Firstly, what are conjoint experiments?

In political science, conjoint experiments have been used to examine preferences for presidential candidates, attitudes toward immigrants, voting behavior (Hainmueller et al., 2014; Breitenstein, 2019), preferences for COVID-19 vaccine allocation, and to evaluate different health care interventions (Johnson et al., 2013; Duch et al., 2021; Beckham et al., 2020).

Conjoint experiments are a set of research methods used to help understand how people make decisions when faced with various possible attributes by showing a respond a series of hypothetical profiles informed by a random combination of attribute levels. Respondents then evaluate these profiles, indicating their preferences, allowing us to make causal inferences about the overall popularity of individual attributes.

An example use-case: Soda Company

Imagine that we work for a soft drink company. We want to launch a new product, but we don’t know what type product to launch. We know that there are a number of attributes that are important for consumers.

Attributes: The features that we want to study.

For soft drinks, some attributes that we might be interested in studying could be: flavour, packaging type, sugar content, or container size.

Each of these attributes can have different levels, or variations in what is possible.

Levels: Intra-attribute variation.

In our soft drink example, some levels we might test for the flavour attribute are: cola, lemon-lime, extreme berry, or Rocket Fizz Pickle-Flavoured Soda.

Once we’ve set the attributes and levels that we are interested in, we can randomly select a level from each attribute to create a profile.

Profile: A configuration of randomly selected levels from each attribute.

In our soft drink example, a complete random profile might be:

Flavour: lemon-lime

Packaging Type: glass bottle

Sugar Content: 0 Sugar

Container Size: 355ml

We now have a few options for how we want to measure attitudes towards possible attributes. We could ask respondents to rate how much they would enjoy the soda, from a scale of 1-10, or we could display two randomly selected profiles to ask respondents for their preferred profile.

Setting Up Your Conjoint Experiment in Qualtrics Using Javascript

Now that we have a basic understanding of conjoint experiments, we can now setup the experiment in Qualtrics using JavaScript. Alternatively, Qualtrics offers an in-house tool for this, which is great, but it does cost extra. The method below is free, and relatively straightforward to implement.

1. In a separate text document, design your conjoint. What are you attributes? What are your levels? Clearly label your experiment so that you can copy text over for the next steps.

2. Somewhere before you intend to implement your conjoint, on an unrelated question, open the JavaScript tool in Qualtrics in the question dashboard to add the following JavaScript code. Replace the existing text with code to randomize the selected level within each attribute. You can use the following, simple, Fisher-Yates shuffle:

Qualtrics.SurveyEngine.addOnload(function()
{
function shuffle(array) {
		var m = array.length, t, i;
		while (m) {
			i = Math.floor(Math.random() * m--);
			t = array[m];
			array[m] = array[i];
			array[i] = t;
		}
		return array;
	}

3. We now need to inform the levels and attributes. I provide a single example below, but you will need to repeat this code for each of your levels:

flavor= shuffle(["cola", "lemon-lime", "extreme berry", "Rocket Fizz Pickle-Flavoured"])[0];

4. Once we have repeated that process for each of our attributes, we need to embed the attributes into our survey. You can use the following code:

Qualtrics.SurveyEngine.setEmbeddedData("flavor",flavor);

5. Save your JavaScript code.

6. Open the survey flow and add embedded variables for each of the attributes. Leave the defined value blank, as this value will be set by your JavaScript code.

7. Create a new question in Qualtrics which builds on the question type that you intend to analyze. For instance, if you are going to use a ‘slider’ type question for analysis (scale of 1-10), then create a slider question to embed the Javascript. Within the question text, embed the variable from your survey flow. This code, for each attribute, should appear as follows:

${e://Field/flavor}

8. Once your attributes are embedded into your survey question, perhaps in a table or as part of vignette, test your survey from before your embedded JavaScript code.

9. Congratulations, you’ve coded a conjoint experiment in Qualtrics!

Collecting and Analyzing Conjoint Experiment Data

Once you respondents have completed your survey, you can export data from Qualtrics for further analysis in statistical software. I recommend the cjoint package in R, but there are other options available in STATA or SPSS.

The central thesis in Segregation by Design is that segregation within, between, and outside American cities is “not simply the result of individual choices about where to live” (4), but rather the result of careful policy design by local policymakers. Trounstine returns to this argument throughout the book by providing overwhelming evidence of local policies which demonstrate that segregation is not only occurring, as evident by zoning laws and variation in public good provision, but that this segregation is intentional, and racist. This argument rests on the premises that “local governments control the location of negative and positive externalities (such as pollution-producing factories or public parks)… the functions that local governments provide are allocated” (141). In contrast with the federal or state-level contexts, Trounstine notes that “while it is difficult to deny particular house-holds access to sewer lines or a local public park, it is much easier to deny particular neighborhoods” (98), resulting in great differences across city boundaries or within neighbourhoods in a city.

Trounstine (16) provides a second, sub-thesis within their work: “segregated places are politically polarized places”. Polarization, in this context, can be partially explained by examining the effects of segregation on political preference: “residents who live in defended neighborhoods (segregated neighbourhoods, favouring a white majority)… are more likely to identify as Republicans and more likely to vote for Republican presidential candidates” (204). This, in turn, encourages local policymakers to make policies that these groups favour, further exacerbating polarization. This feedback cycle contributes to less collective investment in segregated communities, compared with all white or diverse communities. Trounstine (215) normatively opposes this outcome:

It is clear is that if we do nothing about this design, politics will continue to polarize, and inequality in wealth, education, safety, and well-being will continue to worsen. Much is at stake.

Trounstine notes that the research project began with an exploration of public goods allocation and “the relationship between land use control and inequality” (i.). Trounstine uses waterways (158; 206), “roads, policing, parks, sewers” as the main dependent variables throughout the book as a means by which to measure examine the “delivery of public goods to politically powerful constituents” within ‘white protected communities’. This work is related to Mancur Olson research on public goods, “the basic and most elementary goods or services provided by the government” (Olson 1965, 14). Evidently, measuring public good provisions is an appropriate approach to understanding tangible policy outcomes at the local level.

Trounstine utilizes a mixed-methods policy research design, drawing “on more than 100 years of quantitative and qualitative data from thousands of American cities to explore how local governments generate race and class segregation” (i). This mixed-methods approach is convincing, as the qualitative historical references provide the context necessary for thorough quantitative modelling.

Although Trounstine argues that the “roots of [segregation] are classic models of individual choice” (27), the book would benefit from greater exploration of individual-level, rational choice accounts of policymaking. Trounstine dismisses “theories reliant on individual choices [as] subject to instability in the absence of collective enforcement mechanisms” (28), however this modelling is accounted for in some of the rational choice policy literature. Schneider & Ingram argue that “public policy almost always attempts to get people to do things that they might not otherwise do” (1990, 513) as there are policy tools which do not require collective enforcement mechanisms to arrive at outcomes, such as incentive tools (Ibid., 515-516) or capacity tools (Ibid., 517-518).

Trounstine’s Segregation by Design is excellent public policy research. If policy scholars are to adopt Lasswell’s conception of the policy profession as the pursuit of “knowledge that will heal the sick and improve the position of the socially deprived in every category” (1970, 14), then Trounstine’s book is an especially valuable contribution.

Works Cited

Lasswell, Harold. “The Emerging Conception of the Policy Sciences.” Policy Sciences 1, 1970: 3-14.

Olson, Mancur. The Logic of Collective Action: Public Goods and the Theory of Groups. Harvard University Press, 1965.

Schneider, Anne, and Helen Ingram. “Behavioral Assumptions of Policy Tools.” Journal of Politics 52.2, 1990: 510-529.

Trounstine, Jessica. Segregation by Design: Local Politics and Inequality in American Cities. New York: Cambridge University Press, 2018.

Dense urban places in Canada, like Toronto, Vancouver, and Calgary, have high intergenerational economic mobility, whereas regions in rural Manitoba, coastal British Columbia, and Newfoundland are ‘mobility traps,’ places where individuals are unlikely to move up the income ladder (Sanders 2017). Rural communities are not always less economically mobile than urban places, rather “higher mobility communities… tend to be areas with lower poverty, less income inequality, and a higher share of immigrants” (Corak 2017, 2), factors more closely associated with urban areas. Mobility traps are “an important issue in many rich countries” (Corak 2017, 1) because of their impact on income inequality and social mobility.

Some cities are undergoing a ‘new urban crisis,’ defined by “the back-to-the-city movement of the affluent and the educated—accompanied by rising inequality, deepening economic segregation, and increasingly unaffordable housing” (Florida 2017). In Toronto and Vancouver, municipal governments have responded to increasingly unaffordable housing prices by implementing policies designed to discourage wealthy investors from speculating on housing prices. Florida (2017) summarizes this urban policy challenge as:

the central contradiction that stands at the heart of today’s urbanized form of knowledge capitalism writ large. The very same clustering of talent, business, and economic capability in large, dense, knowledge-based places also carves deep divisions into our cities and society.

Attracting talent to mobility trap regions can help break the cycle of limited intergenerational economic mobility and the ensuing divisions. Policies which incentivize skilled workers and businesses to relocate to mobility trap areas, such as enhanced government services, reliable internet, tax breaks, improved infrastructure, and enhanced educational opportunities, may foster long-term growth. By transforming mobility trap regions into hubs of talent and opportunity, governments can promote more equitable economic outcomes across Canada.

Works Cited

Corak, Miles. “Divided Landscapes of Economic Opportunity: The Canadian Geography of Intergenerational Income Mobility.” Human Capital and Economic Opportunity Working Group. Working Papers 2017-043, (2017).

Florida, Richard. “Mapping the New Urban Crisis.” City Lab. April 2017. Accessible online: https://www.citylab.com/equity/2017/04/new-urban-crisis-index/521037/

Sanders, Doug & Tom Cardoso. “A tale of Two Canadas: Where you Grew Up Affects Your Income in Adulthood.” Globe and Mail. June 2017.

Facing a devastating recession in 2008 caused in part by the collapse of the American economy, McGuinty sought out the advice of economic experts, both within the government and in the private sector: “McGuinty was receiving daily briefings on the severity of the economic situation in Ontario” (Lesch 2018, 109). TD Bank released a report warning that an economic collapse in Ontario was imminent unless dramatic steps were taken, identifying “tax reform as a sensible thing that the province could do” (Lesch 2018, 111) to help counteract the effects of a recession. The TD Bank report, compiled by private sector experts, transformed McGuinty’s beliefs so publicly that after the HST was implemented, Jim Flaherty, then President of the Treasury Board, confided with friends that he believed that McGuinty’s “shocking policy shift was because ‘Bay St. had gotten to McGuinty and convinced him’” (Benzie 2010).

The implementation of the HST was as much McGuinty’s idea as it was the idea of experts who defined the solution, problem, and potential punishments for failure. Experts instilled in McGuinty the idea that a disaster was impending and that something significant had to be done. Experts also provided the solution. The TD Bank report, prepared “was critical to McGuinty’s change of heart” (Benzie 2010).

Works Cited

Benzie, Robert. 2010. “Why Ottawa and Queen’s Park Embraced the HST.” Toronto Star.

Lesch, M., 2018. Playing with fiscal fire: the politics of consumption tax reform. University of Toronto (Canada).