Category: Uncategorized

The United States is not going to meet President Joe Biden’s target of 70% of the eligible population vaccinated by July 4th. This is not because of a lack of supply or capacity: in every state in the country, any eligible adult can get a vaccine.* Although issues of access surely explain why some Americans have not been vaccinated yet, it almost certainly does not explain the United States’ failure to meet President Biden’s target. The bigger issue is that many eligible Americans are choosing not to get vaccinated.

What explains this disappointing result? For a couple of weeks now, people have been noticing that there is a strong relationship between President Trump’s 2020 vote share and vaccination rates. Here is Seth Masket:

https://twitter.com/smotus/status/1404483957647831046?s=20

But state-level results are still a pretty coarse measures–just compare 71.4% vaccinated in Tompkins County, NY (where I live) to 50.7% in Yates County, NY (not far from here). But we can repeat the same analysis at the county level, across the American states, and here is what we find.

There is a strong negative correlation across nearly every state in the union between county-level Trump vote share in 2020 and vaccination rates, measured using data maintained by the CDC.

This would seem to be pretty clear evidence of a link between Trump support and vaccine hesitancy. But there are a lot of reasons why this correlation might exist that have nothing to do with Trump itself. Here are some alternative explanations:

• Trump-supporting counties are rural, and rural counties have lower vaccination rates due to supply, capacity, or distance-of-travel issues.
• Trump-supporting counties how small populations, and in counties with smaller populations the urgency of vaccination is lowers than in counties with large populations.
• Trump-supporting counties are also Republican counties, so we’re not picking up something particular to Trump, but rather an artifact of partisanship in 2020.
• Trump-supporting counties are majority white, and whites have lower vaccination rates. Now, this idea gets at the racial dimensions of vaccine hesitancy, although it runs exactly counter to the expectation that vaccine hesitancy is higher–and, critically, vaccine access is lower–among Black and Hispanic populations.

There are plenty of other ideas that we could explore here too. To do so, we can use the tried-and-true method of multiple regression.

The analysis below shows the correlation between county-level vaccination rates (18+) and a range of predictors that can capture the ideas above:

• Trump Swing: the county-level swing towards Trump between 2012 and 2020, to distinguish between a correlation due to Republican support and a correlation due to Trump support.
• Black, Hispanic, and Native population shares from the 2019 American Community Survey: to pick up racial and ethnic dimensions of vaccine hesitancy.
• County-level population (in log terms), also from the ACS.
• Indicators for how urban or rural a county is.
• State effects: to capture whatever differences in counties are associated with the state that the county is in.

Here is what happens when we enter these all in a regression.

The findings are clear: vaccination rates are negatively correlated with county-level Trump support in 2020, but not the county-level Trump swing, suggesting that whatever “Trump effect” there is is due to partisanship rather than Trump. Conditional on other variables, we also see lower vaccination rates in counties with larger Black and Hispanic population shares. There is no general relationship between county population or urban-rural factors.

Another way to slice this, though, would be to recognize that there are probably differences between rural counties in New York and rural counties in Wyoming, or Orange County, CA versus Orange County, FL, two large metropolitan counties. To capture this, I’ve created a new analysis that uses “state-by-urban/rural-county” fixed effects. Here is what we find:

How can we make sense of these findings? When we look at the table of scatterplots at the beginning of this post, we do see that the relationship between Trump support and vaccination rates is different in different parts of the country. Why might this be? To investigate, we can estimate a multilevel/hierarchical regression model that allows for the county-level correlations to themselves depend on state factors: state-level population, racial/ethnic share, and so forth. My analysis shows no evidence that those factors explain differences in county-level patterns across states: in other words, knowing the state-level support for Trump doesn’t help us to explain anything about vaccination rates that we cannot figure out using county-level support for Trump, and knowing the state-level Black population share doesn’t give us any more explanatory power than county-level Black population share, etc.

However, we can also check to see if there are geographical differences by allowing county-level correlations to vary by census division, a geographic unit defined by the U.S. Census Bureau.

Estimating such a model produces a mess of coefficients, interactions, and variance components that are hard to interpret. So to see how geography matters, I’ve plotted the estimates for four important variables across census divisions.

There is a ton to learn from this figure, so let’s take our time with it. Each plot shows you a “coefficient” by a census division: read these, for instance, as “the correlation between black population share and vaccination rates in the Pacific division, controlling for other factors.” The lines reflect 95% confidence intervals. We learn that

• There is always a negative correlation between county-level partisanship and vaccination rates, although the size of this correlation is stronger in some parts of the country (i.e. the West) than in others (i.e. New England).
• The distinctly Trumpian relationship between Trump support and vaccination rates is confined primarily to the middle Atlantic, the Midwest, the Middle South, and the Mountain regions. The Trump effect seems to be mostly a Rust Belt phenomenon.
• There is also always a correlation between Black population share and vaccination rates, net of other factors like urban/rural differences, state effects, and so forth. And importantly, this is not just something that we find in the South: it is evident everywhere, and actually tends to be smaller in the South** than in other parts of the country.
• There is no general pattern that we can see between Hispanic population share and vaccination rates, once we account for other factors in this more comprehensive model.

A fuller and more complete analysis of the political and social correlates of vaccination rates will have to wait for another time. But I have placed all of these data and replication commands online to allow anyone to recreate these analyses, update the data with new or more complete vaccination figures, and add new variables (partisanship of the governor! county-level measures of poverty!) that might further refine these preliminary findings.

NOTE

* The enormous privilege of it all. Right now there are private companies advertising three week trips for Indonesians to travel to the U.S. to get their vaccine, given the slow rollout of vaccines there.

** However, these coefficients are most precisely estimated in the South.

This post is written especially for current students in GOVT 3443/ASIAN 3334, Southeast Asian Politics.

Earlier this semester we briefly discussed the great kingdoms of pre-colonial Southeast Asia, from Dai Viet to the Khmer Empire to Pagan to Majapahit. One kingdom that we mentioned briefly was the Srivijayan Empire, a maritime state whose territory spanned Sumatra and the Malay peninsula. Here is the map that I showed you.

Later we mentioned Srivijaya one more time, in addressing the rise of the Malacca sultanate.

A theme in this part of our lectures was (1) the difficulty of describing the polities of pre-colonial Southeast Asia, owing to the incomplete and fragmentary evidence available to us today, (2) the role of colonial powers and colonial-era scholarship in producing the knowledge that we do have, and (3) the attendant result that it appears that we only can start talking about “politics” in depth in Southeast Asia when we start to get European records.

We tried our best to react against this, noting that in addition to the local evidence that has survived in the form of monuments, temples, constructions, inscriptions, and others, we do have other records left by Chinese and later Arab and European traders. We discussed the concept of the mandala as the dominant pre-colonial political form, of Zomia as outside of the lowland mandala polities, and the distinction between hulu and hilir. Still, we didn’t have much by the way of concrete discussion of any of these empires; for that, you can look to other great classes here for more on these empires’ history, architecture, art, and religion.

However, because I am not a historian of pre-colonial Southeast Asia (and neither are any of you! at least not yet), I presented these pre-colonial empires as basically facts. So, too, did your SarDesai reading. It was not up for debate whether or not the Khmer Empire was a great empire–it was–or whether Ayutthaya was the central political actor in the Chao Phraya valley. It would be unthinkable for me to even debate that.

As it turns out, the same is not true of the Srivijayan Empire.

This blog post by Liam Kelley, a historian of Vietnam, introduces a striking argument that has sparked a serious debate about the status of the Srivijayan Empire. The author’s claim is that the sources that scholars have used to describe Srivijaya as a great empire are talking about something else–in the author’s view, Angkor. The argument is not that there was no such thing as Srivijaya, as there are inscriptions that use the word Srivijaya that have been found in southeastern Sumatra. Rather, the point is that this Srivijaya is not the same thing as the polities described in the important pre-colonial sources that have served as our main evidence for the Srivijayan Empire.

He makes this argument by analyzing the primarily Chinese accounts that serve as the evidentiary basis for describing Srivijaya as an important pre-colonial polity, such as accounts of Chinese traders spending months in Srivijaya learning Sanskrit before later traveling to what is today India. We do not have local evidence of this, we have only the accounts of others, as well as the names of the places that they used, written in the Chinese of the time. Kelley argues that those words (such as Shi-Li-Fo-Shi) do not describe Srivijaya. You can read his posts for all the gory details.*

Remember the important distinction between Srivijaya and the empires of the mainland. Whereas many of those formed around great riverine systems (Mekong, Irrawaddy, Red River, etc.) that allowed the empires to amass large population bases through the intensive cultivation of rice, Srivijaya was a maritime-facing polity. It has been described recently as a thalassocracy: an empire with a maritime focus. Majapahit on Java was a thalassocracy. But Majapahit also left reams of evidence in Java itself of its own existence, and of its own greatness. The same is not true of Srivijaya.

I must insist that I am not qualified to evaluate the argument that Kelley presents. I must also insist that even if it were true that the sources used to describe Srivijaya are actually talking about some other place, this does not logically entail that there was no such thing as Srivijaya: this word appears in inscriptions found in Sumatra, so it describes something.** But there are some important points to take away from this emergent debate, even if it turns out that Kelley is entirely wrong.

First, the evidentiary basis for what we know about Srivijaya is very incomplete. Kelley is not the first to note that the analyses of Srivijaya rest to a large degree on the accounts from others traveling through the region. There is precious little evidence of Srivijaya that comes from the territories where it was located. To say anything about the politics of pre-colonial Southeast Asia in this case requires us to work very hard to assemble an evidentiary base.

Second, the effort to discover and analyze Srivijaya is intimately tied with colonial-era scholarship. I did not fully appreciate, for example, that the first concrete proposal of the existence of a Srivijayan Empire came from George Cœdès–a French archeologist–in 1918. That is not that long ago! He has been described as having “discovered” Srivijaya. A lot of scholarship about Srivijaya rests on his interpretations of words in Old Malay found in contemporary Thailand, and those interpretations are much more contested than I realized.

Third, these facts interact in what might be uncomfortable ways when it comes to post-colonial scholarship and our understanding of the pre-colonial polities of Southeast Asia. The concept of the Srivijayan Empire is important to the concept of Indonesia itself, much like Majapahit is, as a pre-colonial antecedent to the post-colonial state. Much of the post-colonial scholarship on Southeast Asia sought to uncover what John Smail called in 1961 an “autonomous history of Southeast Asia.” That is a history of Southeast Asia that sees the region in its own local terms, rather than merely as a reflection of Indic, Chinese, Arab, and European influences*** as they spread culturally, economically, religiously, and politically throughout the region.

I value this search for an autonomous history of Southeast Asia as well. And yet I am forced to think critically about the possibility of writing that history when the accounts that we use to do so were not produced by Southeast Asians in Southeast Asia.

NOTES

* Also the graphics are great. I wish that these existed as a series of TikToks too.

** But careful. Kelley suggests that Srivijaya describes a person, not a polity. And indeed, to anyone familiar with the Sanskrit influence on naming conventions in Southeast Asia, when you stop to think about it, “Sri Vijaya” sounds like a royal title.

*** Here we grapple with the question of Orientalism in the study of Southeast Asia, because the region itself is the subject to the external gaze of others in the East, not just Europeans.