11/19 Philip Waggoner

[ehuppert]

Comment below with questions or thoughts about the reading for this week's workshop.

Please make your comments by Wednesday 11:59 PM, and upvote at least five of your peers' comments on Thursday prior to the workshop. You need to use 'thumbs-up' for your reactions to count towards 'top comments,' but you can use other emojis on top of the thumbs up.

[JadeBenson]

Thank you for this fascinating and incredibly relevant research! I am particularly interested in your broad definition and conceptualization of "epidemics." As you mention, "epidemic" has been used to describe practically any public health issue and has been extended far beyond infectious diseases to include crime, obesity, opioid use, loneliness, heart disease, car accidents, the list really is all encompassing. Although all of these problems might be rhetorically considered "epidemics," they will necessitate different policy action. For example, the negativity of COVID-related policies might reflect urgency and severity, rather than a public health issue that operates on a different time scale, like obesity. I would love to hear your thoughts on how your conceptualization of "epidemics" might influence these results? As a follow-up, do you think your results reflect public health policy as a whole from the 1970s onward given the liberal use of "epidemics"?

[sabinahartnett]

Thank you for sharing your research with us Professor Waggoner!

In the conclusion of your working paper Pandemic Policymaking: Learning the Lower Dimensional Manifold of Congressional Responsiveness you write that "future work might pick up on these results by considering… different substantive topics (e.g. the economy, elections, and so on)” and so my question is about topic clustering:

Many critics of the American system (largely focused on capitalism) have highlighted the ways in which the COVID pandemic has exacerbated existing issues, divides, and instabilities of the country (arguing that these are some of the greatest threats of the pandemic). I’m wondering if, in line with this argument, COVID has not necessarily introduced new legislative demands, but rather exacerbated the urgency for existing demands (i.e. affordable health care, small business support as monopolies increase their reach...). If this is the case, I would expect there to be thematic clustering of the COVID bills with pre-COVID bills addressing similar issues (for example: do bills addressing unemployment benefits cluster together) - is this something that you were able to investigate or something you think worth considering in the cluster analysis?

[chrismaurice0]

Very interesting research, Professor Waggoner. I am impressed at the speed at which you were able to process your data and write a thorough, thought-provoking paper. My question is on the broader implications of your papers. While you focus on the differences between pre-covid and covid legislating, I think there is a sub-narrative on how the institution of Congress is slow to change and the legislating has remained constant despite greater polarization, more divisiveness, and even new members of congress. Given the results of your papers, how would you respond when a candidate running for congress promises to shake up the system, “drain the swamp”, or bring change to congress?

[mikepackard415]

Thank you for sharing this research Professor Waggoner! I'm looking forward to hearing you discuss it more.

I'm interested in your description of "(restored) hope in the prime institution of policymaking" derived from the stability that your empirical research identified between the pre-COVID and COVID periods. Would it be an equally valid reaction to feel that the current environment should induce a drastic change in policymaking patterns, and therefore to interpret the stability in a more negative light? In other words, I'm curious why, in your view, consistency in policymaking ("a formulaic approach") in the face of unprecedented crisis would be celebrated. Let me know if I'm misreading you! Thanks!

[adarshmathew]

Thank you for presenting your working paper in our workshop, Prof. Waggoner!

This was a fun paper to read and deconstruct, especially since you take the approach of unsupervised learning to describe phenomena. It gives us a great opportunity to understand the descriptive capabilities of these methods and how we can use them to test hypotheses. I have a few questions:

is policymaking in response to COVID-19, which is occurring in this divided era, substantively different from policymaking in response to similar epidemics in the past? (Page 2)

• Your key question, copied above, makes the assumption that there were no epidemics in other periods of high polarization, along with a later claim that pandemic/epidemic policies are 'apolitical'. I think these are big assumptions to make, especially for COVID, given that it's an election year. Also, the AIDS epidemic was fairly political in its lifecycle: the Reagan administration drastically changed how they handled the AIDS epidemic and funding due to political pressures, moving from characterizing it as a 'homosexual disease' to something that was killing off young, white men. It was also a key issue in the Reagan-Clinton election, on public health and 'morality' lines. Without a measure of polarization, how would you explain this assumption?

This suggests that there is less of an “evolutionary trend” in pandemic policymaking, where instead there is striking uniformity in this type of Congressional policymaking, despite currently operating an era of hyperpolarization [25], deepening mass political polarization [26], and ineffective governance [27]. (Page 4, summary of finding)

bill type, committee configuration for each bill, state of prime sponsor, chamber of prime sponsor, date of introduction, and date of final action. (Page 6, features for unsupervised UMAP)

bill type, cosponsors,committee configurations, state of primary sponsor, and the chamber of the primary sponsor. (Page 15, features for supervised projection)

• The first quote here, from a summary of your results, builds on my concerns from the previous question. Given that your feature space uses a stable taxonomy of bills and a stable configuration of committees, the feature space hasn't changed much in pre- and post-COVID worlds. Bills are still proposed by Republicans or Democrats, there are certain committees which generate these bills, and there's some sort of continuity in the region of congresspeople putting up these bills. Can't your result be attributed to the stability/rigidity of the institution of Congress rather than any trend in policymaking? That ideas/proposals are coming from the same spaces as before, indicating the rigidity of the institution? And as such, your feature space is limited in unearthing meaningful differences between the different versions of Congress (1973-2020)? (Shout-out to @mikepackard415's question)

• Additionally, since you're attempting to capture polarization and how different iterations of Congress approach an ostensibly public health crisis in different manners, does it make sense to include a host of other (feasible) variables like: the party in power at the Executive; the partisan split in Congress at large and in the committees; are the sponsors and co-sponsors senior or junior legislators; do they come from a safe seat or a swing state (building on Grimmer's paper using text analysis of press releases); or is it an election year. Do you think these variables would be able to capture the minutiae of polarization and electoral politics?

• Finally, on a methodological note: You assume that these epidemic bills exist along a common manifold. While I get your theoretical reason to argue that, are we sure UMAP is unearthing that? I ask this because I've not been able to defend my use of representation learning methods in the past satisfactorily, other than saying "We hope that it captures the manifold". Do you have suggestions on how to showcase the validity of the learned representation? Would exploring the cases along the manifold -- where do AIDS bills lie, where do H1N1 bills lie on the manifold, how are they substantively different -- help us in establishing the validity of the learned manifold?

Thank you again!

[MkramerPsych]

Thank you for sharing your research with us! This is the first paper I have read on UMAP, so forgive me if my question belies my ignorance.

To further @JadeBenson 's question, I am interested in the relative contributions of public health bills versus other national issues (such as gun control, tax code changes, etc) to the results of the UMAP? You make a point of saying that US politicians do not usually author legislation on international issues, but COVID-19 presents a very specific case of a singular public health issue at the international level. If we observe a difference between COVID (and presumably legislature affected by the state of the world as a result of COVID) and other "epidemics", is it actually a safe assumption to assume that they exist along a common manifold? Is there a way of performing a multi-phase analysis that separates these issues by the nature of the problem (public health, economic recovery,...) that would be beneficial to learning a lower-dimensional representation?

[wanitchayap]

Thank you for your interesting research! I am looking forward to your presentation.

I have a small question about your future direction in the manifold paper. You suggested that this approach could be applied to the other governments beyond the US'. I assume that we can use the same approach to apply to any government if we are going to compare the same government (as of the same country) across different timelines or across different political parties. However, I am not sure if this is also applicable if we want to compare the governments of different countries...? My knowledge in the field is very limited, so please pardon me if I am wrong, but I think that there would be a lot of cultural/national differences that it might not make sense to project the policies (whether it is from the bills' metadata or the bills' text) from different governments onto the same dimensions...? Could you explain this more?

Ps. And pls answer @adarshmathew questions!

[Anqi-Zhou]

Thanks for sharing this inspiring paper! I understand your visualization figures showing the pandemic policy making is distinct in the periods, but what's the potential mechanism behind the change? Are there any insights or hypothesis?

[boyafu]

Thanks for sharing! I was fascinated by the analysis of how certain events have different influences on history. I was wondering how concrete the conclusions are to attribute these differences to certain events, since there might be other confounding variables such as the changing political climates for both parties.

[bakerwho]

Thanks for presenting your research!

With PCA, the weighting on the original dimensions for each component allows us to interpret the reduced dimension plots with some real-world variable adherent. Is this possible with UMAP? How can we better characterize the differences the UMAP plot highlights in terms of real-world adherent concepts?

Big +1s to the questions from @adarshmathew and @wanitchayap

[Raychanan]

Hi Dr. Waggoner, I think you've been emphasizing stability and continuity of policy. But my question right now is, why do we place so much emphasis on policy stability and continuity? Or is there an advantage to policy instability and discontinuity in COVID-19? Does instability provide an advantage in the face of changing events?

[YuxinNg]

Thanks for the presentation. My question is the same to @wanitchayap, applying similar approach to the other governments beyond the US could be interesting, but any national, cultural, political, systematical difference may also need to be considered. I am wondering if you could spend some time talking about this issue? Also, in this article, you focused on pandemic-related policies, I am wondering if we could apply similar approach to other incidents-related policies, maybe incidents like terrorism? Thank you!

[wu-yt]

Thanks for presenting! I my question is that the current pandemic is different from the past events, so is it enough to do a direct comparison like this without consideration for more factors?

[Yutong0828]

Thanks for sharing the wonderful papers! My question is quite general, which is related to the dataset you choose to study, the bill data. It seems to be a very comprehensive dataset which contains various information for studying government policymaking, however, I find it a bit hard to understand the significance and details of these bill data, and it would be very helpful if you could introduce some additional information about the bill dataset, and its role in political science study. For example, how did you find out that the bill data are appropriate being used for your research aim? Thanks very much!

[nwrim]

Thanks for presenting! I had a very similar question with @adarshmathew and @JadeBenson (although not as in quite a detail) on the fact that "epidemic" is a loosely defined umbrella term and that we had never really seen anything like the COVID-19 pandemic before. I will not be able to elaborate more on the detailed questions above and hope there could be a good discussion regarding the issue!

From a more meta-perspectives, I have seen many people advising any attempts on using COVID-19 data until the pandemic is actually over. I think this is because the pattern of the pandemic are apt to change as it progresses, and the whole aftermath cannot be quite captured unless we can declare it over. With an astonishing amount of cases still surging in America and other countries, this pandemic is far from over (but finger-crossed for the vaccine). Do you have any wisdom to share on this matter?

[mintaow]

Thanks for sharing the paper as well as your note! My question lies more on the methodology side, where you implement a supervised learning method as supplementary to the original unsupervised study. I am wondering how you tackle the labeling through this transition. It is an interesting approach and also is "substantively motivated for present purposes", as you have mentioned in the article, but isn't it going to introduce some bias during the pre-Covid data imputation process? Thanks!

[a-bosko]

Thank you for sharing your papers, Prof. Waggoner! I found these papers to be extremely interesting and relevant to today's society. I appreciate your approach towards understanding pandemic policymaking and really enjoyed learning about the subject.

I think that everyone else this week has asked some great questions! I am also curious on what you define as an "epidemic", since the word is often times tossed around without thought. Also, you mentioned that "political ideologies (liberal and conservative) is stronger than ever before in American history." What do you believe is the extent of the effect that this political divide has on how America is handling the current pandemic? What do you believe the government or society could have done differently?

Thank you!

[bowen-w-zheng]

Thanks for sharing your research Prof. Waggoner! I have some questions with regard to the methodology. Related to the last question by @adarshmathew, I wonder how you would justify the assumption that all legislators act in a common space. I am not familiar with UMAP, but it seems like it's using nearest neighbours to restore local structures of the manifold. Is this method robust to the shape of the true underlying manifold, and are there methods to check the validity of the learning outcomes?

[luckycindyyx]

Thank you for sharing such an interesting research! I have a general question. As the COVID-19 pandemic progresses to its tenth month, people are becoming more and more aware of covid-19. In which area do you think future research on COVID-19 pandemic will be more focused? And what will the changes be in research methods and algorithms?

[xzmerry]

This is a really quick response to reality! Thanks for sharing these papers, especially the pandemic one. An interesting and meaningful finding that "uniformity in Congressional policymaking related to these types of large-scale crises".

My question is a kind of technical question. When you compare whether policymaking at different periods, you seem to derive the conclusion based on their recognizable distribution and patterns on graphs directly. As I have never used this method before, I wonder whether we can conclude that policymaking of different stages are similar or different based on the recognizable patterns of their visualization, or we'd better use some quantitive measures to get their similarity and verify it in a similar way as a hypothesis test.

And besides, how you cope with possible biases caused by using bill-level data? Because ways to make enact a policy could have changed a lot over a period of around fifty years.

[vinsonyz]

Thank you for sharing this exciting research. As students majoring in computational social science, what research questions should we focus on?

[bjcliang-uchi]

This is a very interesting and up-to-date paper! I also like how much it can be an exemplar of how to conduct an NLP-based political science research project. My questions are, first, is there any conclusion you mention in the paper that has to be derived from a quantitative approach? I mean, party-based polarization, Republican's emphasis on coronavirus and international affairs, and Democrats' emphasis on social security and healthcare are expected even in qualitative research. Second, I am wondering how to interpret "good" vs "bad" or "positive" vs "negative", because I suspect that the "good" for Republicans could be "bad" for Democrats given the level of polarization.

[NikkiTing]

Thank you for sharing your work! I have the same question as @mikepackard415 on the interpretation of stability in policymaking. In addition, would this interpretation hold if the same study is applied to other countries? Thank you.

[linghui-wu]

It is really exciting to see social science research focusing on COVID-19. I have a similar inquiry regarding the generalization of the conclusion as mentioned by @wanitchayap, though the perspective I am interested in is slightly different. What do you think of the trade-off between the generalized application of the research results and the peculiarities possessed by different societies that we endeavor to comprehend?

[SoyBison]

Hi Dr. Waggoner, thanks for putting yourself on the altar for us this week!

My concern has to do with the validity of doing research on a world event while it's going on. While the writing of your paper seems to be distanced from the emotional impact of the pandemic, I think we all know that it can be somewhat difficult in practice. The other side of this coin is the fact that data takes time to be validated, it may take a year or two (fast in the grand scheme of things) for us to get really good data on these events as they unfold. For example, after the election solidifies we may see a shift in legislator behavior which would change the impact of your analysis.

Eager to hear your thoughts on this.

[Bin-ary-Li]

Hi. This is very interesting work.

My question is how do you check whether the data conform to the assumptions of UMAP? Also, what do you think about the challenge of interpreting and communicating the components from dimensional reduction (linear or nonlinear)?

[WMhYang]

Thank you very much for your papers. My question is to some extent related to your corpus construction. Could you please explain more about what epidemics, or pandemics, took place from 1973 - 2018? Is there an obvious cluster in the time dimension, e.g. more bills around 2009 because of the H1N1? Could we curtail the pre-COVID corpus further by characeterizing the related epidemics (AIDS/HIV, H1N1, etc)? Thanks.

[jinfei1125]

Thank you for sharing your excellent papers. They are very related to the current period and interesting to read! Correlated to this week's Perspective courses, I am extremely curious about how you come up with the questions you want to answer in your papers. These are two questions in the CCSE paper:

Topic models will help address related questions like, does evolution exist, or are bill topics relatively stable? And, are foci of topic structures similar between periods or not? Sentiment analysis will also help address the evolutionary question, but will help address slightly different questions like, what is the tone, and does it shift over time? Do we see differences across chambers?

Could you talk about how you come up with these questions? Do you have other questions that interest you but you don't work on? I am looking forward to your answer. Btw I also love the parts you describe in how you collect, clean and analyze your data (it reminds me of a programming assignment of text analysis in Jon's CS-SSA class and I am looking forward to learning more skills and models in your course next quarter! :D).

[NaiyuJ]

Wonderful study! I have a question on the sentiment analysis part. In my sense, unlike social media posts or other more life-like scenes, the tones or the words that are used in the content of proposed legislation may tend to be more fixed and more negative (because the ultimate goal is to improve policy-making?). Do you think conduct sentiment analysis on legislative proposals is as meaningful as sentiment analysis on social media posts or other everyday articles?

[afchao]

Thank you for sharing your research with our group! As do many of my colleagues, I have a question about the decision to fold public health designations of "pandemic" in the context of COVID-19 and "epidemic" in the context of COVID-19 into the same conceptual category while also equivocating the latter with "epidemic" in its more colloquial usage (e.g. "the opioid epidemic"). Given the result of your analyses and the fact you mention of that politicians rarely author legislation on international issues, is it reasonable to consider the American policy approach to COVID-19 as a reduction of the international issue to a national one? If so, could that play a role in the similarities you described? Would you expect substantial differences in policy structure between legislation focused on the nation vs. legislation focused on its relationship with other nations? Intuitively, it seems strange for the national approach to an international issue to have so much in common with prior approaches to domestic issues (although that may just be a result of my relative ignorance when it comes to legislative sausage-making)

[chiayunc]

Hi, thank you for sharing your work. I have two questions regarding the data you are using. Since the data you are using is bill-level, I suspect the static-ness of the data would be natural given that it is meant to gather most consensus, and the law is known to be slow-reacting to current issues. I was wondering if you conduct this experiment across all different topics of bills, would you expect this trend to appear?

Secondly, I think it would be quite interesting to also look at the text of the bill itself, as it will certainly reveal new paradigms of governmental management. Could you share your thoughts on this?

[ChivLiu]

Thank you for sharing this presentation! I wonder that for the state government and the federal government, do they have different systems to calculate their bills for the policies? I know some states are wealthy, so they could spend a lot helping people maintain basic livings, but some states have to ask the federal government to help them. Therefore, what concepts do we need to think of when measuring the bills for different states?

[MengChenC]

Thank you for the two pieces of research. I am curious that due to the continuously surging cases and seemingly endless war of the pandemic in the U.S., the policy making and negotiation in congress is still with uncertainty, what can we expect from the computational methods (either mentioned in the paper or not) to deconstruct the mechanisms of policy making? Thank you.

[k-partha]

Thank you for presenting! How would the rather broad (and non-homogenous) conception of "epidemic" affect the generalizability of this research?

[mingtao-gao]

Thank you for sharing these interesting papers! The findings and results are very appealing to me. In the first paper, it shows even though the political contexts have changed but patterns of policymaking responding to pandemics stays stable over years. This is a very interesting result as it implies no matter which party is in power the policymaking responding to COVID-19 will be similar.

My question is that "is it possible that political ideologies and images can largely affect the results of pandemics?"

Specifically, take COVID-19 as an example, based on research result, the congressional pandemic policymaking follows the same pattern as previous pandemics, which gives a baseline efficiency. However, the political ideologies responding to pandemics, for example "covid-19 is just a flu" and the idea of not wearing masks, can be largely affected by which political party is in power or is more influential. Thus, the efficiency of policy like "wearing masks in public" can be largely reduced due to those ideologies.

[Dxu1]

Thank you for sharing both interesting papers. I have some conceptual questions. What exactly is the pattern and approach of policy-making that remained stable over time? Is it the process of legislating a bill? In the second paper, you noted that despite the approach to policy-making might be time-consistent, the tone around it gets more divided over time. What do you think of its implication? It is not intuitive to conceive it that partisan or ideological difference would not essentially alter policy-making (and hence potentially put people's welfare at risk) at pandemic times, despite a "nosier" debate around it?

[TwoCentimetre]

Thanks for sharing. In both paper, it seems that our computational methods can only play as a role to uncover the pattern of behaviors of the government. It is unclear if the computational methods can actually facilitate the process of policy making, which I think is more interesting and important. Also, it seems that in both paper, it is assumed that political ideologies play a role in policy making. I am confused that the control of pandemic is not a traditional political issue so why it is assumed to be something related in this case?

[Qiuyu-Li]

Thank you for providing this wonderful presentation. My question is much consistent with previous comments: how would you address the comparability of COVID-19 with other events? I can understand that "continuity" and "uniqueness" are meaningful concepts in uncovering the policy-making patterns from a historic perspective, but how does this method of retrospecting help us evaluate the effectiveness of current policies? In what situation does a historic perspective applicable?

[LFShan]

Thank you for sharing both papers. As someone who is not very familiar with politics. It is very interesting to learn that computational methods can be applied to politics where quantitative data are often not available. I would like to know if different presidents from different parties could influence the results in the paper?

[hihowme]

Thanks a lot for your presentation! I am wondering is there any research related to this, but from a economic perspective? Thanks a lot!

[Yaweili19]

Thank you for sharing these two interesting research papers! They will be very helpful to my dissertation. I hope to learn more about these studies in your presentation. The methodology part is kind of beyond my knowledge range though, and I would appreciate it if you could go over that part in a greater detail.

[cytwill]

This is an interesting study. And I am excited to see that you use Umap in your analysis and your interpretation of UMAP as a combination of supervised and unsupervised learning is quite attractive. I would like to try it later~

However, there are some issues I hope you would like to clarify more about: 1) The projection part: As you hope to confirm the closeness between the policymaking traits before and after COVID19, I think it might be more convincing to calculate some similarity metrics (e.g, cosine similarity) between the representation of the before and after trait vectors instead of explaining it in a descriptive way (more than visualization). 2) It seems that you have excluded the textual data and time-dependent data from your analysis, however, I think both of them are super important to this issue. If there are any differences between the two patterns, I think they are very likely hidden in time-dependent data, so you might need to find out some approaches to include the time-dependent data and standardized them to a "uniform" scale. And for textual data, you can transfer them into vectors with many available NLP methods (word2vec, tf-idf, topic modeling....)

[ydeng117]

Thank you for presenting such interesting articles. It is surprising to find out that even the tone toward the pandemic had changed significantly, the pandemic policymaking remains largely stable over time. I wonder what contributes to the stagnant policymaking pattern in the US. Compared to countries which did a better job in controlling the spread of the virus, such as China, Korean, Japan, etc, do those governments also have the same stagnant policymaking traits over time?

[lulululugagaga]

Thanks for your sharing. I think adarshmathew's questions pretty cover most of the things that I'm confused of. Maybe just one more: from your experience, what factors might affect such policy-making mechanism in a similar way like pandemic? Since pandemic happens in a quite low frequency(hopefully), I wonder if the result of this study can be replicated to other factors?

[FrederickZhengHe]

Thanks a lot for your paper! I just have a question that is not so relevant: I know you will lecture in the Perspective course next quarter, so what will be the primary language of the course? Is it Python or R?

[Yilun0221]

Thank you for the presentation! I think this topic is very interesting which reminds me of the presentation that Prof. Gary King made previously in our workshop. I am very interested in the part where you collected policies related to pandemic. I think it is very hard to be comprehensive since every goverment in different countries, states or even counties (or cities) will have different websites to post the policies where the format will be very diverse. I wonder whether you can elaborate more in this part during the workshop?

[yutianlai]

Thanks for sharing! I'm wondering if the result could be applied to other countries.

[luxin-tian]

Thanks a lot for sharing! I was wondering how different policy-making constitution can be captured in this study.

[MegicLF]

Thanks for your presentation in advance! I wonder why it is reasonable to access current COVID using the past pandemics as they all have different scales.

[PAHADRIANUS]

Thank you for sharing your work! Through these two papers you effectively demonstrated that while on the one hand the policy responses to COVID-19 bear signs of increasing political fracture and division, on the other hand evidence can be found that certain stability and predictability in policy making were consistent with pre-COVID periods. My question is, then, whether this display of stability is a cheerful indication: we have long observed how political bickering impeded the enactment of potentially progressive government policies, yet in the past such impediments were conducted with courtesy and decorum. The late emergence of Trumpism simply made the divisions more exposed and public. I wonder if we can find (I am quite confident of so) quantitative evidence of partisan difference in reacting to the previous HIV or H1N1 crises.