However, this phenomenon may be more influenced by athletes disproportionately ranking the quad low than anything else.</p><p>It is also interesting to note that athletes tended to favor Adams, Lowell, and Quincy less than their non-athlete counterparts — perhaps due those houses being a bit farther from the athletic facilities than other river houses. But in the end, if it’s not the quad, athletes applaud.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/890/0*thh041LRdG04ySOa" /></figure><h3>Across Concentrations, Quad Met With Damnation</h3><p>We chose to look at the preferences of the three most common concentrations at Harvard: Economics, Computer Science, and Government. Computer Science could be especially interesting because of the new Allston campus set to open in Fall 2020. Among all three concentrations, trends were similar to the overall rankings.</p><blockquote><strong><em>Interestingly, Computer Science concentrators ranked Dunster above Quincy on average more than Government and Economics concentrators. However, it looked like the new Allston campus development had minimal impact on house preference.</em></strong></blockquote><p>Additionally, Economics concentrators preferred Eliot to Quincy while the other two concentrations did not favor Eliot as much. However, differences in both cases were quite small.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*PH97mXf-NCo3sU8F" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*D3DloKmkFFcpXPNn" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*ODWV1753-_QDnQzP" /></figure><h3>We Just Want a House</h3><p>So, we know that different types of people have different preferences for housing. “So what?” you might ask. Our next article will hopefully answer that question for you — we’ll be analyzing the housing results, which will be released soon. So even if you’re disappointed by the House that (storms your dorm) (emails you) Zoom-storms you (?), don’t worry. Whatever house you eventually get, you will learn to love it!</p><p>Wash your hands and stay safe!</p><p><em>This article was an analysis by the </em><a href="http://hodp.org/"><em>Harvard Open Data Project</em></a><em>, a student-faculty group that analyzes public Harvard data to hold Harvard institutions accountable. To see our raw data, look </em><a href="https://github.com/avcheng/housingpreferences2020"><strong><em>here</em></strong></a><em>.</em></p><p><em>Want to work on more projects like this?</em><a href="http://hodp.org/about/"><em> Join </em><strong><em>us</em></strong><em>!</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=836f5003d36c" width="1" height="1" alt=""><hr><p><a href="https://medium.com/harvard-open-data-project/harvards-most-and-least-desired-houses-updated-housing-day-2020-836f5003d36c">Harvard’s Most (and Least) Desired Houses Updated: Housing Day 2020</a> was originally published in <a href="https://medium.com/harvard-open-data-project">Harvard College Open Data Project</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>