Go to the U of M home page

Thursday, March 5, 2020

Mapping the 2020 Minnesota primary results

Minnesota's presidential primary was held March 3rd on ‘Super Tuesday’, which was contested by Democratic candidates and in every practical sense, uncontested for Republican incumbent Donald Trump. A big surprise was Democratic candidate Joe Biden receiving the largest share of votes (38.6%), even though an earlier poll found him at a distant 4th place.

Biden’s vote share was followed by Bernie Sanders (29.9%), Elizabeth Warren (15.4%), Mike Bloomberg (8.3%) and Amy Klobuchar (5.6%). Klobuchar dropped out of the race a day before the election, giving her endorsement to Biden amid a slow showing in other states and criticism by Black Lives Matter and LGBTQ leaders for her past actions relating to civil rights issues. Being the former state frontrunner, Klobuchar's endorsement of Biden undoubtedly fueled his stronger performance in the state.

The Leading Candidates

As the below maps show, Biden gained large vote shares throughout the state, including in rural areas and middle and outer suburban areas in the Twin Cities metro, places where Klobuchar gained a foothold in the 2018 General Election.

Sanders and Warren’s vote gains were strongest in Minneapolis and Saint Paul, with Sanders also gaining large shares of votes in precincts near Indian Reservations in the state. Bloomberg, who has since also dropped out, tended to have larger vote shares scattered in outstate precincts. Of the 27 precincts where he gained over 30% of the vote, none were in the 11-county metro area.

And although Klobuchar dropped out the race before the election, she still gained large shares of votes in rural areas, often in precincts adjacent to where Biden gained large vote shares. Beyond those locations there were insufficient numbers of persons voting for Democratic candidates—rural precincts that were strongly Republican in the last general election.

The primary results suggest that the metro-outstate divide appears to be widening even more. A significant fact is that nearly three-quarters of the votes for Democratic candidates (73%) came from precincts in the 11-county metro area (544,334 of 744,291 of the votes). By comparison, in the last (gubernatorial) primary in 2018, 69% of the vote for Democratic candidates came from the metro.

Party Ideology

It is hard to discern a general pattern about how the results reflect the ideological makeup of voters, due to the large number of candidates in the primary race, among other measurement and validity issues. Although an imperfect method, by grouping progressive-leaning and moderate-leaning candidates according to their perceived ideologies, we can show a clearer pattern of where voters selected such candidates.

The below maps show that progressive candidates performed best in the cities of Minneapolis and Saint Paul (36% of all of their votes came from these cities) while moderate candidates had better results in Outstate Minnesota, outside of highly populated municipalities, and scattered across several outer suburban jurisdictions in the Twin Cities metro. 

These results suggest that the fight over the ideological viewpoint of the Democratic Party is occurring mostly in inner and middle ring suburbs in the Twin Cities, and to some extent in smaller outstate Minnesota cities.

Growth and Decline of Democratic Primary Votes

There have also been changing vote patterns over the last two primary elections, following a 2018 general election where turnout rates were high across the state, and historically high for the state’s central cities, which surpassed rural turnout rates for the first time in 20 years. 

Although data is not available to calculate turnout rates for the state’s precincts, we are able to map out percentage changes in the number of votes cast for democratic candidates over the last two elections: the 2018 gubernatorial primary and the 2020 presidential primary. We did not make comparisons to the last presidential primary in 2016, because votes were made in the party’s caucus system that year rather than the state’s primary system, as is done now in 2020. Incidentally the 2016 caucus gained only 28% of the votes counted in this year’s primary.

When comparing the voting change between the last two Democratic primaries, there are clear spatial patterns in the growth and decline of votes submitted between the primaries. Growth rates are highest in the Twin Cities particularly in growing outer suburban areas, but also in many central city and first-and second-ring suburbs.  

The fact that the current 2020 primary is for a presidential election, while the 2018 election was non-presidential makes it somewhat unsurprising that there are growing numbers of persons casting a vote in precincts for the current primary—an overall growth rate of 27.6% since 2018.

What is more surprising is that Outstate precincts are about as likely to have declining rates of votes for the Democratic primary as they are to have growing rates. Could this be an indication of further declining support for Democratic candidates outstate? It is notable that there are several exceptions to this, with strong growth in the Democratic vote rates in and around the Rochester area in Southeastern Minnesota, in Northwestern Minnesota around Moorhead and in Becker and Beltrami counties as well in the Duluth area.

Candidate Vote Results in Growing and Declining Precincts

The location where the candidates draw votes could have implications for how well they fare in a general election. While moderates gained more votes than progressives statewide, they narrowly won in the metro area, where progressive candidates performed more strongly. Yet if a progressive candidate is to win the general election they would clearly need to gain more support outside of their base in the central cities. 

On the other hand, moderate candidates should not ignore the fact that in the last general election central cities had turnout rates only outmatched by Twin Cities’ suburban areas. Furthermore, progressive candidates received larger shares of their votes from precincts that had the largest increases in votes for the Democratic primary. The chart below shows that moderate candidates Klobuchar, Bloomberg and Biden tended to fare worse in gaining votes from precincts with primary growth rates of 15% or more than their progressive counterparts, Sanders and Warren. 

On the Road to November

Diverse inner and middle-ring suburbs are places where the Democratic Party is competing the most to determine its ideological viewpoint and where many general election races have been settled with Republican candidates over the last decade. Whether there is an ideological split in the party that hampers success in the general election remains to be seen. In the last election, both moderate and progressive candidates won at the state-level, although progressives like Keith Ellison won by smaller margins. 

Another concern for the party is the emerging urban and rural divide that appears to be further widening in the latest primary. Yet again there are some positive outstate results for Democrats from this primary, which showed large growth rates in the Democratic vote in Southeastern and Northwestern Minnesota, as well as in the Duluth area. 

While the presidential primary was largely uncontested for Donald Trump, Republican candidates can look favorably toward the decline in the Democratic vote across large sections of outstate Minnesota, while looking less favorably at gains in the Democratic vote across the developed portion of the large and growing Twin Cities metro area.

Thursday, January 30, 2020

Response to “Research shows progressive places, like Minneapolis, have the worst achievement gaps.”

On January 24th the Star Tribune published an opinion piece by Nekima Levy Armstrong, entitled “Research shows progressive places, like Minneapolis, have the worst achievement gaps.” In the piece, Levy Armstrong argues that racial achievement gaps are smaller in politically conservative cities, and that the public should demand progressive cities like Minneapolis to do better at closing those gaps. For empirical support, Levy Armstrong relies on a report by Brightbeam, a project of the school-reform oriented outlet Education Post. The Brightbeam report analyzes student performance gaps between racial groups in the most progressive and conservative large cities in the U.S.

Unquestionably, as in almost all American cities and regions, Minneapolis and the Twin Cities region are home to significant racial disparities, and these disparities are an important problem that policymakers should address. However, there are numerous substantive and methodological issues contained in the Brightbeam report. These problems cast serious doubt on the report’s conclusion that progressive cities are especially likely to produce gaps , or its suggestion that leaders in more conservative cities merit praise for tackling racial gaps.

The Brightbeam report, entitled “The Secret Shame,” is full of rhetorical claims based on results from a variety of unsound research practices. Most importantly, racial segregation is mentioned only once in the report and is not given consideration in the data analysis, an unacceptable omission given the body of research concluding that segregation is a driving factor in producing racial performance gaps.

The report bases its conclusions on analysis of two groups of 12 cities, one putatively consisting of the most progressive cities in the nation, and the other putatively consisting of the most conservative cities. It does not cite data sources or provide much of a methodological account of the research techniques it utilizes. While the report compares two groups containing a total of 24 cities, it fails to conduct analysis of the remainder of the universe of 67 cities described in the report, which show a range political policy preferences.

The report determines the political lean of each city by using surveys of policy preferences of city residents. However, at the outset, this poses serious methodological issues. First, cities and school districts are separate jurisdictions, and the policy preferences of city leaders may not mirror those of its associated district. Indeed, the survey used does not contain any question about schools or school policy, and instead focuses on respondent attitudes towards topics like the federal government.

An even more fundamental issue is that educational policymaking often does not occur at the city or district level. The governmental entities with the greatest control over questions of school policy are typically the state legislature and state department of education. Brightbeam makes no attempt to incorporate any measure of state policy or political lean, meaning that, in effect, the report assumes that city districts have singular and ultimate control over their schools.

But perhaps the most serious concern about the report’s measurement of citywide political opinion is that it assumes that each city is conterminous with its associated district. In reality, cities and school district boundaries often do not always overlap spatially. In fact, about half the cities listed in the survey are associated with school districts that either comprise a fraction of the municipality, or alternatively, include multiple municipalities and unincorporated areas. Importantly, these mismatches are more common in some parts of the country than others (for instance, in the south, districts are often countywide in scope). The report’s “conservative” cities are more likely to have boundary mismatches—6 of the 12 cities used for the report’s analysis of conservative cities have a major discrepancy.

To illustrate, Oklahoma City actually contains nineteen school districts within the city’s boundary, with the city district the most centrally located. The report offers no explanation of how it addressed these scenarios.

The report also appears to err in its use of school performance data. The authors of the report note that the data are derived by states and that the data between states are not strictly comparable to one another. However, unlike other researchers, Brightbeam has not attempted to standardize the results.

The Brightbeam report claims to control for other factors that might explain differences in gaps between progressive and conservative cities. These factors include student population size, and the racial demographics of the district. However, these are a far cry from all explanatory factors, and the report provides few details on how these controls were applied or why they were selected in the first place.

Here again, the Brightbeam analysis is partially limited by the fact that they have selected 12 of the “conservative” and “progressive” cities but do not include the full set of performance results for all the cities for which data exists.

By simply matching Brightbeam’s ‘conservatism’ scores to 2017-18 performance scores provided by The U.S. Department of Education, EdFacts, we are able to display the full range of performance scores in the 67 cities.

This shows that there is a great deal of variation in student performance and conservatism scores across city districts. For instance, chart 1 shows that there are large differences in the percentage of white and black students proficient in reading after sorting by the level of conservatism in the large cities. Clearly, there are wide racial gaps in the data. However, using the full data, it is much less clear how they relate to a city’s relative political conservatism.

Chart 2 shows the same data, but with test performance measured by white-black student racial gap ratios. We find that, generally speaking, as the conservatism score rises in a city, the racial gaps in student performance drops, but that the relationship between these variables is not strong, with a Pearson’s correlation coefficient of -0.32. There are progressive cities with small racial gaps (e.g. Detroit and New Orleans) and conservative cities with large racial gaps (e.g. Oklahoma City), weakening the correlation between ‘conservatism’ and racial gaps.

There is another important dynamic at play, as well. When breaking achievement gap data out into white and black student proficiency scores, we find that most of the difference between the cities’ racial gaps are due to decreasing white test performance scores in conservative-leaning cities, as shown on chart 3. There is a moderate correlation between white student reading performance and ‘conservatism’ in cities (R = -0.35), but practically no correlation at all when it comes to the performance of black students (R= 0.07).

We find quite similar results when comparing the 67 cities student performance in math, and when comparing the results between white and Hispanic students. Table 1 shows that all correlations between student proficiency and conservatism are weak to moderate (-0.28 to -0.36). Table 2 shows that there is virtually no relationship between black or Hispanic student performance and conservatism in cities. We also find this when limiting the data to the 24 cities reported and analyzed by Brightbeam, a finding not noted by the report’s authors.

Needless to say, this is not an overwhelming finding. Indeed, it suggests that conservative cities, far from being places to emulate, are simply providing uniformly mediocre education.

This is not to say that racial gaps in school performance are not a persistent issue, or that Minneapolis does not have large gaps. In fact, Minneapolis public schools have one of the largest white-black student performance gaps and socioeconomic segregation rates of school districts in the U.S., according to the Educational Opportunity Project at Stanford University. The Stanford project addressed this question with a full complement of standardized neighborhood, school and district data covering the entire nation. The researchers at Stanford find that the gaps can be linked to something that the Brightbeam authors neglect entirely: school segregation. The Stanford authors argue that to narrow such gaps, a reduction in segregation is necessary.

Unfortunately, the Brightbeam authors do not make any such policy recommendation. Their report simply suggests there is something magic in the water in conservatives cities, and use their report to laud the leaders of those cities while chiding progressive leaders. They offer no guidance for Minneapolis city leaders in the midst of important education decisions, such as the new Minneapolis Schools District Redesign plan.

We would argue that a more constructive stance is to perhaps disregard the superficial progressive-conservative divide altogether. Educational policy does not fit on a neat left-right spectrum, and some of the more progressive ideas about addressing segregation and racial gaps in schools come from more conservative-leaning areas, like Louisville, Kentucky, Raleigh, North Carolina and San Antonio, Texas. Putting the blame on progressive city leaders and residents also misses the mark, as segregation driven racial gaps are not limited to city boundaries. Instead they cut across city boundaries, metropolitan regions and political geographies. What is truly needed is a more serious and concerted effort by everyone to address them.