Background

Canada has maintained an agricultural policy that manages the supply of dairy, poultry and eggs since the early 1970s. Supply management is a long-standing and little-changed public policy that is often criticized for its adverse effects on consumers by academics, think tanks (Busby and Schwanen, Reference Busby and Schwanen2013; Sarlo et al., Reference Sarlo, Martin and Lee2012) and commentators in widely read and popular outlets such as the Globe and Mail (2018) and the National Post (Reference Mullinix, Leeper, Druckman and Freese2018). Yet Canada's political parties, with the exception of the People's Party of Canada (PPC), remain steadfastly committed to ensuring that the supply management regime remains intact.

Although Pierre Elliott Trudeau's Liberal government officially implemented the policy in 1970, the federal government had been intervening in the dairy market since the 1950s (Goldfarb, Reference Goldfarb2009: 3; Hall Findlay, Reference Haider-Markel and Joslyn2012: 5). In a passing word on the 1957 election, Peter C. Newman (Reference Newman1973) noted that the St. Laurent Liberals were not above intervening in the dairy market—in fact proposing direct government intervention in the dairy market by banning the importation of cheddar cheese and providing a price floor for poultry. These measures, however, were “interpreted as last-minute bribes” (Newman, Reference Newman1973: 55). Federal governments since then have not shied away from intervening in the dairy market and have remained steadfastly committed to Canada's supply-management regime.

Supply management has remained a stable agricultural policy in Canada for over 40 years. Skogstad (Reference Skogstad2008) argues that, resulting from trade negotiations at the beginning of the twenty-first century, a “second generation” of supply management has emerged with modifications concerning pricing policy instruments, quota allocation and the prices received by processors. Nonetheless, the current policy still maintains a “striking resemblance to the first generation” (Skogstad, Reference Skogstad2008: 175). Supply management's resilience is a result of weak internationalization in supply-managed sectors, as well as its economic and, perhaps more importantly, political viability with political elites concerned about upsetting a powerful lobby and voters in critical ridings in Ontario and Quebec (Skogstad, Reference Skogstad2008).

At its core, the policy rests on three pillars: (1) prices are set based on a cost-of-production formula, (2) production is limited to an internally derived supply-and-demand estimate and (3) border measures are enacted to prohibit foreign competition (Barichello et al., Reference Barichello, Cranfield and Meilke2009: 204). In theory, price and production stability should be beneficial for both producers and consumers alike. Consumers are afforded locally sourced milk that is heavily controlled and regulated by a central authority to ensure its safety. Stable prices benefit both producers and consumers because it allows the former to receive a stable income that was worth their labour while the latter would not be faced with price shocks due to sudden increases in the price of milk.

There are costs to this stability, however. A general consensus exists among economists and mainstream policy analysts on the harmful effects of supply management. Their work has found that (1) smaller provincial dairy boards free-ride off the decision of boards of larger provinces (for example, Ontario and Quebec) to increase their lobbying expenditures (Baylis and Furtan, Reference Baylis and Furtan2003); (2) changes in the milk production quota scheme in Ontario led to widespread inefficiency in the quota (production) market (Cairns and Meilke, Reference Cairns and Meilke2012); (3) the prices of supply-managed goods in Canada are higher than in the United States (Hall Findlay, Reference Hall Findlay2012: 5); and (4) supply management has had a regressive effect on low-income earners—particularly those with children (Cardwell et al., Reference Cardwell, Lawley and Xian2015; see Doyon et al., Reference Doyon, Bergeron and Tamini2018, for a response).

That being said, it is important to acknowledge the reality that subsidies and protectionism are the norm in agricultural sectors cross-nationally. The Organisation for Economic Co-operation and Development's (OECD) Trade Directorate (for Food, Agriculture and Fisheries) contends that “milk producers, in virtually every OECD country and in many non-member economies, benefit from government interventions that boost the prices they receive for their raw milk production” and that “government support and protection for milk producers is also more widespread than for any of the other commodities for which the OECD calculates producer support estimates” (OECD, 2005: 97).

Canada's system of supply management differs mostly in the type of support it provides compared to other countries. Proponents of supply management have often argued that the policy is unsubsidized, unlike dairy production in the United States (Alberta Milk, n.d.; British Columbia Dairy Association, n.d.). By unsubsidized, proponents mean that there are no out-of-pocket government expenses in normal times. Exogenous shocks to the existing system, such as free trade agreements, can move supply-managed industries into abnormal times, necessitating direct government aid. Citing negotiations in the Comprehensive Economic and Trade Agreement (CETA) and the Comprehensive and Progressive Agreement for Trans-Pacific Partnership (CPTPP), the Canadian government agreed to compensate supply-managed industries affected by CETA and the CPTPP to the tune of $1.75 billion over eight years for dairy farmers and $691 million for poultry farmers over 10 years (Government of Canada, 2020)Footnote ¹ in the 2019 Canadian federal budget.

Nevertheless, for much of supply management's history it was de jure unsubsidized but de facto resulted in a consumer-to-producer transfer. But as pressures to liberalize trade mount, direct government assistance is likely to rise in the future.

Ideology and Support for Government Regulation

At the moment, little is known about the determinants of public support for supply management. However, Angus Reid's finding that opposition is higher among Conservative party voters is potentially instructive of a relationship between right-leaning ideology and opposition to supply management.

As an ideology, conservatism is hard to define. O'Sullivan (Reference O'Sullivan, Freeden and Stears2013: 293) notes that modern conservative ideology is often reduced to an “anti-ideology inspired by self-interest and fear of change” but that conservatism itself is malleable and can largely be divided into four schools of thought: reactionary, radical, moderate, and the New Right. The extant research on policy preferences has, to a degree, confirmed O'Sullivan's argument that conservatism is perhaps malleable but nevertheless committed to the defence of limited politics.

The malleability of conservatism has been expanded upon in the literature, which has largely found that conservatives are particularly capable of sectionalizing and differentiating their priorities (Cochrane, Reference Cochrane2010, Reference Cochrane, Farney and Rayside2013). So where O'Sullivan identifies four schools of conservative thought, Zumbrunnen and Gangl (Reference Zumbrunnen and Gangl2008) argue that there exist three types of conservatives: the market conservative, the limited government conservative and the cultural conservative. They defined the limited government variant as one that recognizes government is “necessary for performing certain functions [but] its economic, social, and moral influence must be kept within very definite limits” (2008: 207). Similarly, Thorisdottir et al. (2007: 179) define economic conservatism as involving “an ideological commitment to capitalism, private enterprise, and the value of competition among individuals and corporations in the context of a free market.” Economic conservatism, then, prioritizes a free-market, private enterprise and competition. It is distinct from cultural conservatism, defined as the “rejection of qualitative social change as well as a belief in the importance of religion, traditional family arrangements, and conventional gender roles” (Thorisdottir et al., Reference Thorisdottir, Jost, Laiviatan and Shrout2007: 179). The differentiation between social and economic conservatism is important given that both forms significantly affect ideology but economic conservatism appears to have no impact on one's stances toward social issues (Zumbrunnen and Gangl, Reference Zumbrunnen and Gangl2008; see also Johnson and Tamney, Reference Johnson and Tamney2001).

The connection between economic conservatism and the preference for limited politics is confirmed by Bartels (Reference Bartels2018: 1485), who finds that the likelihood of conservative self-identification increases with support for limited government and the belief that the government in Washington has gotten too powerful. Furthermore, survey items that most strongly reflect support for limited government include opposition to the welfare state and egalitarian economic policy including but not limited to reducing income inequality (see also Aspelund et al., Reference Aspelund, Lindeman and Verkasalo2013; Jost et al., Reference Jost, Glaser, Kruglanski and Sulloway2003; Thorisdottir et al., Reference Thorisdottir, Jost, Laiviatan and Shrout2007), the provision of publicly funded healthcare, increasing the minimum wage and progressive taxation (Bartels, Reference Bartels2018: 1484).

This economic policy-focused conservatism that favours less government involvement has also been prominent in Canada. Farney and Rayside (Reference Farney, Rayside, Farney and Rayside2013) and Patten (Reference Patten, Farney and Rayside2013) note that ideological shifts within the Canadian Conservative party have led the decline of Red Toryism—a belief that stresses communitarianism and social order but allows for considerable government intervention—in favour of neoliberalism. Patten (Reference Patten, Farney and Rayside2013: 61) further argues that neoliberalism has become a core ideology within the modern Conservative party, leading to the embrace of the free-market, the desire to roll back the state and the privileging of individualism and self-reliance.

Separate from social conservatism, economic conservativism is associated with support for free markets, limited government and competition. From this perspective, then, the core tenets of economic conservatism are antithetical to supply management's structures. The parochial nature of the policy—where production and prices are set by a marketing board and competition and imports are strictly limited—should result in less support for the policy among economic conservatives, though we have no such expectations for cultural or social conservatives:

Framing Supply Management

In addition to evaluating the correlates of support for supply management, we test how responsive this support is to framing. Framing has been defined as “the process by which people develop a particular conceptualization of an issue or reorient their thinking about an issue” (Chong and Druckman, Reference Chong and Druckman2007: 104) and as “alternative definitions, constructions, or depictions of a policy problem” (Nelson and Oxley, Reference Nelson and Oxley1999: 1041). Frames, then, are how the issue is presented to the recipient. They “prime” individuals to conceive of the topic through the lens of the frame (Nelson and Oxley, Reference Nelson and Oxley1999: 1043).

Framing effects have been defined as when a “speaker's emphasis on a subset of potentially relevant considerations causes individuals to focus on these considerations when constructing their opinion” (Druckman, Reference Druckman2001: 1042) and when often small “changes in the presentation of an issue or an event produce (sometimes large) changes of opinion” (Chong and Druckman, Reference Chong and Druckman2007: 104). A successful frame will therefore lead to a framing effect that causes the individual to focus on the frame's considerations (Druckman, Reference Druckman2001). These framing effects work by making new beliefs available, accessible or applicable (Chong and Druckman, Reference Chong and Druckman2007: 111).

Studies of framing and cueing effects have increased notably in recent years in Canada. Doberstein and Smith (Reference Doberstein and Smith2019) find that deservingness cues related to mental health increase support for government aid to the homeless across the ideological divide. Harell et al. (Reference Harell, Soroka, Iyengar and Valentino2012) use a conjoint experiment to show that cues signalling that individual immigrants are high in skill level increase support for granting their citizenship. Gravelle (Reference Gravelle2018) finds that Trump endorsement cues undermine Canadian public support for related trade policy, though not for energy policy. These works, however, focus on cues rather than emphasis frames.

Other work has examined the role of emphasis frames. Lachapelle et al. (Reference Lachapelle, Montpetit and Gauvin2014) show that expert framing can have impacts on source credibility, which comes under strain when experts frame issues in ways that are a threat to people's predispositions. They, however, do not look at its effects on policy attitudes. Closest to our focus here, Gravelle (Reference Gravelle2020) finds that Canadians’ views toward foreign trade are subject to framing effects in a survey experiment. Respondents were less likely to support closer economic ties with the United States when respondents were provided information about the size of Canada–US trade. Presenting respondents with information highlighting Mexico's and China's human rights abuses reduced their desire to seek closer economic ties with said states as well.

However, no research to date has looked at framing effects on agricultural policy in Canada. And this area of policy is likely ripe for strong framing effects because of its lack of salience. Not many people are likely to have firmly anchored beliefs related to supply management or agriculture policy more generally. Research has typically found that framing effects are weaker when people have stronger prior beliefs about a policy (Druckman and Nelson, Reference Druckman and Nelson2003; Haider-Markel and Joslyn, Reference Haider-Markel and Joslyn2001), though with a caveat that high levels of knowledge increase frame accessibility, which will be discussed more below.

The limited work that has been conducted so far on agriculture policy framing supports this contention. Jensen and Shin (Reference Jensen and Shin2014) find that public attitudes toward agricultural subsidies are incredibly malleable. When respondents were presented with a vignette that frames the current American subsidization regime as less generous than in other countries, an additional 12 per cent voiced support for increasing agricultural subsidies (Jensen and Shin, Reference Jensen and Shin2014: 306). To account for the success of their frames, the authors argue that “individuals’ lack of knowledge of agriculture policy provide tremendous opportunities to frame agriculture policy to generate support for farmers” (306). This lack of knowledge further allows policy makers to adopt the frame that is most conducive to public support, thus attempting to pull public opinion towards the preferences of the policy maker.

We focus on three frames for the purposes of this article. The first is the pro-farmer frame. All major political parties in the 2019 Canadian election maintained a commitment to supply management and demonstrated in their platforms their support for farmers. The LPC argued that a Liberal government will “continue to defend supply management” and will continue to implement support for dairy farmers affected by trade agreements (LPC, 2019: 25). The Conservatives promised to “defend Canada's system of supply management” in order to “defend our farmers’ way of life” (CPC, 2019: 31). The NDP promised to protect supply management given that it “protects our family farms, rural communities, and hundreds of thousands of jobs” (NDP, 2019: 40). The Bloc Québécois forcefully defended the supply management regime and promised to propose a bill in the House of Commons to guarantee compensation promised for dairy farmers (as a result of trade agreements), as well as compensation for poultry and egg farmers (Bloc Québécois, 2019: 15).

The major parties are wedded to the notion that supply management is crucial for ensuring a fair return for farmers. Because dairy and poultry farmers’ incomes are volatile in a globalized world, these incomes must be protected by the federal government. As a result, parties are themselves framing supply management as a necessity to preserve the Canadian farm and the farmers’ way of life. Yet the family farm frame employed by political parties is not entirely accurate. In fact, the number of dairy farms with shipments of milk has declined in every subsequent year from 1967 (the first year for which we have data) to 2019, dropping from 174,139 farms in 1967 to 10,371 in 2019 under a supply management regime (Canadian Dairy Commission, 2020a).

Nevertheless, pro-farmer frames are prevalent in supply management discourse, and such sentiment is widely available, at least in the American public, especially in regards to family farms. Ellison et al. (Reference Ellison, Lusk and Briggeman2010: 352) find that people tend to support smaller farms over larger farms, with a large majority in favour of increasing farm subsidies for small farms and 71 per cent willing to reduce subsidies provided to large farms.Footnote ² Similarly, Variyam et al. (Reference Variyam, Jordan and Epperson1990: 259) find that 57 per cent agree with the statement that “government should have a special policy to ensure that family farms survive,” while Variyam and Jordan (Reference Variyam and Jordan1991) find that positive views toward the family farm are linked with support for farm subsidies.

The second frame we consider is the free-market frame. This frame emerges from existing research demonstrating that supply-managed goods are often more expensive in Canada than in America (Cardwell et al., Reference Cardwell, Lawley and Xian2015) and that, consequently, Canadian consumers pay more for dairy and poultry products. The free-market frame is often invoked by opponents of supply management—and particularly by opinion columnists in the National Post, who argue that the current agricultural regime limits Canada's trade capacity and is an unfair burden on consumers (Dorosh, Reference Dorosh2018; Ivison, Reference Ivison2013, Reference Ivison2015; Lau, Reference Lau2020; Morgan, Reference Morgan2018; National Post, 2018). Consequently, eliminating the policy would allow Canadian dairy farmers to seek new external markets and would provide internal competition to Canadian consumers, thereby lowering the price of supply-managed goods. Thus, consumers would see considerable gains in a system where price and production are set by market forces.

Variyam et al. (Reference Variyam, Jordan and Epperson1990) show that there is pocketbook-based opposition toward agricultural subsidies in the United States. They find that only 21 per cent agree (59 per cent disagree) with the statement that “most consumers would be willing to have food prices raised to help preserve the family farm”; only 35 per cent agree (39 per cent disagree) that “family farms should be supported even if it means higher food prices”; and a plurality of respondents (44 per cent) agree that “farmers should compete in a free market without government support” (Variyam et al., Reference Variyam, Jordan and Epperson1990: 259). The free-market frame, then, asks respondents to consider supply management from the perspective of their pocketbook.

The last frame we consider is a social inequity argument that emphasizes the asymmetrical effect of higher prices for staple goods for low-income and single-parent households. Indeed, Cardwell et al. (Reference Cardwell, Lawley and Xian2015) argue that the average household faces a financial burden of $444 per year as a result of supply management. More to the point, there is an annual burden of $466 and $585 imposed on low-income households and households with children, respectively. The $466 burden on low-income households equates to an implicit tax of 2.29 per cent, which is roughly five times larger than the burden placed upon high-income households (see Doyon et al., Reference Doyon, Bergeron and Tamini2018, for a direct response).

Our expectations are simply that the farmer frame should heighten support for supply management, while the free-market and inequity frames should reduce support. We test these expectations using a framing experiment included in wave 7 of a study of the 2019 study of the Canadian election by the Digital Democracy Project (DDP). We randomly assign respondents into four groups: a control condition with no frame, and three conditions that match our three frames. We expect the following:

Framing effects are likely to be heterogeneous across several dimensions. First, they are moderated by individual predispositions where people with strong values are less amenable to contradicting frames (Chong and Druckman, Reference Chong and Druckman2007). Indeed, these values may manifest as “value frames” that are situated “between a value and an issue [and] carr[y] an evaluative implication: It presents one position on an issue as being right (and others being wrong) by linking that position to a specific core value” (Brewer, Reference Brewer2001: 46). Ultimately, we expect the effectiveness of the frame to be determined by its alignment with an individual's values. Free-marketers should find the free-market frame more persuasive, while economic progressives should be persuaded by the important social inequities fostered by supply management:

Second, framing effects should be moderated by knowledge. The direction of this effect, however, is contested. On the one hand, having more knowledge about a particular issue “increases the likelihood that the consideration emphasized in a frame will be available or comprehensible to the individual” (Chong and Druckman, Reference Chong and Druckman2007: 112). On the other hand, having knowledge about an issue may inoculate respondents from the influence of a frame (Chong and Druckman, Reference Chong and Druckman2007). Perhaps not surprisingly then, empirical studies have found evidence in both directions (see Bullock and Vedlitz, Reference Bullock and Vedlitz2017; Haider-Markel and Joslyn, Reference Haider-Markel and Joslyn2001; Kinder and Sanders, Reference Kinder and Sanders1990; Nelson et al., Reference Nelson, Clawson and Oxley1997; Slothuus, Reference Slothuus2006), though most evidence points to a reinforcing effect of political knowledge on frame acceptance (Bullock and Vedlitz, Reference Bullock and Vedlitz2017). Druckman and Nelson (Reference Druckman and Nelson2003) argue that negative results in the literature are likely a result of a conflation of prior beliefs with knowledge. Strong priors inoculate respondents from framing effects, while knowledge itself allows frames to be accessible and comprehensible. Our expectation is that framing effects will be stronger among those with high levels of policy knowledge, at least when controlling for prior beliefs.

Data and Method

Our data come from wave 7 of the DDP's study of the 2019 Canadian election, fielded from October 4 to October 13, 2019 (N = 1,545). The sample was collected by Qualtrics, a company that provides access to online panel data, in addition to its better-known survey platform. Qualtrics draws potential respondents from a number of different panels maintained by other sample providers. Specific panels were identified to the DDP team with an embedded data field that contained the respondent's panel identification.

Online non-probability samples are increasingly used by social scientists as a means of reliable, low-cost data collection. These scholars have been bolstered by a growing literature that has shown that while online non-probability samples provide somewhat divergent point estimates from probability-based modes (though not always, see Ansolabehere and Schaffner, Reference Ansolabehere and Schaffner2014), there is limited evidence of mode-based differences in the estimation of bivariate or multivariate relationships (Bytzek and Bieber, Reference Bytzek and Bieber2016; Pasek, Reference Pasek2016; Sanders et al., Reference Sanders, Clarke, Stewart and Whiteley2007; Stephenson and Crête, Reference Stephenson and Crête2011; Yeager et al., Reference Yeager, Krosnick, Chang, Javitz, Levendusky, Simpser and Wang2011). In the Canadian context, Breton et al. (Reference Breton, Cutler, Lachance and Mierke-Zatwarnicki2017) compare the online opt-in component of the 2011 Canadian Election Study with the traditional phone survey and find the latter less likely to match population quantities and more likely to suffer from self-selection and social desirability bias.Footnote ³ Even more importantly, sample average treatment effects and conditional average treatment effects recovered from survey experiments are comparable across modes (Mullinix et al., Reference Mullinix, Leeper, Druckman and Freese2015; Coppock et al., Reference Coppock, Leeper, Leeper and Mullunix2018).

The DDP set quotas on gender (male, female), age (18–34, 35–54, 55+), region (Atlantic, Quebec, Ontario, West) and language (English, French) to match population benchmarks provided by the 2016 census. Our data are further weighted for our observational analyses within region by age and gender to match 2016 Canadian census population benchmarks. Sample characteristics can be found in Table S1 of the supplementary materials. They compare favourably to these benchmarks.

Outcome measure and treatment conditions

We provide respondents with a primer on supply management and then ask them to provide their level of support for supply management (strongly oppose; somewhat oppose; neither oppose, nor support; somewhat support; strongly support; don't know). Approximately 25 per cent of respondents opposed supply management at some level in the control group, compared to 22 per cent who support it; 53 per cent of respondents either took a neutral position or indicated they didn't know. The text of the prime can be found in the online supplement. We code “don't know” as missing to create a 0–1 scale.

We randomly assign respondents into four groups. In the control condition, respondents received only the prime. The other respondents were exposed to one of three frames. The first was a message consistent with the dairy lobby and Canada's major political parties, arguing that supply management stabilizes prices and provides farmers with a fair return (farmer frame). The second was consistent with the position of critics who argue that supply management privileges some farmers over others who are not under supply management, while increasing prices on consumers (free-market frame). The third emphasizes the socially regressive nature of supply management, in that its burden falls most strongly on less fortunate Canadians (social inequity frame). Our treatment conditions are described in Table 1, along with example text. The full text of the treatment conditions can be found in the online supplement.Footnote ⁴ We expect the farmer frame to increase support for supply management (H_2A), while the free-market and social inequity frames should reduce support for supply management (H_2B, H_2C).

Table 1. Treatment Conditions

A few elements of the design are worth highlighting. All respondents receive some information on the nature of supply management in the control condition. This means that our estimate of support is taken after all respondents are given some information on the nature of the program. Rather than an estimate of public opinion as currently constituted, our estimate represents what public opinion would look like if people were given the information we provide. There are a few reasons why we chose this approach.

First, supply management is a low-salience issue. Many Canadians know nothing about the nature of the program. It is unclear whether there is much value in examining public opinion in such a state. Opinion is likely to change when people are exposed to a bare minimum level of information if elite debate on the issue was to take hold (Key, Reference Key1961), and politicians may seek to anticipate the public's reaction to this change in salience (Merkley and Owen, Reference Merkley and Owen2020). An estimate of public support for supply management given a bare minimum understanding of the policy is important, in and of itself.

Second, and perhaps more importantly, we wanted to preserve a high degree of experimental control. Providing respondents with frames would, in effect, also provide them with some factual information about supply management (that it is an agricultural-support program that maintains a system of quotas and tariffs to guarantee a return to farmers). In the absence of factual information about supply management in the control condition, treatment effects we observe would reflect a bundling of effects from factual information about the program and the frame itself. That being said, future research should explicitly examine the effect of providing factual information related to supply management on policy support.

Another feature of this design is that we are testing the effects of frames independently of one another. We do this in order to isolate the effect of each of these frames compared to the control condition. An alternative approach would be to cross these conditions in a 2 × 2 × 2 design. As Chong and Druckman (Reference Chong and Druckman2007) note, frames often exist in competition to each other in political discourse. Understanding the degree to which these effects hold in a competitive environment is thus important.

We did not implement such a design for a number of reasons. First, we were not able to measure the strength of frames in advance of the implementation of our survey module, so we use three frames that are most common in this policy debate and infer frame strength from the magnitude and significance of the persuasion effects. Scholarship has found that strong frames dominate weak frames, and strong frames tend to cancel out the effects of other strong frames. In the absence of prior measure of frame strength, we had limited expectations as to the interactive effects we would observe with a 2 × 2 × 2 factorial design.

Second, it requires a lot of statistical power to precisely estimate heterogeneous effects with a three-way interaction. Our principle interest for this study is to estimate the effects of treatment compared to control and how these effects are moderated by observational factors such as policy knowledge and political predispositions. Future work should examine how the framing effects we identify change in light of competition from one another.

One final point on the design is that we do not examine moderation effects by the type of messenger, such as by political elites or economists. It is important to note that framing effects can be conditional on perceived credibility and trust in the messenger. For instance, Druckman (Reference Druckman2001) finds that respondents exposed to frames were more likely to alter their opinion toward a KKK rally with more credible messengers. We do not test the effects of source cues here, or the interaction between source and message. This is an important limitation of the external validity of this study. Aggregate public opinion on this question may be shaped more by the credibility of the sources on each side of the debate (that is, political parties vs. supply management's critics) rather than the particular message they are using. Nonetheless, testing the effectiveness of important frames on this topic is a reasonable first step of inquiry.

Observational model

We estimate a model using ordinary least squares (OLS) regression that regresses supply management support on economic and social conservatism, our controls (X) and a set of binary variables indicating experimental treatment (T).

$$supply\;management\;support = \alpha + \beta _1economic\;conservatism + \beta _2social\;conservatism + \beta X + \beta T + \varepsilon $$

We expect β ₁to be negative and significant to support H₁. Breusch-Pagan tests allow us to rule out the presence of heteroscedasticity, so our models are estimated with OLS standard errors.

Experimental models

To evaluate the effects of our frames on support for supply management, we estimate an OLS regression where supply management support is regressed on each of our treatment conditions. There are no covariates in this estimation:

$$supply\;management\;support = \alpha + \beta _1farmer + \beta _2freemarket + \beta _3inequity + \varepsilon $$

We expect β ₁ to be positive and significant to support H_2A, while β ₂ and β ₃ should be negative and significant to support H_2B and H_2C, respectively.

For our experimental moderation analyses, we first estimate a pair of models where we interact economic conservatism with either a binary variable indicating a free-market frame (baseline = control, model 1) or a social inequity frame (baseline = control, model 2) to test H_3A and H_3B.

$$supply\;management\;support = \alpha + \beta _1\,frame + \beta _2economic\;conservatism + \beta _3\,frame\,\ast \,conservatism + \varepsilon $$

We illustrate these results with marginal effects plots.

Then we estimate a trio of models where we interact policy knowledge with a binary indicator of each treatment condition independently (baseline = control condition). Policy knowledge, however, is observational. It may be highly correlated with other factors that moderate the treatment effect. We need to control for these factors in order to isolate the moderation effect of policy knowledge (Kam and Trussler, Reference Kam and Trussler2017). In particular, strong priors may dampen treatment effects while being correlated with policy knowledge (Druckman and Nelson, Reference Druckman and Nelson2003). Druckman and Nelson (Reference Druckman and Nelson2003) use the “need to evaluate” index as a means to account for these priors. This election study lacks such an index, so we instead use political interest, partisan strength and ideological extremity to absorb the moderating effects of strong priors (X).

Our measure of policy knowledge is based on a question asking respondents to identify the commodities that fall under the umbrella of supply management from a list. We construct a variable classifying respondents as low knowledge if they failed to identify any such commodity from the list and as high knowledge if they correctly identified all three commodities. We score them as medium knowledge if they correctly identified only some of these commodities. Twenty-six per cent of our respondents correctly identified the three commodities under supply management from the list.Footnote ⁷ We enter dichotomous variables in our model for medium- and high-knowledge respondents to account for potential nonlinear relationships, as shown below. For each model, we restrict ourselves to respondents in either the control or respective treatment condition. This allows us to directly test H₄ for each frame.

$$supply\;management\;support = \alpha + \beta _1\,frame + \beta _2midknow + \beta _3highknow + \beta _4\,frame\,\ast\, midknow + \beta _5\,frame\,\ast\, highknow + \beta X + \beta frame\,\ast\, X + \varepsilon $$

We again illustrate these results with marginal effects plots. For our experimental estimates, we use HC2-robust standard errors to account for heteroscedasticity. The full estimates can be found in the supplementary materials.