《社会保障概论》课程教学资源（课后阅读资料）APSR_Volume112. Issue 4.Nov 2018_On_the_Limits_of_Officials_Ability_to_Change_Citizens_Priorities_A_Field_Experiment_in_Local_Politic

American Political Science Review (2018)112.4.860-873 doi:10.1017/S0003055418000473 American Political Science Association 2018 On the Limits of Officials'Ability to Change Citizens'Priorities: A Field Experiment in Local Politics DANIEL M.BUTLER University of California,San Diego HANS J.G.HASSELL Florida State University e test whether politicians'communications shape their supporters'policy priorities by conduct- ing a field experiment in collaboration with several local elected officials.In the experiment, the officials sent out email messages to the constituents on their distribution lists.Half the con- stituents received messages where the official advocated for the priority of a given issue,while the other half received a placebo email.We surveyed the constituents one to two months before the message went out and again the week after the official sent the message.The experiment shows that politicians did not change citizens'priorities in the desired direction.Moreover,citizens who received a message where the official indicated the issue was a priority were not more likely to act when invited to sign a petition on the issue.Elected officials'ability to shape the priorities of the politically active citizens with whom they regularly communicate is limited and can even be self-defeating. nformation is a tool that has the potential to affect (Oliver and Ha 2007:Rugh and Trounstine 2011: opinion and mobilize citizens to action (Foos and Tausanovitch and Warshaw 2014).At the local level, de Rooij 2017a).We test whether local elected offi- many of the political questions are not whether an cials can change the political agendas and issue priori- action should be taken,but whether certain issues ties of the people with whom they regularly communi- should be prioritized (Erie,Kogan,and MacKenzie cate. 2011;Franklin and Ebdon 2004;Guo and Neshkova We study the ability of government officials to in- 2013).For example,given a list of infrastructure fluence citizens'issue priorities because governments projects,many constituents would support action have limited resources.Governments cannot deal with on all the necessary improvements provided there all issues at the same time:they must identify which were sufficient resources.However.given a limited issues will receive the highest priority.The ability to budget,and limited time resources,constituents might affect citizens'priorities can thus have significant im- prioritize certain action items over others. plications for the allocation of political power.As Studies of agenda setting in other contexts have Schattschneider(1960)noted,the ability to define po- found that political elites can drive the political agenda litical priorities and the alternatives changes the na- and priorities of the general public.However,these ture of political conflict and is the"prime instrument studies have almost exclusively focused on the ability of power"(73).Changing citizen priorities and the of nonelected political elites to change citizens'prior- agenda changes the political lines of division within so- ities and the effect that these changes have on poli- ciety and can reallocate power among political actors. tics(e.g.,Iyengar and Kinder 1987;King,Schneer,and Citizens'priorities and political agendas are es- White 2017;McComb and Shaw 1972).Scholars have 115.5010 pecially important to understanding local politics not studied elected officials'ability to shape their con- stituents'priorities. There are both arguments that public officials can Daniel M.Butler is an Associate Professor,University of California, shape the political agenda through their communica- San Diego,Social Sciences Building,9500 Gilman Drive #0521,La tions with constituents and reasons to believe that they Jolla.CA 92093-0521 (daniel.butler@gmail.com). Hans J.G.Hassell is an Assistant Professor,Florida State Univer. cannot.On one hand,there are reasons that commu- sity,Department of Political Science,531 Bellamy Building,Tallahas- nication by elected officials might shape citizens'pri- see,FL 32306 (hans.hassell@gmail.com). orities.Previous studies have found that other political An earlier version of the paper was presented at the Urban Po- elites such as the media and interest groups shape the litical Economy Conference at Vanderbilt University in March 2017 general public's priorities (Iyengar and Kinder 1987; We thank Marc Meredith and other conference participants for com- ments.Replication files are available at the American Political Sci. King,Schneer,and White 2017;McComb and Shaw ence Review Dataverse:https://doi.org/10.7910/DVN/KTCXTE 1972).What these actors talk about and highlight of- ten become the general public'priorities and set the Received:May 15,2017:revised:January 16,2018:accepted:July 5. 2018.First published online:September 4,2018. agenda for political debate.Likewise,communication by other political elites,particularly the media or in- 1 Schattschneider(1960)identifies two ways of changing the balance terest groups,has a strong effect on citizens'priorities of power:First,by defining the political priorities and,second,by enlarging or reducing the scope of conflict and the participants in and can propel citizens to action (Gerber,Karlan,and that conflict.Sarah Anzia's(2014)and Justin de Benedictis-Kessner's Bergan 2009).There is also some evidence that elite (2018)work on changing the timing of elections to correspond with communication can sometimes shape public opinion national elections are examples of the few studies we are aware of (e.g.,Broockman and Butler 2016).2 that examine the ability of local public officials to enlarge or reduce the scope of conflict.While we fully believe that more research is 士 needed to understand the ability or inability of public officials to However,changing a mind over a single issue (that is perhaps not draw in new participants into the political conflict,in this paper we well understood)is different than changing a citizen's issue priorities focus on their ability to change their constituents'political priorities and their preferred political agenda. 860

American Political Science Review (2018) 112, 4, 860–873 doi:10.1017/S0003055418000473 © American Political Science Association 2018 On the Limits of Officials’ Ability to Change Citizens’ Priorities: A Field Experiment in Local Politics DANIEL M. BUTLER University of California, San Diego HANS J.G. HASSELL Florida State University We test whether politicians’ communications shape their supporters’ policy priorities by conduct￾ing a field experiment in collaboration with several local elected officials. In the experiment, the officials sent out email messages to the constituents on their distribution lists. Half the con￾stituents received messages where the official advocated for the priority of a given issue, while the other half received a placebo email. We surveyed the constituents one to two months before the message went out and again the week after the official sent the message. The experiment shows that politicians did not change citizens’ priorities in the desired direction. Moreover, citizens who received a message where the official indicated the issue was a priority were not more likely to act when invited to sign a petition on the issue. Elected officials’ ability to shape the priorities of the politically active citizens with whom they regularly communicate is limited and can even be self-defeating. I nformation is a tool that has the potential to affect opinion and mobilize citizens to action (Foos and de Rooij 2017a).We test whether local elected offi￾cials can change the political agendas and issue priori￾ties of the people with whom they regularly communi￾cate. We study the ability of government officials to in￾fluence citizens’ issue priorities because governments have limited resources. Governments cannot deal with all issues at the same time; they must identify which issues will receive the highest priority. The ability to affect citizens’ priorities can thus have significant im￾plications for the allocation of political power. As Schattschneider (1960) noted, the ability to define po￾litical priorities and the alternatives changes the na￾ture of political conflict and is the “prime instrument of power” (73).1 Changing citizen priorities and the agenda changes the political lines of division within so￾ciety and can reallocate power among political actors. Citizens’ priorities and political agendas are es￾pecially important to understanding local politics Daniel M. Butler is an Associate Professor, University of California, San Diego, Social Sciences Building, 9500 Gilman Drive #0521, La Jolla, CA 92093-0521 (daniel.butler@gmail.com). Hans J.G. Hassell is an Assistant Professor, Florida State Univer￾sity, Department of Political Science, 531 Bellamy Building, Tallahas￾see, FL 32306 (hans.hassell@gmail.com). An earlier version of the paper was presented at the Urban Po￾litical Economy Conference at Vanderbilt University in March 2017. We thank Marc Meredith and other conference participants for com￾ments. Replication files are available at the American Political Sci￾ence Review Dataverse: https://doi.org/10.7910/DVN/KTCXTE. Received: May 15, 2017; revised: January 16, 2018; accepted: July 5, 2018. First published online: September 4, 2018. 1 Schattschneider (1960) identifies two ways of changing the balance of power: First, by defining the political priorities and, second, by enlarging or reducing the scope of conflict and the participants in that conflict. Sarah Anzia’s (2014) and Justin de Benedictis-Kessner’s (2018) work on changing the timing of elections to correspond with national elections are examples of the few studies we are aware of that examine the ability of local public officials to enlarge or reduce the scope of conflict. While we fully believe that more research is needed to understand the ability or inability of public officials to draw in new participants into the political conflict, in this paper we focus on their ability to change their constituents’ political priorities. (Oliver and Ha 2007; Rugh and Trounstine 2011; Tausanovitch and Warshaw 2014). At the local level, many of the political questions are not whether an action should be taken, but whether certain issues should be prioritized (Erie, Kogan, and MacKenzie 2011; Franklin and Ebdon 2004; Guo and Neshkova 2013). For example, given a list of infrastructure projects, many constituents would support action on all the necessary improvements provided there were sufficient resources. However, given a limited budget, and limited time resources, constituents might prioritize certain action items over others. Studies of agenda setting in other contexts have found that political elites can drive the political agenda and priorities of the general public. However, these studies have almost exclusively focused on the ability of nonelected political elites to change citizens’ prior￾ities and the effect that these changes have on poli￾tics (e.g., Iyengar and Kinder 1987; King, Schneer, and White 2017; McComb and Shaw 1972). Scholars have not studied elected officials’ ability to shape their con￾stituents’ priorities. There are both arguments that public officials can shape the political agenda through their communica￾tions with constituents and reasons to believe that they cannot. On one hand, there are reasons that commu￾nication by elected officials might shape citizens’ pri￾orities. Previous studies have found that other political elites such as the media and interest groups shape the general public’s priorities (Iyengar and Kinder 1987; King, Schneer, and White 2017; McComb and Shaw 1972). What these actors talk about and highlight of￾ten become the general public’ priorities and set the agenda for political debate. Likewise, communication by other political elites, particularly the media or in￾terest groups, has a strong effect on citizens’ priorities and can propel citizens to action (Gerber, Karlan, and Bergan 2009). There is also some evidence that elite communication can sometimes shape public opinion (e.g., Broockman and Butler 2016).2 2 However, changing a mind over a single issue (that is perhaps not well understood) is different than changing a citizen’s issue priorities and their preferred political agenda. 860 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

On the Limits of Officials'Ability to Change Citizens'Priorities On the other hand.there are also reasons to believe reduce the participation that public officials want from that public officials'communication with constituents constituents (Levine 2015). may not change their priorities.Agenda setting by We study how elected officials'messages influence elected officials may work differently than agenda set- constituents'priorities and actions by collaborating ting by other elites who are not necessarily in a position with elected officials in four different municipalities to to directly affect policy and whose messages reach a dif- conduct a field experiment.In the experiment,the part- ferent audience.One reason elite communication may nering officials sent out email messages to constituents not affect constituent priorities and action is that pub on their distribution lists.We had the partnering of- lic officials communicate most often with constituents ficials send messages to individuals who already sup- who likely already hold well formulated opinions.The ported the issue but who had indicated that the issue citizens that public officials can most easily reach often was lower on the political agenda.We surveyed these have intentionally opted to receive updates and infor- individuals one to two months before the message went mation from their representatives.They are likely to out and again the week after the official sent the mes- already be well-informed.People who have well formu- sage to test how these citizens responded to the mes- lated opinions and who traditionally follow politics are sage from their city official. less likely to be affected by the messages they receive The key aspect of the experiment is that half of the (Arceneaux and Johnson 2013:Krosnick 1990:Zaller residents received messages where the official advo- 1992).3 These individuals are also likely to be strong cated that a given issue should be placed higher on policy demanders whose principles and priorities are the political agenda,while the other half received a harder to change(Bawn et al.2012;Fenno 1978;Kros- placebo email (i.e.,an email that did not advocate for nick 1990:Masket 2009).In short,previous research the importance of an issue but was otherwise similar) showing that politicians can change constituents'opin- We included the placebo email (as opposed to a control ions may not extend to the influence of public officials group that received nothing)to carry out a placebo de- on the coalitions to which they have the most direct sign (Nickerson 2005).We could carry out a placebo 4r元 access. design because we tracked who opened the emails that In addition,even if public officials can change the the official sent.This allows us to compare the peo- political agendas of those they communicate with,we ple who were actually treated (because they opened have reason to doubt that these changes have any up the email)to the individuals in the placebo control meaningful influence on political actions.Holding group who also opened up their email.To be clear,our public opinions and using those public opinions in the sample is not representative of voters.We intentionally ballot booth are not the only way that citizens influence study the effect of officials'communication on those political outcomes(Bergan 2009:Kam and Zechmeis with whom they most frequently communicate. ter 2013).Politicians must often rely on the willingness Our experiment finds that public officials'messages of individuals or groups to become involved on a par- did not,on average,change the political agenda and ticular issue to achieve their policy goals (Schlozman priorities of the citizens they contacted.If anything. 1984).Given the reliance of political elites on the in the public officials'messaging decreased the likelihood volvement of other groups,can public officials'efforts that individuals thought the issue should be a priority to change priorities affect constituents'actions in sup- and caused citizens to be less likely to act when invited port of a policy priority? to sign a petition on the issue.5 In some cases,public officials'communication about their priorities may actually decrease the willingness THE EFFECTS OF ELECTED OFFICIAL of individuals to participate through self-undermining rhetoric(Levine 2015;Levine and Kline 2017).In gen- COMMUNICATION ON POLITICAL AGENDAS eral,actions motivated by purposive goals may actually There are competing expectations about elected of- be depressed by information about the actions that oth- ficials'ability to change their constituents'priorities. ers have taken(Hassell and Wyler 2018:Mutz 1995).In On one hand,many individuals'opinions appear to formation about officials'actions and priorities might be malleable.Rather than citizens using their policy be particularly impactful because officials are able to opinions to influence representatives,many analyses do something about the issue.If individuals recognize conclude that politicians,especially those who share that a public official with political influence is inter- a party identity,shape voters'opinions (Broockman ested in the agenda item,an individual may rational- and Butler 2016;Bullock 2011).Previous research on 四 ize that his or her action is not necessary because the politicians'ability to shape opinion has largely focused issue will be handled by the government.Thus,com on the constituent's positions,yet the quality of rep- munication from public officials about their issue pri- resentation also depends on how well constituents' orities may contain self-undermining components that While,as we note below,our study is slightly underpowered,it pro- These individuals may actually be more likely to have a negative vides strong evidence against the idea that public officials can change reaction to persuasive information and to be more likely to engage their constituents'priorities.While the lack of power reduces our in motivated reasoning (Brehm and Brehm 1981;Redlawsk 2002) ability to draw stronger conclusions about the backlash,our findings 士 Moreover,political outcomes,especially outcomes at the local are consistent with other research that has found a self-undermining level,are not determined solely on the basis of public opinion(Anzia effect of certain types of communication from political elites (Levine and Meeks 2016;Oliver 2012;Peterson 1981). 2015). 861

On the Limits of Officials’ Ability to Change Citizens’ Priorities On the other hand, there are also reasons to believe that public officials’ communication with constituents may not change their priorities. Agenda setting by elected officials may work differently than agenda set￾ting by other elites who are not necessarily in a position to directly affect policy and whose messages reach a dif￾ferent audience. One reason elite communication may not affect constituent priorities and action is that pub￾lic officials communicate most often with constituents who likely already hold well formulated opinions. The citizens that public officials can most easily reach often have intentionally opted to receive updates and infor￾mation from their representatives. They are likely to already be well-informed.People who have well formu￾lated opinions and who traditionally follow politics are less likely to be affected by the messages they receive (Arceneaux and Johnson 2013; Krosnick 1990; Zaller 1992).3 These individuals are also likely to be strong policy demanders whose principles and priorities are harder to change (Bawn et al. 2012; Fenno 1978; Kros￾nick 1990; Masket 2009). In short, previous research showing that politicians can change constituents’ opin￾ions may not extend to the influence of public officials on the coalitions to which they have the most direct access. In addition, even if public officials can change the political agendas of those they communicate with, we have reason to doubt that these changes have any meaningful influence on political actions.4 Holding public opinions and using those public opinions in the ballot booth are not the only way that citizens influence political outcomes (Bergan 2009; Kam and Zechmeis￾ter 2013). Politicians must often rely on the willingness of individuals or groups to become involved on a par￾ticular issue to achieve their policy goals (Schlozman 1984). Given the reliance of political elites on the in￾volvement of other groups, can public officials’ efforts to change priorities affect constituents’ actions in sup￾port of a policy priority? In some cases, public officials’ communication about their priorities may actually decrease the willingness of individuals to participate through self-undermining rhetoric (Levine 2015; Levine and Kline 2017). In gen￾eral, actions motivated by purposive goals may actually be depressed by information about the actions that oth￾ers have taken (Hassell and Wyler 2018;Mutz 1995). In￾formation about officials’ actions and priorities might be particularly impactful because officials are able to do something about the issue. If individuals recognize that a public official with political influence is inter￾ested in the agenda item, an individual may rational￾ize that his or her action is not necessary because the issue will be handled by the government. Thus, com￾munication from public officials about their issue pri￾orities may contain self-undermining components that 3 These individuals may actually be more likely to have a negative reaction to persuasive information and to be more likely to engage in motivated reasoning (Brehm and Brehm 1981; Redlawsk 2002) 4 Moreover, political outcomes, especially outcomes at the local level, are not determined solely on the basis of public opinion (Anzia and Meeks 2016; Oliver 2012; Peterson 1981). reduce the participation that public officials want from constituents (Levine 2015). We study how elected officials’ messages influence constituents’ priorities and actions by collaborating with elected officials in four different municipalities to conduct a field experiment. In the experiment, the part￾nering officials sent out email messages to constituents on their distribution lists. We had the partnering of￾ficials send messages to individuals who already sup￾ported the issue but who had indicated that the issue was lower on the political agenda. We surveyed these individuals one to two months before the message went out and again the week after the official sent the mes￾sage to test how these citizens responded to the mes￾sage from their city official. The key aspect of the experiment is that half of the residents received messages where the official advo￾cated that a given issue should be placed higher on the political agenda, while the other half received a placebo email (i.e., an email that did not advocate for the importance of an issue but was otherwise similar). We included the placebo email (as opposed to a control group that received nothing) to carry out a placebo de￾sign (Nickerson 2005). We could carry out a placebo design because we tracked who opened the emails that the official sent. This allows us to compare the peo￾ple who were actually treated (because they opened up the email) to the individuals in the placebo control group who also opened up their email. To be clear, our sample is not representative of voters.We intentionally study the effect of officials’ communication on those with whom they most frequently communicate. Our experiment finds that public officials’ messages did not, on average, change the political agenda and priorities of the citizens they contacted. If anything, the public officials’ messaging decreased the likelihood that individuals thought the issue should be a priority and caused citizens to be less likely to act when invited to sign a petition on the issue.5 THE EFFECTS OF ELECTED OFFICIAL COMMUNICATION ON POLITICAL AGENDAS There are competing expectations about elected of￾ficials’ ability to change their constituents’ priorities. On one hand, many individuals’ opinions appear to be malleable. Rather than citizens using their policy opinions to influence representatives, many analyses conclude that politicians, especially those who share a party identity, shape voters’ opinions (Broockman and Butler 2016; Bullock 2011). Previous research on politicians’ ability to shape opinion has largely focused on the constituent’s positions, yet the quality of rep￾resentation also depends on how well constituents’ 5 While, as we note below, our study is slightly underpowered, it pro￾vides strong evidence against the idea that public officials can change their constituents’ priorities. While the lack of power reduces our ability to draw stronger conclusions about the backlash, our findings are consistent with other research that has found a self-undermining effect of certain types of communication from political elites (Levine 2015). 861 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

Daniel M.Butler and Hans J.G.Hassell priorities are reflected in governments'priorities icy success often relies on individuals'willingness to (Druckman and Jacobs 2015). take political action in support of the cause.Political In addition.numerous studies have shown that non- leadership requires successfully encouraging others to elected political actors,specifically the media and in- act on a specific agenda.For these reasons,we also terest groups,can affect citizens'priorities and increase look at the effect of officials'communication on con- the salience of issues on the public agenda (Iyengar and stituents'actions. Kinder 1987:King.Schneer.and White 2017:Kiousis A stated priority is not the same as a public action. and McCombs 2004;McComb and Shaw 1972) and opinions often do not conform with actions taken On the other hand,prior studies have focused on either privately (Berinsky 2004)or publicly (LaPiere political elites'ability to change the opinions of the 1934).Studies have shown that individuals are willing general public,rather than the politically active con- to lie or decline to respond when they know their views stituents that officials reach with their messaging. are not perceived as socially acceptable(Berinsky 1999, Although the general public is largely inattentive to po- 2004:Schuman and Presser 1980:Vogel and Ardoin litical affairs,and thus susceptible to persuasion,politi- 2008).It is possible that pressure from politicians may cally involved constituents are more likely to be knowl- change publicly stated priorities without changing un- edgeable about politics and to have priorities that are derlying motivations to participate and engage on an harder to move (Arceneaux and Johnson 2013:Kros- issue. nick 1990;Zaller 1992).While previous studies have Furthermore,communication from public officials shown that even the most informed voters are still in- might be self-undermining by encouraging compla- 元 fluenced by elite communication on single issues,the cency as constituents perceive that the issue is already effect "is swamped by the average absolute effect of being handled (Levine 2015).There is some evidence exposing subjects to details about...policy"(Bullock that descriptions of others taking action or past success 2011,500).The effect may be even further reduced reduces participation relative to information that com- when trying to change priorities (rather than opin- municates a lack of action on the issue (Hassell and ions)that are often already informed by a wealth of Wyler 2018;Levine and Kam 2017).7 knowledge.Thus,while there is evidence that political elites can change voters'opinions,these previous stud- ies have not examined constituents'priorities and have RESEARCH DESIGN not focused on those individuals that politicians typi- We test local officials'ability to affect issue salience and cally reach with their messaging. to encourage participation on an issue by conducting Moreover,there is evidence that attempts to per- embedded field experiments(Foos and John 2018;Foos suade knowledgeable individuals with well-formed and de Rooij 2017b)in collaboration with city officials opinions may prompt a negative backlash (Brehm from four cities across the United States.s The officials 1966;Brehm and Brehm 1981).5 This theory of psy- who worked with us on the study had earlier expressed chological reactance holds that individuals react nega- interest in helping with a research project after they tively to persuasive information when individuals per- had taken a survey administered by one of the authors.? ceive their self-determination about what priorities to Table 1 provides information about the officials and hold and what actions to take being threatened.This the cities they serve in.Two of the officials came from S5.501g perception of threat to self-determination is likely to relatively small towns(with populations under 20,000), be stronger among those with well-formulated opin- another from a mid-sized suburb with a population of ions and priorities.When others try to persuade these about 30,000,and the last a city of over 100,000 that is individuals,this theory holds that they often embrace a key part of a metropolitan area in the Midwest.The the attitude threatened by the attempt at persuasion officials also were diverse in other ways (see Table 1). (Brehm 1966).As such,attempts by public officials to For example,two of the officials were women.while encourage constituents with higher levels of knowl- edge to place more priority on certain issues may cause effects on opinion and behavior that are opposite to 7 Levine and Kam(2017)find that messages that hint at future ac- what was intended (Dillard and Shen 2005:Ringold tion,as opposed to retrospective action,are not self-undermining 2002) However,the messages they test imply the need for support to ac- complish those goals and they come from interest groups rather than elected officials.Elected officials,unlike interest groups,can directly CHANGING PRIORITIES AND take action to change policies.Because public officials are differ eys ENCOURAGING POLITICAL ACTION ent from other political elites,we might expect constituents to react differently to communication from officials than to communication Changing the political agenda alone does not remove rom other political actors many of the barriers to policy outcomes.Achieving pol- The field experiments were approved by the IRB at Washington University in St.Louis. 9 They were around 50 officials who had taken the earlier survey and 6 Recent work by Guess and Coppock (2016)finds that there is no expressed interest in helping with academic research generally (with. backlash among the general public when they are presented with fac- out expressing interest in a specific project).For this experiment,we tual information about a topic.However,their experiments (1)look invited all of them to collaborate with us.We first made the invita- at a general population rather than a sample of politically knowl- tions via email and talked by phone with those who expressed some edgeable and interested individuals and(2)present factual informa- initial interest.Ultimately,only these four officials could collaborate tion rather than information from a source that may have ulterior A few others were no longer serving and the majority who responded motives (such as a publicly elected official). said they were too busy to help at the time. 862

Daniel M. Butler and Hans J.G. Hassell priorities are reflected in governments’ priorities (Druckman and Jacobs 2015). In addition, numerous studies have shown that non￾elected political actors, specifically the media and in￾terest groups, can affect citizens’ priorities and increase the salience of issues on the public agenda (Iyengar and Kinder 1987; King, Schneer, and White 2017; Kiousis and McCombs 2004; McComb and Shaw 1972) On the other hand, prior studies have focused on political elites’ ability to change the opinions of the general public, rather than the politically active con￾stituents that officials reach with their messaging. Although the general public is largely inattentive to po￾litical affairs, and thus susceptible to persuasion, politi￾cally involved constituents are more likely to be knowl￾edgeable about politics and to have priorities that are harder to move (Arceneaux and Johnson 2013; Kros￾nick 1990; Zaller 1992). While previous studies have shown that even the most informed voters are still in￾fluenced by elite communication on single issues, the effect “is swamped by the average absolute effect of exposing subjects to details about…policy” (Bullock 2011, 500). The effect may be even further reduced when trying to change priorities (rather than opin￾ions) that are often already informed by a wealth of knowledge. Thus, while there is evidence that political elites can change voters’ opinions, these previous stud￾ies have not examined constituents’ priorities and have not focused on those individuals that politicians typi￾cally reach with their messaging. Moreover, there is evidence that attempts to per￾suade knowledgeable individuals with well-formed opinions may prompt a negative backlash (Brehm 1966; Brehm and Brehm 1981).6 This theory of psy￾chological reactance holds that individuals react nega￾tively to persuasive information when individuals per￾ceive their self-determination about what priorities to hold and what actions to take being threatened. This perception of threat to self-determination is likely to be stronger among those with well-formulated opin￾ions and priorities. When others try to persuade these individuals, this theory holds that they often embrace the attitude threatened by the attempt at persuasion (Brehm 1966). As such, attempts by public officials to encourage constituents with higher levels of knowl￾edge to place more priority on certain issues may cause effects on opinion and behavior that are opposite to what was intended (Dillard and Shen 2005; Ringold 2002). CHANGING PRIORITIES AND ENCOURAGING POLITICAL ACTION Changing the political agenda alone does not remove many of the barriers to policy outcomes.Achieving pol- 6 Recent work by Guess and Coppock (2016) finds that there is no backlash among the general public when they are presented with fac￾tual information about a topic. However, their experiments (1) look at a general population rather than a sample of politically knowl￾edgeable and interested individuals and (2) present factual informa￾tion rather than information from a source that may have ulterior motives (such as a publicly elected official). icy success often relies on individuals’ willingness to take political action in support of the cause. Political leadership requires successfully encouraging others to act on a specific agenda. For these reasons, we also look at the effect of officials’ communication on con￾stituents’ actions. A stated priority is not the same as a public action, and opinions often do not conform with actions taken either privately (Berinsky 2004) or publicly (LaPiere 1934). Studies have shown that individuals are willing to lie or decline to respond when they know their views are not perceived as socially acceptable (Berinsky 1999, 2004; Schuman and Presser 1980; Vogel and Ardoin 2008). It is possible that pressure from politicians may change publicly stated priorities without changing un￾derlying motivations to participate and engage on an issue. Furthermore, communication from public officials might be self-undermining by encouraging compla￾cency as constituents perceive that the issue is already being handled (Levine 2015). There is some evidence that descriptions of others taking action or past success reduces participation relative to information that com￾municates a lack of action on the issue (Hassell and Wyler 2018; Levine and Kam 2017).7 RESEARCH DESIGN We test local officials’ ability to affect issue salience and to encourage participation on an issue by conducting embedded field experiments (Foos and John 2018;Foos and de Rooij 2017b) in collaboration with city officials from four cities across the United States.8 The officials who worked with us on the study had earlier expressed interest in helping with a research project after they had taken a survey administered by one of the authors.9 Table 1 provides information about the officials and the cities they serve in. Two of the officials came from relatively small towns (with populations under 20,000), another from a mid-sized suburb with a population of about 30,000, and the last a city of over 100,000 that is a key part of a metropolitan area in the Midwest. The officials also were diverse in other ways (see Table 1). For example, two of the officials were women, while 7 Levine and Kam (2017) find that messages that hint at future ac￾tion, as opposed to retrospective action, are not self-undermining. However, the messages they test imply the need for support to ac￾complish those goals and they come from interest groups rather than elected officials. Elected officials, unlike interest groups, can directly take action to change policies. Because public officials are differ￾ent from other political elites, we might expect constituents to react differently to communication from officials than to communication from other political actors. 8 The field experiments were approved by the IRB at Washington University in St. Louis. 9 They were around 50 officials who had taken the earlier survey and expressed interest in helping with academic research generally (with￾out expressing interest in a specific project). For this experiment, we invited all of them to collaborate with us. We first made the invita￾tions via email and talked by phone with those who expressed some initial interest. Ultimately, only these four officials could collaborate. A few others were no longer serving and the majority who responded said they were too busy to help at the time. 862 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

On the Limits of Officials'Ability to Change Citizens'Priorities TABLE 1. Partnering Officials # Location Population Position Gender Constituents in Study Northeast ≈15K Councilor Female 9 South ~7K Mayor Female South 30K Councilor Male 20 Midwest 125K Councilor Male TABLE 2. Issues City 1 City 2 City 3 City 4 Water quality ·Natural trails ·Term limits ·Bike lanes ·Community center ·Rec facilities ·Referenda ·Cleaning up the city ·Expand sewer system ·More special events Benefits for city officials ·Street repair Impact fees for ·Developing a common ·City credit card use Economic development development use area ·Nepotism laws Standards for private Off-shore oil drilling Ethical guidelines for roads city officials 4r元 two were men.One official was the mayor,while the I believe the city should consider this and it should other three were city councilors. be a medium priority for [name of city/town]. The collaborative experiments were carried out dur- I believe the city should consider this and it should ing the spring and summer of 2016.We designed and be a high priority for name of city/town. implemented these experiments to maximize external validity.For example,we had the officials decide what The text of the surveys is provided in Section SI.2 issues they would write about and we had them draft of the Supplementary Material.Because we are inter- the text of the emails.We also had them contact the ested in the ability of officials to affect their supporters' constituents using email because that is how they nor- priorities,we recruited citizens for the study from the mally contacted the constituents in the study.We car- participating officials'email distribution lists.The sur- ried out the study by taking the following steps(which vey recruitment email came from us as researchers and are summarized in Figure 1): explained that we were studying local policy priorities The full text of the invitation to take the survey is pro- vided in SI.1 of the Supplementary Material. 55.501g 1.Identify the issues.We asked each partnering official to identify four to six issues for which they were in terested in building support.We asked them to pick 3.Identifying the Sampling Frame.We used two cri- concrete goals that were relevant for their city and teria to determine which individuals would be in- for which they were willing to write about in com- cluded in the study.First,at the end of the survey we munications with citizens.Table 2 gives an overview asked participants if they would be willing to take of the issues that the officials chose for this study. a follow-up survey.Our study only includes the par- The officials also drafted the text of the issue used ticipants who answered that they would be willing to in the email messages they sent.We had the officials take a follow-up survey.Second,we limited the sam- choose the topics and draft the letter to increase the pling frame to those individuals who agreed with the external validity of the study.Our study looks at the official on an issue but did not think that the issue in effect of the types of messages that elected officials question should be a high priority for the city.10 For would send. 10 One concern might be that individuals who indicated that an is 2.Baseline Surveys.We conducted online surveys sue should be a low or medium priority for the city were actually of expressing weak opposition.This does not seem to be the case as a residents in each city that asked them about their substantial portion of these individuals were willing to sign a peti- positions on the issues that the official had identified. tion on the issue.On the post-treatment survey,14%of those who For each issue,citizens chose one of four responses: expressed the issue should not be a priority for the city were willing to sign the petition.If we include those who expressed that it was a medium priority for the city,that number rises to 28%.The will- I do not support doing this and it should not be a ingness of these individuals to sign a petition on an issue that was /:sony priority for [name of city/town]. not a high priority,or even a medium priority,suggests that this was I believe the city should consider this but it should not something they opposed.Moreover,rerunning the analyses be- low including only those who expressed that the issue was a medium not be a priority for name of city/town. priority for the city does not change the substantive results. 863

On the Limits of Officials’ Ability to Change Citizens’ Priorities TABLE 1. Partnering Officials # Location Population Position Gender Constituents in Study 1 Northeast ∼ 15K Councilor Female 89 2 South ∼ 7K Mayor Female 68 3 South ∼ 30K Councilor Male 20 4 Midwest ∼ 125K Councilor Male 67 TABLE 2. Issues City 1 City 2 City 3 City 4 • Water quality • Community center • Expand sewer system • Impact fees for development • Standards for private roads • Natural trails • Rec facilities • More special events • Developing a common use area • Off-shore oil drilling • Term limits • Referenda • Benefits for city officials • City credit card use • Nepotism laws • Ethical guidelines for city officials • Bike lanes • Cleaning up the city • Street repair • Economic development two were men. One official was the mayor, while the other three were city councilors. The collaborative experiments were carried out dur￾ing the spring and summer of 2016. We designed and implemented these experiments to maximize external validity. For example, we had the officials decide what issues they would write about and we had them draft the text of the emails. We also had them contact the constituents using email because that is how they nor￾mally contacted the constituents in the study. We car￾ried out the study by taking the following steps (which are summarized in Figure 1): 1. Identify the issues. We asked each partnering official to identify four to six issues for which they were in￾terested in building support. We asked them to pick concrete goals that were relevant for their city and for which they were willing to write about in com￾munications with citizens. Table 2 gives an overview of the issues that the officials chose for this study. The officials also drafted the text of the issue used in the email messages they sent. We had the officials choose the topics and draft the letter to increase the external validity of the study. Our study looks at the effect of the types of messages that elected officials would send. 2. Baseline Surveys. We conducted online surveys of residents in each city that asked them about their positions on the issues that the official had identified. For each issue, citizens chose one of four responses: I do not support doing this and it should not be a priority for [name of city/town]. I believe the city should consider this but it should not be a priority for [name of city/town]. I believe the city should consider this and it should be a medium priority for [name of city/town]. I believe the city should consider this and it should be a high priority for [name of city/town]. The text of the surveys is provided in Section SI.2 of the Supplementary Material. Because we are inter￾ested in the ability of officials to affect their supporters’ priorities, we recruited citizens for the study from the participating officials’ email distribution lists. The sur￾vey recruitment email came from us as researchers and explained that we were studying local policy priorities. The full text of the invitation to take the survey is pro￾vided in SI.1 of the Supplementary Material. 3. Identifying the Sampling Frame. We used two cri￾teria to determine which individuals would be in￾cluded in the study. First, at the end of the survey we asked participants if they would be willing to take a follow-up survey. Our study only includes the par￾ticipants who answered that they would be willing to take a follow-up survey. Second, we limited the sam￾pling frame to those individuals who agreed with the official on an issue but did not think that the issue in question should be a high priority for the city.10 For 10 One concern might be that individuals who indicated that an is￾sue should be a low or medium priority for the city were actually expressing weak opposition. This does not seem to be the case as a substantial portion of these individuals were willing to sign a peti￾tion on the issue. On the post-treatment survey, 14% of those who expressed the issue should not be a priority for the city were willing to sign the petition. If we include those who expressed that it was a medium priority for the city, that number rises to 28%. The will￾ingness of these individuals to sign a petition on an issue that was not a high priority, or even a medium priority, suggests that this was not something they opposed. Moreover, rerunning the analyses be￾low including only those who expressed that the issue was a medium priority for the city does not change the substantive results. 863 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

Daniel M.Butler and Hans J.G.Hassell FIGURE 1.Research Design Steps Step Timing Action Step Taken By: 1 Week 0 Official picks issues s/he wants to write on Elected Official V 2 Week 1 Online survey of residents about these issues Research Team Y 3 Week 5 Identify the issues that the survey respondents Research Team could be moved on.Use in sampling frame (when given issue letter). Number movable on at least 1 issue:451 Number movable on 1 issue:244 Number movable on 2 issues:133 Number movable on 3 issues:59 Number movable on 4 issues:14 Number movable on 5 issues:1 V 4 Week5 Randomly assign individuals to treatments Research Team 化Y Placebo (No issue Issue (Info on 1 content).N=227 issue).N=224 丝 5 Weeks 6-8 Official sends assigned emails to citizens Elected Official V 6-7 Week 7-9 Follow-up,online survey of those who opened Research Team emailed (and were thus exposed to treatment). Placebo Group(N=119) Issue Group(N=125) Analysis. Research Team Total Number of People in final sample:244 Number movable on 1 issue:137 Number movable on 2 issues:75 Number movable on 3 issues:27 Number movable on 4 issues:4 Number movable on 5 issues:1 purposes of exposition,we will say that a citizen is leverage because,in some cases,citizens were mov- movable on an issue if in their response during the able on multiple issues.When creating the sampling baseline survey they said that they"believe the city frame,we took steps to privilege working with peo- should consider"the issue but that it"should not be ple who thought the issue was not a priority.We a priority the city"or said that the issue"should be a wanted to focus on people who agreed with the leg- medium priority for the city."Thus,our sample only islator but thought it was "not a priority"because included people who were movable on at least one we felt that this was the population of greatest in- issue.11 terest to officials trying to set the political agenda. For part of our analysis,the unit of observation Thus,when a citizen thought that at least one issue is the individual issue.This allows us to get more was not a priority for the city (but agreed the city should consider the issue),we only included the is- sues that they thought were not a priority in the sam- 11 If someone either disagreed with the official on all the issues pling frame.If someone did not have any issues that or agreed with the official and thought they were all high priority issues-or some combination of those two options-they were ex- they thought were "not a priority for the city,"then cluded from the sample prior to randomizing the treatments. the sampling frame included all the issues for which 864

Daniel M. Butler and Hans J.G. Hassell FIGURE 1. Research Design Steps Step Timing Acon Step Taken By: 1 Week 0 Official picks issues s/he wants to write on Elected Official 2 Week 1 Online survey of residents about these issues Research Team 3 Week 5 Idenfy the issues that the survey respondents could be moved on. Use in sampling frame (when given issue leer). Number movable on at least 1 issue: 451 Number movable on 1 issue: 244 Number movable on 2 issues: 133 Number movable on 3 issues: 59 Number movable on 4 issues: 14 Number movable on 5 issues: 1 Research Team 4 Week 5 Randomly assign individuals to treatments Research Team Placebo (No issue content). N = 227 Issue (Info on 1 issue). N = 224 5 Weeks 6-8 Official sends assigned emails to cizens Elected Official 6-7 Week 7-9 Follow-up, online survey of those who opened emailed (and were thus exposed to treatment). Research Team Placebo Group (N=119) Issue Group (N=125) 8 Analysis. Total Number of People in final sample: 244 Number movable on 1 issue: 137 Number movable on 2 issues: 75 Number movable on 3 issues: 27 Number movable on 4 issues: 4 Number movable on 5 issues: 1 Research Team purposes of exposition, we will say that a citizen is movable on an issue if in their response during the baseline survey they said that they “believe the city should consider” the issue but that it “should not be a priority the city” or said that the issue “should be a medium priority for the city.” Thus, our sample only included people who were movable on at least one issue.11 For part of our analysis, the unit of observation is the individual issue. This allows us to get more 11 If someone either disagreed with the official on all the issues or agreed with the official and thought they were all high priority issues—or some combination of those two options—they were ex￾cluded from the sample prior to randomizing the treatments. leverage because, in some cases, citizens were mov￾able on multiple issues. When creating the sampling frame, we took steps to privilege working with peo￾ple who thought the issue was not a priority. We wanted to focus on people who agreed with the leg￾islator but thought it was “not a priority” because we felt that this was the population of greatest in￾terest to officials trying to set the political agenda. Thus, when a citizen thought that at least one issue was not a priority for the city (but agreed the city should consider the issue), we only included the is￾sues that they thought were not a priority in the sam￾pling frame. If someone did not have any issues that they thought were “not a priority for the city,” then the sampling frame included all the issues for which 864 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

On the Limits of Officials'Ability to Change Citizens'Priorities TABLE 3.Balance Tests Individual level Issue level DV =Assigned to treatment email (1) (2) Regression model: Probit Probit Female -0.25 -0.13 (0.17) (0.14) Age(six categories) 0.07 0.03 (0.07) (0.06) Education(six categories) 0.02 0.02 (0.07) (0.06) Political interest 0.06 -0.04 (0.16) (0.13) Follow local politics -0.15 -0.09 (0.18) (0.15) Constant -0.05 -0.43 (0.84) (0.69) Joint significance test(Likelihood ratio test) Chi-square 4.05 1.52 P-value 0.542 0.911 Observations 238 408 Note:Standard errors in parentheses. they thought"should be a medium priority for the the study.Our study looks at the effect of the types city.” of messages that elected officials would send. 白 4.Treatments and Randomization.In our study,all cit- 5.Message Delivery.About six to eight weeks after in- izens received an email message from his or her city dividuals took the baseline survey,the city official official that highlighted the internet sources where emailed them the message they were assigned to individuals could find out what the city was doing. receive in step 4.Those who did not open the first All email messages included the same subject line email within 48 hours were sent their assigned email (i.e.,this did not vary with treatment condition).We a second time (to increase the probability that they used R to randomly assign individual citizens to re- opened their email prior to taking the follow-up sur- ceive either a treatment or placebo condition.Each vey).13 citizen was randomly assigned with a 50%probabil- 6. Tracking Who Opened the Email.The emails all used ity of being in either condition.The placebo con- the same subject line:"Improving [Name of City]." dition did not include any additional information. The emails were sent out using MailChimp,which However,in the treatment condition,the email in- tracks who opens emails by embedding a small im- cluded an additional paragraph where the official age in the email.We use this information to carry advocated for an issue that was important to them. out a placebo analysis(Nickerson 2005)by limiting Section SI.3 of the Supplementary Material pro- our sample to those who opened their emails.We vides the text of these emails.If a citizen assigned take this step because the treatment can only af- to the treatment condition was only movable on one fect those who opened their emails (just like a door- issue(see step 3),then they received the information to-door treatment in a Get-out-the-vote campaign about that issue.If they were movable on more than is only delivered to those who open their doors) one issue,then we used R to randomly choose which We can do this without introducing bias because we of those issues was provided in the treatment email tracked who in the placebo condition opened their message.2 Finally,as a reminder,we had the officials emails(and thus represent the types of people who draft the letter to increase the external validity of would have been exposed to the treatment had they been assigned to that condition).For the analysis,we 12 For the analysis that looks at the issue-level attitude.the number of restrict the sample to those who opened the email issues on which an individual was movable affected the probability from the city official before taking the survey. that a given issue was treated (Aronow and Middleton 2013).For example,if an individual was movable on only one issue,there was a 50%chance that they received a letter on that issue.However,if an individual was movable on two issues,there was a 25%change on which an individual was movable.Section SI.7 of the Appendix that they received a letter about one of those issues (because there provides the R-code we used to implement the randomization. was a chance of receiving a letter,and-conditional on receiving a One concern is that people who received two emails may have letter.一a %chance that the letter was on the given issue and⅓*k=⅓ been more annoyed and defensive and had an artificially negative MM//:sdny =25%).Table SI.1 provides the probability of treatment assignment reaction as a result.As a robustness check,we reran the analysis only based on the number of issues for which an individual was movable. looking at those who opened up the first email (and therefore never We follow Angrist(1998)and account for the differences in probabil- received a second email).The results are presented in Table SI.3 and ities of assignment by including fixed effects for the number of issues show that the coefficients all point in the same direction. 865

On the Limits of Officials’ Ability to Change Citizens’ Priorities TABLE 3. Balance Tests Individual level Issue level DV = Assigned to treatment email (1) (2) Regression model: Probit Probit Female − 0.25 − 0.13 (0.17) (0.14) Age (six categories) 0.07 0.03 (0.07) (0.06) Education (six categories) 0.02 0.02 (0.07) (0.06) Political interest 0.06 − 0.04 (0.16) (0.13) Follow local politics − 0.15 − 0.09 (0.18) (0.15) Constant − 0.05 − 0.43 (0.84) (0.69) Joint significance test (Likelihood ratio test) Chi-square 4.05 1.52 P-value 0.542 0.911 Observations 238 408 Note: Standard errors in parentheses. they thought “should be a medium priority for the city.” 4. Treatments and Randomization. In our study, all cit￾izens received an email message from his or her city official that highlighted the internet sources where individuals could find out what the city was doing. All email messages included the same subject line (i.e., this did not vary with treatment condition). We used R to randomly assign individual citizens to re￾ceive either a treatment or placebo condition. Each citizen was randomly assigned with a 50% probabil￾ity of being in either condition. The placebo con￾dition did not include any additional information. However, in the treatment condition, the email in￾cluded an additional paragraph where the official advocated for an issue that was important to them. Section SI.3 of the Supplementary Material pro￾vides the text of these emails. If a citizen assigned to the treatment condition was only movable on one issue (see step 3), then they received the information about that issue. If they were movable on more than one issue, then we used R to randomly choose which of those issues was provided in the treatment email message.12 Finally, as a reminder, we had the officials draft the letter to increase the external validity of 12 For the analysis that looks at the issue-level attitude, the number of issues on which an individual was movable affected the probability that a given issue was treated (Aronow and Middleton 2013). For example, if an individual was movable on only one issue, there was a 50% chance that they received a letter on that issue. However, if an individual was movable on two issues, there was a 25% change that they received a letter about one of those issues (because there was a ½ chance of receiving a letter, and—conditional on receiving a letter—a ½ chance that the letter was on the given issue and ½*½ = ¼ = 25%). Table SI.1 provides the probability of treatment assignment based on the number of issues for which an individual was movable. We follow Angrist (1998) and account for the differences in probabil￾ities of assignment by including fixed effects for the number of issues the study. Our study looks at the effect of the types of messages that elected officials would send. 5. Message Delivery. About six to eight weeks after in￾dividuals took the baseline survey, the city official emailed them the message they were assigned to receive in step 4. Those who did not open the first email within 48 hours were sent their assigned email a second time (to increase the probability that they opened their email prior to taking the follow-up sur￾vey).13 6. Tracking Who Opened the Email.The emails all used the same subject line: “Improving [Name of City].” The emails were sent out using MailChimp, which tracks who opens emails by embedding a small im￾age in the email. We use this information to carry out a placebo analysis (Nickerson 2005) by limiting our sample to those who opened their emails. We take this step because the treatment can only af￾fect those who opened their emails (just like a door￾to-door treatment in a Get-out-the-vote campaign is only delivered to those who open their doors). We can do this without introducing bias because we tracked who in the placebo condition opened their emails (and thus represent the types of people who would have been exposed to the treatment had they been assigned to that condition). For the analysis, we restrict the sample to those who opened the email from the city official before taking the survey. on which an individual was movable. Section SI.7 of the Appendix provides the R-code we used to implement the randomization. 13 One concern is that people who received two emails may have been more annoyed and defensive and had an artificially negative reaction as a result. As a robustness check, we reran the analysis only looking at those who opened up the first email (and therefore never received a second email). The results are presented in Table SI.3 and show that the coefficients all point in the same direction. 865 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

Daniel M.Butler and Hans J.G.Hassell 7 Follow-up Survey.Three days after the city official Attrition can also be a concern if it is related to treat- sent the email message,we emailed citizens for the ment assignment.In other words,if a given treatment is follow-up survey to measure the impact of the treat- causing people to systematically drop out of the survey, ment.Only citizens who took the baseline survey this can introduce bias.Columns 1 and 3 of Table 4 test and agreed to the follow-up survey were contacted. for whether treatment assignment is related to drop- In addition to the initial invitation,we sent two re- ping out of the survey.We again look at the results minder emails for those who had not taken the sur- both at the individual level(column 1)and the issue vey yet.The initial invitation and the reminder in- level (column 3).These probit regressions look at all vitations were spaced two to three days apart.Most the observations that were in the sampling frame when people who took the follow-up survey did so within we randomized (i.e.,because they completed the pre- a week of receiving the email from their city offi- treatment survey and were eligible for treatment).The cial.For our placebo design,we restrict the sample dependent variable is simply whether the observation to those who opened the email from the official and is in the final sample (i.e..because the respondent an- who did so before taking the follow-up survey(Nick- swered the question in the post-treatment survey).The erson 2005) results show that the missing observations(missing be cause of attrition)are missing independent of treat- INFORMATION ABOUT THE SAMPLE ment assignment.In other words,individuals (or the issues we asked about)that were assigned to the treat- We achieved a high follow-up rate on the post- ment group were not more or less likely to attrite from treatment survey (especially given that no incentives the study. were provided),with 68%of respondents(310 out of Finally,attrition can affect the population we learn 455)from the first round also taking the follow-up sur- about in the study.Columns 2 and 4 of Table 4 analyze vey.For the analysis,we use the 80%of respondents whether certain types of people were more likely to from this subset(244 out of 310)who also opened the drop out of the study.As Column 2 shows,there was no 4号元 email sent from the official(as tracked by MailChimp) systematic attrition at the individual level.There was before taking our survey.We use this subsample in some attrition at the issue level,with older individuals our analysis because they are the ones who were ex- and men being more likely to be in the final sample. posed to the intended message (either the treatment Table 5 provides a more general overview of our or placebo message).When we analyze the individual- sample.Our study investigates the reactions of individ- level data.we have 244 observations.When we look at uals who are reached by public officials.These individ- individuals issue priorities,we have 415 observations uals differ from the general population in systematic (408 of whom answered all the pre-treatment demo- ways and this is reflected in our sample.Our sample graphic questions).14 is older,politically interested,and highly likely to fol- Attrition can be a problem if it leads to imbalance low local politics.These individuals may respond differ- 是 between the treatment and placebo groups.We test for ently than the general public to messages from politi- balance by regressing the randomly assigned treatment cians.These differences,however,are intentional;we (1 treatment,0=placebo)on the demographics that are interested in politicians'ability to influence the au- were gathered in the first wave of the survey (and thus dience to whom they are regularly communicating. S5.501g measured pre-treatment).The independent variables include gender,education (six-point scale),age (six- point scale),level of political interest (four-point scale) RESULTS and how much they follow local politics(three-point scale).The wording for these questions is provided in Because we are interested in the ability of politicians to the Section SI.2 of the Supplementary Material.Be- change their constituents'priorities and to encourage cause we analyze the results at both the individual level their constituents to take political action,we estimate and at the issue level,Table 3 presents the balance tests the effect of the issue priority email treatment on two for both levels of the data(column 1 for the individual outcomes.First,we test whether the elected official's level and column 2 for the issue level).These probit message increases the priority of the issue for the re- regressions test the significance of each variable indi- spondent.In the pre-treatment survey,we asked indi- vidually and all the variables jointly (see the bottom of viduals about their attitudes on the issue,using a four- the table for the results of the joint significance tests) point scale(that included the degree to which the issue The variables fail to achieve statistical significance both is a priority).We include the same question on the post- individually and jointly.We have balance on these pre- treatment survey (for all the issues in that city)to ana- treatment characteristics. lyze the impact of the official's message.The question wording is provided in Section SI.2 of the Supplemen- 14 The sample size affects power.In Section SI.6 of the Supplemen- tary Material.Because these questions included four tary Material,we present simulations to investigate how much power categories,we use an ordered probit model to analyze we have for our specific sample size at different treatment effect this outcome. sizes (Coppock 2013).While not drastically underpowered,the re- As noted in the procedures above,when constituents sults show that the power for the analyses are roughly between 0.6 and 0.65 for the treatment effects we find.The results of the simu- were moveable on more than one issue.we random- lation,along with the R-code to produce them,are given in Section ized which issue the official would write about in his SI6 of the Supplementary Material. or her email message.For the analysis,we maximize 866

Daniel M. Butler and Hans J.G. Hassell 7. Follow-up Survey. Three days after the city official sent the email message, we emailed citizens for the follow-up survey to measure the impact of the treat￾ment. Only citizens who took the baseline survey and agreed to the follow-up survey were contacted. In addition to the initial invitation, we sent two re￾minder emails for those who had not taken the sur￾vey yet. The initial invitation and the reminder in￾vitations were spaced two to three days apart. Most people who took the follow-up survey did so within a week of receiving the email from their city offi￾cial. For our placebo design, we restrict the sample to those who opened the email from the official and who did so before taking the follow-up survey (Nick￾erson 2005). INFORMATION ABOUT THE SAMPLE We achieved a high follow-up rate on the post￾treatment survey (especially given that no incentives were provided), with 68% of respondents (310 out of 455) from the first round also taking the follow-up sur￾vey. For the analysis, we use the 80% of respondents from this subset (244 out of 310) who also opened the email sent from the official (as tracked by MailChimp) before taking our survey. We use this subsample in our analysis because they are the ones who were ex￾posed to the intended message (either the treatment or placebo message). When we analyze the individual￾level data, we have 244 observations. When we look at individuals issue priorities, we have 415 observations (408 of whom answered all the pre-treatment demo￾graphic questions).14 Attrition can be a problem if it leads to imbalance between the treatment and placebo groups.We test for balance by regressing the randomly assigned treatment (1 = treatment, 0 = placebo) on the demographics that were gathered in the first wave of the survey (and thus measured pre-treatment). The independent variables include gender, education (six-point scale), age (six￾point scale),level of political interest (four-point scale), and how much they follow local politics (three-point scale). The wording for these questions is provided in the Section SI.2 of the Supplementary Material. Be￾cause we analyze the results at both the individual level and at the issue level,Table 3 presents the balance tests for both levels of the data (column 1 for the individual level and column 2 for the issue level). These probit regressions test the significance of each variable indi￾vidually and all the variables jointly (see the bottom of the table for the results of the joint significance tests). The variables fail to achieve statistical significance both individually and jointly. We have balance on these pre￾treatment characteristics. 14 The sample size affects power. In Section SI.6 of the Supplemen￾tary Material, we present simulations to investigate how much power we have for our specific sample size at different treatment effect sizes (Coppock 2013). While not drastically underpowered, the re￾sults show that the power for the analyses are roughly between 0.6 and 0.65 for the treatment effects we find. The results of the simu￾lation, along with the R-code to produce them, are given in Section SI.6 of the Supplementary Material. Attrition can also be a concern if it is related to treat￾ment assignment. In other words, if a given treatment is causing people to systematically drop out of the survey, this can introduce bias. Columns 1 and 3 of Table 4 test for whether treatment assignment is related to drop￾ping out of the survey. We again look at the results both at the individual level (column 1) and the issue level (column 3). These probit regressions look at all the observations that were in the sampling frame when we randomized (i.e., because they completed the pre￾treatment survey and were eligible for treatment). The dependent variable is simply whether the observation is in the final sample (i.e., because the respondent an￾swered the question in the post-treatment survey). The results show that the missing observations (missing be￾cause of attrition) are missing independent of treat￾ment assignment. In other words, individuals (or the issues we asked about) that were assigned to the treat￾ment group were not more or less likely to attrite from the study. Finally, attrition can affect the population we learn about in the study. Columns 2 and 4 of Table 4 analyze whether certain types of people were more likely to drop out of the study. As Column 2 shows, there was no systematic attrition at the individual level. There was some attrition at the issue level, with older individuals and men being more likely to be in the final sample. Table 5 provides a more general overview of our sample. Our study investigates the reactions of individ￾uals who are reached by public officials. These individ￾uals differ from the general population in systematic ways and this is reflected in our sample. Our sample is older, politically interested, and highly likely to fol￾low local politics.These individuals may respond differ￾ently than the general public to messages from politi￾cians. These differences, however, are intentional; we are interested in politicians’ ability to influence the au￾dience to whom they are regularly communicating. RESULTS Because we are interested in the ability of politicians to change their constituents’ priorities and to encourage their constituents to take political action, we estimate the effect of the issue priority email treatment on two outcomes. First, we test whether the elected official’s message increases the priority of the issue for the re￾spondent. In the pre-treatment survey, we asked indi￾viduals about their attitudes on the issue, using a four￾point scale (that included the degree to which the issue is a priority).We include the same question on the post￾treatment survey (for all the issues in that city) to ana￾lyze the impact of the official’s message. The question wording is provided in Section SI.2 of the Supplemen￾tary Material. Because these questions included four categories, we use an ordered probit model to analyze this outcome. As noted in the procedures above, when constituents were moveable on more than one issue, we random￾ized which issue the official would write about in his or her email message. For the analysis, we maximize 866 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

On the Limits of Officials'Ability to Change Citizens'Priorities TABLE 4.Attrition Tests Individual level Issue level DV=Remained in sample (1) (2) (3) (4) Regression model: Probit Probit Probit Probit Treatment email 0.09 0.06 (0.12) (0.09) Female -0.15 -0.27 (0.12) (0.09) Age(six categories) 0.07 0.10 (0.05) (0.04) Education(six categories) 0.06 0.06 (0.05) (0.04) Political interest 0.02 -0.00 (0.12) (0.09) Follow local politics -0.08 -0.12 (0.14) (0.10) Constant 0.06 -0.25 -0.06 -0.34 (0.08) (0.62) (0.06) (0.46) Joint Significance Test(Likelihood Ratio Test) Chi-Square 0.52 5.17 0.47 19.49 P-value 0.471 0.395 0.491 0.002 Observations 451 442 852 839 Note:Standard errors in parentheses TABLE 5. Descriptive Statistics of the Sample Individual level Issue level Female 47% 46% Age 18-25 0% 26-35 89% 36-45 46-55 56-65 66+ Education Less than high school High school graduate or equivalent(GED) Some college,but no degree Associate degree Bachelor's degree Graduate degree(masters,professional,or doctorate) Follow local politics Not at all Somewhat A lot Political interest Not at all Not very much A fair amount All the time Observations 867

On the Limits of Officials’ Ability to Change Citizens’ Priorities TABLE 4. Attrition Tests Individual level Issue level DV = Remained in sample (1) (2) (3) (4) Regression model: Probit Probit Probit Probit Treatment email 0.09 0.06 (0.12) (0.09) Female –0.15 –0.27 (0.12) (0.09) Age (six categories) 0.07 0.10 (0.05) (0.04) Education (six categories) 0.06 0.06 (0.05) (0.04) Political interest 0.02 –0.00 (0.12) (0.09) Follow local politics –0.08 –0.12 (0.14) (0.10) Constant 0.06 –0.25 –0.06 –0.34 (0.08) (0.62) (0.06) (0.46) Joint Significance Test (Likelihood Ratio Test) Chi-Square 0.52 5.17 0.47 19.49 P-value 0.471 0.395 0.491 0.002 Observations 451 442 852 839 Note: Standard errors in parentheses. TABLE 5. Descriptive Statistics of the Sample Individual level Issue level Female 47% 46% Age 18-25 0% 0% 26-35 8% 8% 36-45 11% 11% 46-55 20% 17% 56-65 29% 30% 66+ 31% 34% Education Less than high school 0% 0% High school graduate or equivalent (GED) 6% 6% Some college, but no degree 22% 22% Associate degree 13% 12% Bachelor’s degree 33% 37% Graduate degree (masters, professional, or doctorate) 26% 23% Follow local politics Not at all 5% 5% Somewhat 55% 55% A lot 40% 39% Political interest Not at all 45% 45% Not very much 45% 45% A fair amount 9% 9% All the time 1% 1% Observations 244 415 867 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

Daniel M.Butler and Hans J.G.Hassell TABLE 6. The Effect of Messaging on Issue Priorities DV Priority of issue (1) (2) (3) Regression model Ordered probit Ordered probit Ordered probit Treatment email -0.29 -0.25 -0.23 (0.12) (0.13) (0.13) Lagged position:Low priority -0.94 -0.96 (0.12) (0.12) Female 0.25 (0.11) Age(6 categories) -0.01 (0.04) Education(6 categories) 0.06 (0.05) Political interest -0.07 (0.12) Follow local politics -0.09 (0.14) Missing Covariate 0.25 (0.42) FE for strata？ Yes Yes Yes Observations 415 415 415 Clusters(Individual) 235 235 235 Log-likelihood 505.5 -476.2 472.0 Distribution of outcome(using inverse-probability weights) Against Low Medium High Policy Priority Priority Priority Issue letter treatment 14% 39% 35% 12% Placebo condition 8% 33% 43% 16% Notes:Standard errors in parentheses. our power by using the individual issue as the unit of those authors present the treatment effects when con- observation and including all of the individual-issue trolling for the lagged dependent variable,we provide observations that were part of the sampling frame.Cor- the same model for the sake of comparability.Third. 235.5010 respondingly,we cluster our standard errors at the re- we estimate the relationship while also controlling for spondent level.We also follow Angrist's(1998)advice the full range of pretreatment covariates we have from for studying situations where there are different strata the pretreatment survey(column 3).Missing variables at which the randomization is carried out by includ- are imputed and an additional dummy variable is in- ing fixed effects for these strata.In our case,the strata cluded in the model to indicate that the observation are represented by the number of issues on which the included missing covariates to avoid the introduction individual was moveable.15 Angrist (1998)shows that, of bias. under some modest conditions,including fixed effects We find that public officials are not able to change for these randomization strata,recovers the causal ef- these citizens'priorities.In fact,the treatment effect fect.Finally,at the bottom of Table 6,we also present is negative.When the public officials wrote that an the results when using inverse-probability weights to issue was important,the constituents who saw those adjust for differences in the probability of treatment messages were less likely to move toward saying that assignment (Aronow and Middleton 2013) the issue should be a priority.The bottom of Table 6 eys We estimate the treatment effect with three models compares distribution of treatment and placebo groups and present the results in Table 6.First,we estimate across the different categories.The individuals in the the effect when not including any controls (column 1). treatment condition were 12 percentage points less Second,we estimate the treatment effect when control- likely to say that the issue was either a high or medium ling for the respondent's lagged position on the issue priority compared with the placebo condition(47%in (column 2).We present this model because our design the treatment condition versus 59%in the placebo con- closely follows Broockman and Butler(2016).Because dition).This suggests that the messages from the pub- lic officials may even cause a backlash among those to 15 See footnote 12 for an explanation for why the probabilities differ. whom they are most easily able to communicate. Table SI.1 provides the probability of treatment assignment based on Second,we test the message's effect on the likeli- the number of issues for which an individual was movable. hood that the citizens would respond to an invitation 868

Daniel M. Butler and Hans J.G. Hassell TABLE 6. The Effect of Messaging on Issue Priorities DV = Priority of issue (1) (2) (3) Regression model Ordered probit Ordered probit Ordered probit Treatment email − 0.29 −0.25 − 0.23 (0.12) (0.13) (0.13) Lagged position: Low priority −0.94 − 0.96 (0.12) (0.12) Female 0.25 (0.11) Age (6 categories) − 0.01 (0.04) Education (6 categories) 0.06 (0.05) Political interest − 0.07 (0.12) Follow local politics − 0.09 (0.14) Missing Covariate 0.25 (0.42) FE for strata? Yes Yes Yes Observations 415 415 415 Clusters (Individual) 235 235 235 Log-likelihood − 505.5 −476.2 − 472.0 Distribution of outcome (using inverse-probability weights) Against Low Medium High Policy Priority Priority Priority Issue letter treatment 14% 39% 35% 12% Placebo condition 8% 33% 43% 16% Notes: Standard errors in parentheses. our power by using the individual issue as the unit of observation and including all of the individual-issue observations that were part of the sampling frame. Cor￾respondingly, we cluster our standard errors at the re￾spondent level. We also follow Angrist’s (1998) advice for studying situations where there are different strata at which the randomization is carried out by includ￾ing fixed effects for these strata. In our case, the strata are represented by the number of issues on which the individual was moveable.15 Angrist (1998) shows that, under some modest conditions, including fixed effects for these randomization strata, recovers the causal ef￾fect. Finally, at the bottom of Table 6, we also present the results when using inverse-probability weights to adjust for differences in the probability of treatment assignment (Aronow and Middleton 2013). We estimate the treatment effect with three models and present the results in Table 6. First, we estimate the effect when not including any controls (column 1). Second, we estimate the treatment effect when control￾ling for the respondent’s lagged position on the issue (column 2). We present this model because our design closely follows Broockman and Butler (2016). Because 15 See footnote 12 for an explanation for why the probabilities differ. Table SI.1 provides the probability of treatment assignment based on the number of issues for which an individual was movable. those authors present the treatment effects when con￾trolling for the lagged dependent variable, we provide the same model for the sake of comparability. Third, we estimate the relationship while also controlling for the full range of pretreatment covariates we have from the pretreatment survey (column 3). Missing variables are imputed and an additional dummy variable is in￾cluded in the model to indicate that the observation included missing covariates to avoid the introduction of bias. We find that public officials are not able to change these citizens’ priorities. In fact, the treatment effect is negative. When the public officials wrote that an issue was important, the constituents who saw those messages were less likely to move toward saying that the issue should be a priority. The bottom of Table 6 compares distribution of treatment and placebo groups across the different categories. The individuals in the treatment condition were 12 percentage points less likely to say that the issue was either a high or medium priority compared with the placebo condition (47% in the treatment condition versus 59% in the placebo con￾dition). This suggests that the messages from the pub￾lic officials may even cause a backlash among those to whom they are most easily able to communicate. Second, we test the message’s effect on the likeli￾hood that the citizens would respond to an invitation 868 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473

On the Limits of Officials'Ability to Change Citizens'Priorities TABLE 7.The Effect of Messaging on Signing a Petition DV =Signed petition (1) (2) (3) Regression model Probit Probit Probit Treatment email -0.36 -0.35 -0.33 (0.17) (0.17) (0.18) Lagged position:Low priority -0.20 -0.21 (0.19) (0.19) Female 0.22 (0.18) Age(six categories) 0.05 (0.08 Education(six categories) 0.07 (0.07) Political interest -0.07 (0.18) Follow local politics 0.04 (0.19 Missing covariate 0.22 (0.63) Constant -0.40 -0.26 -0.92 (0.12) (0.18) (0.91) Observations 244 244 244 Log likelihood -143.1 -142.5 -140.9 Distribution of outcome by treatment condition Signed petition Issue letter treatment 22% Placebo condition 34% Note:Standard errors in parentheses. to visit a page to sign a petition to take action on the again estimate the treatment effect in three ways:(1) issue.We included this item to measure whether re- with no control variables(column 1),when controlling spondents would take action (and not simply change only for the respondent's lagged position on the issue their stated opinions).We measured citizens'willing- (column 2),and when controlling for the full range of ness to sign a petition by including an item on the pre-treatment covariates (column 3). follow-up survey that asked if they wanted to be di- The treated individuals were also less likely to be rected toward a page where they could sign a peti- interested in signing the petition.Overall the indi- tion on the issue:"We wanted to make you aware viduals were highly interested in signing the petition, about a petition that is circulating to encourage [Name likely reflecting the fact that this was a group of in- of City]'s Town/City Council to...[issue specific lan- dividuals who were very interested in local politics guage]...Would you want to sign this petition?If you and these were all issues that individuals agreed with want to sign this petition,click yes below and you will (see Table 4).However,among this group of atten- be provided a link to the petition."All citizens were tive citizens,the treatment effect was negative and only asked about one issue (the issue that had been statistically significant.The individuals sent the treat- randomly chosen prior to the official sending out mes- ment email were 12 percentage points less likely to sages).For those in the treatment group,they were thus visit the webpage with the petition when offered the asked about the issue that the official wrote about.For chance. those in the placebo group,it was the issue that they would have received a message about had they been in the treatment group.The full texts the various ques- tions are provided in Section SI.2 of the Supplemen- previous studies(Hassell and Visalvanich 2015),we coded this as a tary Material.In Table 7,we use a probit model to an- 1 if the participants took 20 seconds or more on the page with the alyze this binary outcome in the analysis.6 Finally,we link.Those who spent less than 20 seconds or who said they were not interested in signing the petition are coded as 0.The results of those analyses reach the same conclusions as those presented in the 16 As a robustness check,we also created and analyzed a variable body the paper.The results are presented in the Appendix Section that took the time spent on the petition page into account.Following SI.4(see Table SI.2). 869

On the Limits of Officials’ Ability to Change Citizens’ Priorities TABLE 7. The Effect of Messaging on Signing a Petition DV = Signed petition (1) (2) (3) Regression model Probit Probit Probit Treatment email –0.36 –0.35 –0.33 (0.17) (0.17) (0.18) Lagged position: Low priority –0.20 –0.21 (0.19) (0.19) Female 0.22 (0.18) Age (six categories) 0.05 (0.08) Education (six categories) 0.07 (0.07) Political interest –0.07 (0.18) Follow local politics 0.04 (0.19) Missing covariate –0.22 (0.63) Constant –0.40 –0.26 –0.92 (0.12) (0.18) (0.91) Observations 244 244 244 Log likelihood –143.1 –142.5 –140.9 Distribution of outcome by treatment condition Signed petition Issue letter treatment 22% Placebo condition 34% Note: Standard errors in parentheses. to visit a page to sign a petition to take action on the issue. We included this item to measure whether re￾spondents would take action (and not simply change their stated opinions). We measured citizens’ willing￾ness to sign a petition by including an item on the follow-up survey that asked if they wanted to be di￾rected toward a page where they could sign a peti￾tion on the issue: “We wanted to make you aware about a petition that is circulating to encourage [Name of City]’s Town/City Council to… [issue specific lan￾guage]… Would you want to sign this petition? If you want to sign this petition, click yes below and you will be provided a link to the petition.” All citizens were only asked about one issue (the issue that had been randomly chosen prior to the official sending out mes￾sages). For those in the treatment group, they were thus asked about the issue that the official wrote about. For those in the placebo group, it was the issue that they would have received a message about had they been in the treatment group. The full texts the various ques￾tions are provided in Section SI.2 of the Supplemen￾tary Material. In Table 7, we use a probit model to an￾alyze this binary outcome in the analysis.16 Finally, we 16 As a robustness check, we also created and analyzed a variable that took the time spent on the petition page into account. Following again estimate the treatment effect in three ways: (1) with no control variables (column 1), when controlling only for the respondent’s lagged position on the issue (column 2), and when controlling for the full range of pre-treatment covariates (column 3). The treated individuals were also less likely to be interested in signing the petition. Overall the indi￾viduals were highly interested in signing the petition, likely reflecting the fact that this was a group of in￾dividuals who were very interested in local politics and these were all issues that individuals agreed with (see Table 4). However, among this group of atten￾tive citizens, the treatment effect was negative and statistically significant. The individuals sent the treat￾ment email were 12 percentage points less likely to visit the webpage with the petition when offered the chance. previous studies (Hassell and Visalvanich 2015), we coded this as a 1 if the participants took 20 seconds or more on the page with the link. Those who spent less than 20 seconds or who said they were not interested in signing the petition are coded as 0. The results of those analyses reach the same conclusions as those presented in the body the paper. The results are presented in the Appendix Section SI.4 (see Table SI.2). 869 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000473