Risking Peace: The Effects of Financial Market Exposure on Votes and Political Attitudes

Transcription

1 Risking Peace: The Effects of Financial Market Exposure on Votes and Political Attitudes Saumitra Jha Graduate School of Business Stanford University Moses Shayo Department of Economics The Hebrew University of Jerusalem November 17, 2015 Abstract Conflict tends to harm the economy, but voters may not fully internalize these risks. Can exposure to broader financial markets change attitudes towards conflict, and even voting decisions? Can this happen even in the context of persistent ethnic conflict? Prior to the 2015 Israeli elections, we randomly assigned financial assets to likely voters and gave them incentives to actively trade for up to seven weeks. These assets included stocks of Israeli and Palestinian companies with randomly assigned divestment dates and initial endowments. We find that this financial market exposure caused systematic shifts in voting behavior and support for policy concessions to achieve peace. There is little evidence that the effects reflect changes in voters political information, overall wealth, or changes in well-being. Instead, we find evidence consistent with financial market exposure changing both individuals evaluation of the risks to the broader economy due to peace and conflict, and their evaluation of national policy platforms. Importantly, these effects are larger for the risk averse, who otherwise tend to support status quo policies, and for inexperienced investors. In fact, the votes of inexperienced investors become more aligned with experienced investors that already support platforms favoring concessions for peace. We are particularly grateful to Kate Casey, Jim Fearon, Avner Greif, Nir Halevy, Ori Heffetz, Keith Krehbiel, Jessica Leino, Neil Malhotra, Joram Mayshar, Huggy Rao, Ken Singleton, the Stanford GSB political economy group, the GSB s CIRCLE team and seminar participants at Bocconi, Harvard, Hebrew University, IPES, the Chapman-Irvine Religion and Culture conference, NYU, LSE, Stockholm and UCSD for valuable suggestions, and to an I-Award from the Stanford Institute for Entrepreneurship in Developing Economies (SEED) for financial support. Shayo gratefully acknowledges support from the European Union, ERC Starting Grant project no Gurpal Sran and Ohad Dan provided much valued research assistance.

2 1 Introduction Can providing incentives to trade in financial markets change attitudes towards peace and conflict, and even voting decisions? This question is of grave importance given the role played by persistent conflict in much underdevelopment around the world. 1 In many conflict environments, even though the broader economy may suffer, the costs and risks of conflict are borne very unevenly by individuals. With public discourse often focused on the political, and more emotive, aspects of conflict territorial disputes, loss of life and ethnic animosities in particular the economic risks of conflict may seem less important and hard to gauge. Even among those who are directly affected by conflict, the solutions may not be clear. With low trust and limited third party enforcement, individuals may perceive a threat of hold up: that making current concessions for peace with their current adversary may weaken their group s future bargaining position and threaten their future security (Fearon, 1996). This may lead risk averse individuals to prefer policy platforms that aim to maintain status quo levels of conflict rather than risk making concessions for peace. For both those indifferent to conflict and those for whom conflict is already salient, therefore, mechanisms that align individuals interests with the risks and returns faced by the broader economy may allow them to internalize and more precisely compare the national risks and returns incurred by a status quo of conflict versus those from peace initiatives. In this paper, we test whether a historically important, but nowadays relatively neglected, mechanism exposure to broader financial markets can change attitudes towards conflict and peace, and even individuals voting decisions. We examine this in a context of highly persistent ethnic conflict that between Israelis and Palestinians. This conflict is not only of profound importance for regional and global political economy, but a legacy of more than eighty years of repeated violence on polarized attitudes, interests 1 See, for example, Blattman and Miguel (2010) and World Bank (2011). Eckstein and Tsiddon (2004) estimate that reduced investment and reallocation of resources due to conflict reduced the level of Israeli GDP per capita by 10% during the Second Intifada ( ) alone. 2

3 and lack of trust has led many to consider it intractable. Within Israel, the rival platforms on these issues in the elections of 2013 and 2015 have been broadly between parties on the right that favor the status quo viewing concessions for peace as likely to lead to a major deterioration of the security situation and parties on the left, that see status quo policies, such as permitting settlements in the West Bank, as already costly or likely to lead to further increased risks, and instead favor restarting the peace process with the goal of finding a permanent solution to the conflict. 2 Financial markets may have a particularly significant effect on attitudes towards national policies because, compared to common daily transactions, they expose individuals to more encompassing national and international sets of risks and considerations. However, measuring the effect of financial market exposure on political attitudes is very difficult in observational data given the selection processes through which financial markets develop and investors choose to hold and choose to retain financial assets. 3 This paper presents results from the first study to experimentally assign individuals financial assets, allow them to trade in those assets, and trace the effects on their political views and behavior. 4 2 Indeed, on the eve of the elections, on March , the leader of the right-wing Likud party, Prime Minister Benjamin Netanyahu, explicitly stated that Whoever moves to establish a Palestinian state or intends to withdraw from territory is simply yielding territory for radical Islamic terrorist attacks against Israel, and stated that he would not allow a Palestinian state if elected (Reuters, 2015). For a useful (brief) overview, see Natan Sachs Why Israel Waits: Anti-Solutionism as a Strategy Foreign Affairs, November/December Due to individual selection, important works examining the relationship between shareholding and firm outcomes are often explicit in interpreting correlational relationships (see eg Brav, Jiang, Partnoy, and Thomas (2008).) An interesting advance is by Bursztyn, Ederer, Ferman, and Yuchtman (2014), that assign a financial asset randomly among those that chose to purchase it, and find that this has effects on take up by peers. On factors that influence national financial development, see, e.g. Rousseau and Sylla (2008), Rajan and Zingales (2003), Levine (2005), Haber and Perotti (2008). 4 Arguably the closest paper is Jha (2015), who exploits the coincidence of individual politicians abilities to sign legally binding share contracts with novel share offerings by overseas companies to identify the effect of shareholding on support for parliamentary supremacy in the English Revolution. We also build on other papers that use observational data to examine the empirical relationships between political attitudes and investment. For example, Kaustia, Knüpfer, and Torstila (forthcoming) exploit a regional matched comparison of Finnish areas where mutual funds became publicly listed companies to others where they did not and document a relationship between the extent of demutualization in an area and the proportion of right-of-center votes. Hong and Kostovetsky (2012) find that fund managers who donate to Democrats in the United States tend to invest more in environmentally-friendly companies, and less in tobacco companies etc. Bonaparte and Kumar (2013) find a relationship between the extent 3

4 A month and a half prior to the highly contested 2015 Israeli elections, we randomly assigned 1,345 Israeli voters to either an asset treatment or a control group. Individuals in the asset treatment received endowments of assets that tracked the value of specific funds or company stocks from both Israel and the Palestinian Authority, or an endowment of cash they could invest in an asset that tracked the Tel Aviv 25 index. They were also given incentives to monitor the performance of their asset and to make weekly decisions to buy or sell part of their portfolio. We further randomly assigned dates at which individuals would be entirely divested of their portfolio to be before and after the elections, and randomly assigned the initial value of the portfolio ( either NIS 200 or NIS 400 (roughly US $50 or $100)). In parallel to the financial study, individuals from both treatment and control groups regularly participated in social and political surveys. This allowed us to track not only their investment behavior throughout the period of the study, but also their social and political views and ultimately their voting behavior. Importantly, participants did not know that the political surveys were in any way linked to the financial study, thus overcoming potential social desirability biases or experimenter demand effects (Section 2 details how this was achieved and verified). We find that this exposure to the financial markets causes large and systematic shifts in voting behavior as well as in support for policy concessions to settle the conflict between Israelis and Palestinians. We find that exposure to the stock market increases the likelihood of voting for the left in the 2015 elections by 5-6 percentage points (relative to a left vote share of 25% in the control group) and reduces the likelihood of voting for the right by about 4 percentage points (relative to 36% in the control). 5 Further, exposure to the stock market increased individuals willingness to support not only the to which households are politically active and the extent to which they participate in the stock market. 5 A desirable feature, from the standpoint of an academic study, is that Israel has a proportional representation system, with the whole country comprising a single constituency of 5.9 million eligible voters. Thus there was approximately zero chance that our study would affect the election outcomes themselves. 4

5 general principle of a two-state solution, but also a range of specific concessions for peace. Importantly, exposure to financial markets mainly affects the voting decisions of individuals who were not actively trading prior to the experiment by 7-8 percentage points for the left and has little effect on those that already claimed to be actively trading prior to the experiment. These ex ante active traders already tended to vote for pro-peace parties by around 10 percentage points on average. 6 Thus the experiment appears to align the political attitudes of new traders with those of existing traders. To shed light on the mechanism, we examine whether these effects reflect spillovers in voters political information beyond the financial markets, or are due to wealth effects and changes in well-being. We find little evidence for these alternatives. We focus instead on comparing the evidence for two main economic channels. First, individuals holding stocks on Election Day have skin-in-the-game and may be more likely to vote for parties that favor the peace process, given that peace overtures tend to raise both Israeli and Palestinian asset prices (Zussman, Zussman, and Nielsen, 2008). The second is an exposure effect: being exposed to broader national risks and returns through financial markets may lead individuals to more precisely evaluate the relative risks and returns to the economy of status quo policies versus initiatives for peace. This exposure and learning would shift political attitudes even among those divested of their assigned assets prior to the elections, and those endowed mainly with cash. The evidence largely supports the exposure channel, although among those assigned Israeli stocks skin-in-the-game also seems to matter. Those endowed with Israeli stocks were more likely to vote for the left when they held financial assets on the day of the election than if they were assigned to be divested before the election, though they are as likely to reduce their vote for the right. In contrast, those initially endowed with cash they could invest in the TA-25 or with Palestinian stocks actually show stronger effects 6 Those who traded before the experiment were 4 percentage points more likely to vote for the left relative to others with similar income and education levels. Of course, such previous trading was not randomly assigned, which is why a field experiment approach is desirable. 5

6 on their vote if they realized gains and divested before the election. For those endowed with Palestinian stocks, realized gains before the election resulted in an increase of 8.6 percentage points on the vote for the left and a reduction of 8.5 percentage points in the probability of voting for the right. Instrumenting for asset holdings on the day of the election using the portfolio of a passive investor shows no additional effect to skin-in-thegame beyond that of initial exposure. We next provide further evidence consistent with the effect of financial market exposure in causing individuals to reevaluate the costs and risks of the status quo of conflict relative to a peace settlement. We find that individuals exposed to financial assets become more likely to predict that a two state solution would improve Israel s economy. We then test a further implication: if financial market exposure causes individuals to reevaluate risks, the effects on behaviour and attitudes should be greater for those who care more about risk. Indeed, we show that while risk averse individuals in the control tend to be much more opposed to making concessions for peace, risk averse individuals who are exposed to financial assets become significantly more supportive of peace concessions, more likely to vote for the left and more likely to expect that a two state solution will improve, not only Israel s economy and security but their own. Exposure also appears to generate familiarity with the asset markets and human capital that is reflected in medium term behaviours as well. In a survey four months after the study, we find that individuals that we exposed to the financial markets are more likely to continue to read financial newspapers regularly. In a companion paper, we show that these individuals also exhibit higher financial literacy four months after the study, and for those exposed to particular assets, a greater willingness to hold similar assets in the future. The fact that the asset treatment moves the inexperienced and the risk averse more and makes them more like those with experience in the financial markets in their political decisions is suggestive that these effects are not temporary but rather reflective of financial literacy and awareness that can have longer-term effects as well. 6

7 Our research naturally links to important literatures on conflict, financial development and the effects of markets, and financial innovations, in particular, in mitigating political economy problems that stifle growth. An important body of research shows that ethnic differences can have large negative effects on public goods provision and economic performance (e.g. Alesina and La Ferrara, 2000, 2005, Montalvo and Reynal- Querol, 2005). Social and cultural biases appear to be strongly linked to ethnic conflict and have been shown to be persistent over time (e.g. Bisin and Verdier, 2011, Nunn and Wantchekon, 2011, Voigtländer and Voth, 2012, Sambanis and Shayo, 2013, Shayo and Zussman, 2014). In particular, the long history of violence between Arabs and Jews in the Holy Land during the twentieth century has led many to believe that the deep animosities and a perceived zero sum nature of the contest over symbolically important territory make the conflict highly intractable. 7 Yet, at least as early as Montesquieu (1748), economic interests from capitalist activity have been seen as means to offset the passions that may be excited by violence (Hirschman, 1977). Inter-ethnic complementarities in trade appear to have been able to mitigate incentives for violence and foster cultures of tolerance over long periods (Jha, 2013b, 2014). However, to the best of our knowledge, no one has measured the role of exposure to financial markets in serving this role. This experiment is motivated by the theoretical promise, and in some key instances, historical success, of financial innovations in mitigating conflict (Jha, 2012, 2013a, 2015). In the benchmark model of portfolio choice, in the absence of transaction costs, all individuals should hold the market portfolio of risky assets. This may also align political incentives, as all individuals stand to gain from policies that improve the returns or lower the risks of the market portfolio, including those stemming from political instability and conflict (Jha, 2012). However, naturally individuals cannot hold shares that they do not know exist, and may face costs to learning both about specific stocks and about how 7 For a recent discussion of how these are reflected in public opinion in the Israeli context, see Manekin, Grossman, and Mitts (2015). 7

8 to invest in the market, and the factors that shape the risks and returns of different assets (e.g. Merton, 1987, Coval and Moskowitz, 1999, Huberman, 2001, Grinblatt and Keloharju, 2001). Thus exposure to the stock market and an opportunity to learn by doing in a simplified manner could lead to individuals to have greater alignment with the gains of the broader economy and appreciation of the risks faced by the economy from conflict. Indeed, exposure to novel financial assets appear to have had historical success at aligning the incentives of assetholders and subsequently mitigating social conflict in three revolutionary states that subsequently led the world in economic growth: England, the United States and Japan (Jha, 2012, 2015, Jha, Mitchener, and Takashima, in progress). The possibility of building broader political support of private property rights protection also motivated privatization in former Communist states (e.g. Boycko, Shleifer, and Vishny, 1994, Jones, Megginson, Nash, and Netter, 1999, Megginson and Netter, 2001, Biais and Perotti, 2002). The comparative experiences of these cases suggest important design elements, particularly of gradually lowering barriers to secondary sales and divestment and of providing regular incentives to allow initially non-sophisticated investors to learn by doing, that directly influenced our experimental design. However, whether such approaches could have a causal effect on political attitudes and behavior in a highly polarized contemporary setting with persistent ethnic conflict remains an open question. It is to address this that we now turn. 2 Basic Experimental Design We recruited 1,681 anonymous individual participants from among Jewish Israeli citizens who had previously voted and who participate in a large Israeli internet panel. 8 This larger panel of about 60,000 participants is nationally representative in terms of age and sex, and is commonly used for commercial market research, political opinion polling and 8 By limiting our sample to past voters, we automatically excluded minors. The panel also screened out all Israelis on the US Treasury watchlist. 8

9 academic studies. 9 Though using the panel has some drawbacks our sample is not representative of the population in all dimensions (please see below) the use of the internet panel has a particularly useful feature double-blindedness in the identity of the respondents from our perspective and in the originators of different surveys from the respondents perspective. This feature allows us to avoid social desirability biases that often plague research on peace-building initiatives. We describe this in more detail below. Individuals were invited to a study on investor behaviour, and also told that they would be asked questions on various issues (please see Appendix). They were informed that they would be entered into a lottery to win a portfolio that would track both Israeli stocks and stocks from elsewhere in the Middle East. We over-sampled nonorthodox center voters at twice their vote share (see also Figure 2). 10 Among those that consented to the study as a whole, we conducted two parallel sets of surveys: a weekly series of financial surveys where they are asked to allocate assets in a portfolio, and a set of social surveys. Importantly, participants did not know that the social surveys were linked to the financial surveys: according to the panel, they were among 110 survey invitations they received between February and March from a variety of different anonymous sources. Indeed, when asked an open-ended question in the endline survey what the participants believed that researchers learned from the study, only eight mentioned politics or elections, despite the surveys running during the period of the elections (see Figure 1.) The modal responses were that the study was about gauging economic knowledge, risk attitudes, capital market behaviour and investor choices, which are accurate responses given that we study these in a companion paper. [ Figure 1 ] 9 We will show that the main effect of being exposed to assets appears to have similar effects for those with income at or below the Israeli average as that above. 10 That is, individuals who voted for Yesh Atid, Hatnu ah or Kadimah in We did not over-sample the Shas orthodox party, which positioned itself in the center in 2013, but tends to represent specific ethnic interests. 9

10 All respondents were asked to fill out an initial financial survey on investment behavior and financial literacy (please see Online Appendix). These included their prior investment history (including whether they had traded stocks in the last six months), a battery of questions measuring financial literacy, adapted from Van Rooij, Lusardi, and Alessie (2011), risk aversion and time preference (from Dohmen, Falk, Huffman, Sunde, Chupp, and Wagner (2011) and Benjamin, Choi, and Strickland (2010)), and well-being (Benjamin, Heffetz, Kimball, and Szembrot, 2014). A few days later they were invited to an initial social survey which included questions on political behavior and social and political attitudes. Of the 1,681 who completed the initial financial survey, 1,418 completed the initial social survey as well. We restricted our sample to those who had voted in 2013, and screened out those who provided incomplete answers, those who had been inconsistent when asked at different times and those who had completed the survey extremely quickly. This left us with 1,345 participants who we then assigned to the various treatments. 11 The combined result of this sampling strategy is that the sample used for random assignment approximates the broader Jewish population of Israel in terms of geographical region and sex, but tends to be more educated, more secular and with fewer individuals over 55 than are present in the Jewish Israeli population as a whole [Please see Table A1]. Among these 1345 respondents, we employed a stratified block randomization procedure designed to increase balance across treatment groups in political and demographic variables. 12 Following the initial financial survey, all 1345 respondents received a bat- 11 The main reason for screening out was extremely quick completion of the survey, which could raise a concern regarding the reliability of the responses. Specifically, the initial financial survey included 33 questions and we screened out 53 subjects who completed the entire survey in less than 180 seconds (the median completion time was 461 and the mean was 600 seconds). The remaining 20 individuals were screened out due to incomplete or inconsistent answers. In particular, we screened out 14 respondents whose answer to our question about vote for in 2013 elections was significantly different from the answer in the internet panel s database (significant in the sense that it moved a person between right and left parties). 12 Specifically, we created 104 blocks of 13 (less for one block), with the blocks created to stratify sequentially on: 2013 vote choice (with parties ordered from left to right), sex, a dummy for whether they traded stocks in the last 6 months, a measure of their willingness to take risks, and discrepancies in their reported voting in the 2013 elections. This creates relatively homogeneous blocks. Within each 10

11 tery of three anonymous surveys on their social, political attitudes and voting intentions, conducted prior to the treatment (beginning on February 1, 2015), after two weeks of treatment (March 1, 2015), and two days following the Israeli elections (March 19, 2015). The latter included a question about their actual vote. A sample of 309 served as a control group that only received these surveys. The treatment group of 1036 were randomly assigned to a set of different financial assets, half of which were initially valued NIS 200 (around $50), and half NIS 400 (around $100). These are comparable to the average Israeli daily wage of around NIS 312 in December The 202 individuals assigned an endowment of cash could choose to hold or use up to 10% a week to buy and sell an asset that tracked the performance of the Tel-Aviv 25 Index individuals were assigned an endowment of financial assets rather than cash, and every week were allowed to sell or buy this specific asset. Of these, 414 were assigned assets from Israel, evenly and randomly distributed between the Tel Aviv 25 Index, a commercial bank, Bank Leumi, and a telecoms company, Bezeq. Similarly, 416 were assigned stocks from the Palestinian Authority, distributed evenly between the Palestine Stock Exchange General Index, a commercial bank, the Bank of Palestine, and a telecoms company, PALTEL. 14 The specific companies were selected along three criteria: comparability in the economy, name recognition to our subjects and relative orthogonality to the Israeli peace process. PALTEL is the largest private employer in the Palestinian Authority while the Bank of Palestine is its largest commercial bank, with market capitalizations 6.01% and 3.52% of Palestinian GDP respectively. Bank Leumi literally means National Bank, and is a well-known but middle-tier bank in Israel, while Bezeq was the former state block we then randomize individuals into the subtreatments. 13 To avoid affecting the asset prices directly for those that are thinly traded, these assets were technically a derivative claim on the authors research funds rather than an actual purchase of the underlying asset. We also disallowed short sales, though is a subject of our future research in a different context. 14 Since the Palestinian and other assets were listed in foreign currency such as Jordanian Dinars, we fixed the exchange rate for the experiment on their assets so that there was no exchange rate risk for the Palestinian or other cross-national stocks. 11

12 telecoms monopoly, with market capitalizations 6.19% and 6.22% of Israeli GDP, respectively. We chose bricks and mortar banks and telecoms companies rather than holding companies, companies with extensive business in the West Bank or companies with overt ties to national defense so as to increase the concreteness of the assigned asset for our subjects and to avoid priming individuals towards thinking about the peace process. 15 All members of the treatment group were invited to complete an instructions survey in which they were informed of their asset allocation, the rules of the game and were asked a series of quiz questions to make sure they understand the nature of the assets they were assigned and how the value of their assets would be determined. 840 participants completed the instructions survey and agreed to continue. This may be partly due to server overload, but also may reflect non-random unwillingness to take up assets. Because we were able to survey the outcomes of those that failed to take up assets as well as those that did, we can estimate conservative Intent to Treat effects (ITT). These measure the effect of being assigned to treatment whether or not an individual actually took up the assets. Further, we can use the random assignment to treatment as an instrument for actual treatment status to measure the Treatment Effect on the Treated (TOT). The 840 participants received weekly updates about their financial portfolio at the close of business at the start of every weekend, and had until the opening of the stock market the following week to trade their assigned asset. The trades were implemented by constructing a series of survey questions on their buy and sell decisions that mimicked the functions of an online trading platform. We provided links to third party websites, 15 The only defense company in the TA-25, Elbit Systems, has a weight of only 3.26%, and the two defense companies in the TA-100 have a combined weight of of 3.34%. The valuations of such firms might respond negatively to peace overtures as in Guidolin and Ferrara (2007). Similarly, to avoid social desirability bias and to understand the effect of assets that are neutral at the national level, we had invited individuals to a study that would allow them to trade stocks from across the Middle East, and not just Israeli and Palestinian stocks, and thus included stocks from Cyprus, Egypt, Jordan and Turkey. However, as this is a novel study establishing a proof of concept, we subsequently decided at the time of assignment that it was better to simplify the study and expand the sample available for the main Israeli and Palestinian sub-treatments. Thus we postponed these alternative asset allocations for future research. To avoid deception, we did however randomly assign four individuals to be able to buy assets that tracked the Amman General Index, Cyprus/FTSE 20 General Index, the Cairo EGX 30 and Istanbul BIST 30 Index. 12

13 particularly the Hebrew version of investing.com to allow individuals to independently track and verify the performance of their stocks. Figure 4 provides a timeline of the the surveys and shows the performance of the assigned stocks over the course of the experiment. [ Figure 4] Participants were given incentives to enter a trading decision every week by a potential loss of the 10% of their portfolio should they not do so. The decision could simply be to hold their existing allocation, but we required a decision to be entered in order to avoid a penalty. 48.8% of those who completed the instructions survey and agreed to continue entered a trading decision every week prior to the elections, and 84.3% entered a trading decision in all but one week. A random sample of 346 was told at the beginning of the study that its members would be divested of their experimental portfolios on March 12 (five days before the March 17 election), while the remainder continued to trade their portfolios until April 2nd, after the election. By limiting purchases and sales to 10% of an individual s portfolio a week, we ensured that those assigned stock endowments could (non-experimentally) divest at most 34% of their initial allocation prior to the election. We also asked individuals every week whether they invested or divested in stocks outside the experiment to assess whether the treatment was leading to investment spillovers or being undone by outside trades. Two days after the elections, on March 19, we surveyed all individuals on their vote choice as well as attitudes towards the peace process. On the voting question, we were able to draw upon and compare not only individuals responses to our post-election survey but also their subsequent responses on the voting question to the survey company s own questionnaire. There were very few discrepancies between the two, again consistent with an absence of social desirability bias. 16 Between the two surveys, we were able to observe % or 1286 respondents claimed to have voted for the same party in both our post-election survey and in their later response to the panel. The coefficient on asset treatment from a regression of the probability of having a match in the two surveys is (0.0086). 13

14 the vote of 1307 out of 1345, of which 97.3% of individuals were among those assigned to asset treatment and 97.4% to the control. There is no evidence of differential attrition across sub-treatments (Table A2.) 17 3 Data Table 1 shows the means of a range of pre-treatment variables across those assigned to financial assets and the control group. We restrict our attention to those for whom we have the 2015 vote outcome. As expected from stratified random assignment with low attrition rates, for almost all variables there are no significant differences across treatment and control. Individuals are very similar in their ex ante political positions, including their voting decision in the 2013 election and their likelihood of voting in Around 36% of our sample in both treatment and control groups reported having traded stocks in the six months prior to the experiment. Both groups show similar attitudes to equality, belief in the trustworthiness of others, and financial literacy, based upon a test that we administered (adapted from Van Rooij, Lusardi, and Alessie (2011)). They also are balanced by basic demographic characteristics, including sex, income, education and geographical location. There are two exceptions that show significant differences: those in our asset treatment are slightly older (41.25 vs 39.1 years old) and believe themselves slightly more willing to take risk ( an average of 4.69 on a 1-10 scale, compared to 4.36 for the control). 18 We control for age with both linear and quadratic terms in our regressions, and will show that our effects are in fact stronger on the risk-averse. [Table 1] 17 There was slightly higher attrition on the attitudes towards the peace process, we were able to get a response rate of 95% (1277/1345). 18 This difference is consistent with a separate measure of risk aversion we constructed based upon hypothetical lotteries. 14

15 4 Main Results We now turn to the main results, starting with the central question: whether exposure to financial markets changes votes in the 2015 elections and attitudes towards the peace process. Figure 2 provides the raw comparison of the vote shares between the control and those participants that received the asset treatments. The left panel shows vote shares in the 2013 elections (prior to our intervention) whereas the right panel shows the 2015 vote shares. Recall that we over-sampled centrist voters at twice their 2013 vote share, as the left panel suggests. Once again, both panels show only those 1309 participants whose 2015 voting decisions are known. Consistent with Table 1, treatment and control groups had very similar distributions of votes across left, right and center parties in However, even without controls, there are substantial differences in their voting decisions in % of the treatment voted for the left in 2015 compared to 24.6% of the control. This difference is mirrored on the right. Right parties won 31.3% of the votes of the treated group, relative to 35.9% of the control. [Figure 2] We will now show that these differences are robust. Recall that 840 out of 1036 (81%) assigned to the asset treatment actually received the assets (having completed the instructions and quiz). To be conservative, for the most part we will present Intent to Treat Estimates, comparing the control group to those that were experimentally assigned the treatment regardless of whether they actually received and traded in the assets. Later on, we will also report Treatment on the Treated measures, using instrumental variables estimates using assignment to treatment as an instrument for receiving assets. Table 2 presents the OLS effects of assignment to asset treatment on the probability of voting for the left in the 2015 elections. Column 1 shows the difference between treatment and control groups without controlling for other factors. Consistent with Figure 3, the asset treatment increases the probability of voting for the left by 6.4 percentage 15

16 points. Given that the 2015 vote share for the left among the control group was 24.6% (a proportion similar to the 25.3% overall vote share of Jewish left parties in the population in 2015), these are substantial effects. As seen in the histogram, even though we stratified on the 2013 vote, there are slight differences in the initial vote pattern in 2013 between the treatment and control. In Column 2, we therefore control for an individual s 2013 vote. In addition, we expect that those with experience of trading in shares prior to the experiment might vote differently. We thus also control for whether an individual had bought or sold shares with six months prior to the experiment. Notice that explanatory power (R 2 ) of the regression increases from to with the addition of these regressors. However, the estimated treatment effect remains fairly stable at 5.6 percentage points. It is also interesting to note that those that had invested prior to the experiment are already more likely to vote for the left by 10 percentage points even controlling for the treatment. Thus, it appears that those who were invested due to the treatment become more like existing investors in the political choices. Naturally, past financial market activity is not randomly assigned, and this correlation may in part reflect a number of other factors. However, the fact that existing investors and experimentally assigned new investors become aligned in their vote choice suggests that the effects on political choices and attitudes of even the temporary financial exposure that we provide are more likely to be persistent than if the vote choices of existing and experimentally assigned new investors diverged. Since our randomization was conducted within 104 stratification blocks, in Column 3, we add fixed effects for each block. Recall that these blocks were constructed to sequentially stratify upon the 2013 vote, sex, past trading of stocks, geographical region, survey discrepancies and risk aversion. Controlling for these blocks further increases the R-squared to 0.42, while the treatment effect remains at 5.4 percentage points. Column 4 presents the full specification. Along with past vote, trading history and fixed effects for each stratum, this includes further controls for differences in demographics 16

17 and economic preferences measured prior to the experimental intervention. These include sex, age (and age squared), four education categories (post-secondary, BA student, college graduate, relative to high school and below), four religiosity categories (traditional, religious and ultra-orthodox, relative to secular), seven regions (North, Haifa, Jerusalem, Tel Aviv, the West Bank and the South, relative to the Center), five income categories, an indicator for married, self-reported willingness to take risks (on a 1 to 10 scale), an indicator for patience above the median (inferred from a series of hypothetical choices) and the individual s financial literacy score (based upon the number of correct answers to a standard set of seven questions from Van Rooij, Lusardi, and Alessie (2011)). As the Column shows, many of these variables predict an individual s propensity to vote for the left in More religious individuals are less likely to vote for the left, whereas the educated are more likely to vote for the left. However, even controlling for these factors, the treatment effect remains at 6 percentage points. Experienced traders have naturally been exposed to financial markets and so one may expect them to be less likely to respond to the treatment. Further these individuals are also better positioned to be able to undo the treatment, selling assets so as to retain their preferred portfolios. In Columns 5 and 6, we therefore split the sample among the 35.8% that reported having traded within six months prior to the experiment and the majority that had not. As Column 5 reveals, the asset treatment had no effect on the propensity to vote left among those with trading experience (as we saw in Column 2, they were already more likely to do so by about 10 percentage points). Perhaps not surprisingly then, as Column 6 shows, most of the treatment effect seen earlier appears to come from those without trading experience, who are around 8.3 percentage points more likely to vote for the left. These differences in the treatment effects between experienced investors and the inexperienced also appear inconsistent with the presence of Hawthorne effects stemming from mere exposure to the study, that should therefore be common for all 17

18 treated participants. 19 [Table 2] Table 3 repeats the exercise examining the propensity to vote for the right (relative to the left or center). Overall, the patterns are similar and consistent, though the effect sizes are weaker: individuals exposed to the asset treatment are around 4 percentage points less likely to vote for the right across specifications. Experienced traders are once again 4 percentage points less likely to vote for the right, controlling for the treatment (Column 2), though here the point estimates of the treatment effect are similar for experienced investors and the inexperienced (Columns 5-6). 20 [Table 3] As Figure 2 shows, there were substantial changes in the vote shares of parties in the left, right and center in our sample between the 2013 and 2015 elections. These reflect not only changes due to the treatment effects but also trends related to potential voter polarization, the composition of parties and their ideological stances, among others. 21 To take account of these trends, we estimate the following difference-in-difference equation: Vote it = β 0 + β 1 AssetTreatment i + β 2 Post t + β 3 AssetTreatment Post it + α i + ɛ it (1) Where Vote it is an indicator for the vote of individual i in year t {2013, 2015}; Post t is an indicator for 2015, and ɛ it is an error term clustered at the individual level. In 19 As we discuss below, we also find effects on individuals engagement with financial news 4 months after the study ended. In a companion paper, we also show that there are effects on individuals financial literacy scores 4 months after the study, as well as reported investment and willingness to invest in specific types of stock in the future. 20 Using a multinomial logit specification, we find that the asset treatment significantly reduces the probability of voting for the right-of-center Likud and centrist Yesh Atid relative to the left-of-center Zionist Union (results not shown). 21 In particular, one of the main center parties in 2013, Hatnuah, created a joint list with the Labor Party. The centrist Kadima party disappeared. On the other side, Moshe Kahlon, a former member of the Likud, created a new centrist party called Kulanu. 18

19 some specifications, we will also include α i : an individual fixed effect. While β 1 captures any potential differences in the propensity to vote for a particular block across treatment and control and β 2 captures differences between the 2015 and 2013 elections, our main interest is in β 3 which captures the difference in the change in the vote between 2013 and 2015 for the treated individuals relative to the control. [Table 4] The first four columns of Table 4 present the effects on voting for the left, while the latter for four present the effects on the right. Consistent with the Figure 2, there appears to be have been a polarization of individuals propensity to vote between 2013 and 2015, with an increase in the vote share of both left and right by 12.3 percentage points and 11.3 percentage points respectively (Columns 1-2, 5-6). Consistent with random assignment, those treated with assets did not display an average inclination to vote for either left (Column 1, second row) or right (Column 5) until after actually being exposed just prior to the 2015 elections. Consistent with our previous findings, individuals treated with assets were 5 percentage points more likely to vote for the left, and 4 percentage points less likely to vote for the right (the latter effect is not precisely estimated). These differences remain precisely the same when we include individual level fixed effects (Columns 2 and 6). In Columns 3-4 and 7-8, we split the sample between experienced traders and nontraders. Once again, the non-traders show larger effect sizes (of 7.2 percentage points relative to 1.5 percentage points) on their propensity to vote for the left, though both show similar effects (of 4 percentage points) on their vote for the right. 4.1 Attitudes towards Peace So far we have shown that exposure to financial assets moves individuals votes in 2015 elections in favor of parties of the left seen as more supportive of the peace process. However, individuals may have voted for the different parties due to their positions on 19

20 issues other than the peace process. In order to evaluate whether this change in vote was due to increased support for peace, in the post-election survey, we also asked them the following series of questions: To what extent do you agree or disagree with each of the following criteria for settling the conflict between Israel and the Palestinians? [Possible responses were: disagree, tend to disagree, tend to agree, agree] A Two states for two peoples [57% agree/tend to agree] B The 1967 Borders will be the borders between the two countries with a possibility of land swaps [40% agree/tend to agree] C Jerusalem will be divided into two separate cities: Arab and Jewish [27% agree/tend to agree] D Palestinian refugees will receive adequate compensation and be allowed to return to the State of Palestine only [40% agree/tend to agree] Question A asks a broad question about support for a two state solution. Though there is some disagreement, more than 57% of our sample support this general principle. 22 However, Questions B through D ask more specific questions about the concessions individuals are willing to make to achieve peace. On these issues, those that tend to agree falls considerably, with only 27% supporting the splitting of Jerusalem and 40% supporting the principle of returning to the 1967 borders with potential land swaps, and an equal number supporting providing compensation and the right to return to Palestine for Palestinian refugees from the war and their descendants. [ Table 5 ] Below we show the effect of the asset treatment on an index composed of all four questions. However, it is illuminating to consider each individually as well. The first three columns of Table 5 present ordered probit estimates of the asset treatment on individual s 22 These questions are drawn from, and thus can be compared to the 2013 Index of Arab-Jewish Relations in Israel (Smooha, 2015). Among Jews in their representative sample, they find: support for the two state solution: 61.5%, 1967 borders with land swaps: 40.3%, the splitting of Jerusalem: 22.6% and right of return for refugees to Palestine: 48.2%. 20

21 responses on each potential peace arrangement on a four point scale including: disagree, tend to disagree, tend to agree and agree. Each regression includes the full set of controls from Table 2, Column 4. Column 1 includes the entire sample. Overall, we find a positive effect of the asset treatment on the extent to which individuals agree with the four principles underlying a potential peace agreement. The effects are stronger for the more specific and less widely accepted concessions (Panels B-D). Consistent with the results on the vote for the left, the effects on support for peace concessions are more pronounced among those that did not trade prior to the experiment (Columns 2 and 3). To help unpack and interpret these results, we present OLS estimates among non-traders of their propensity to choose agreement rather than disagreement (Column 4) and to voice strong disagreement, relative to the other three options, on each question (Column 5). Those without prior trading experience that were exposed to the asset treatment are 7 percentage points more likely than others with trading experience to support the two state solution. They also display small, but not precisely estimated increases in agreement for the other principles (Column 4). At the same time, those without prior trading experience are 7 percentage points less likely to voice strong disagreement to the two state solution, 8.7 percentage points less for the 1967 borders and 11 percentage points less likely to disagree to the return of refugees to Palestine (Column 5). [ Table 6 ] So far we have presented Intent to Treat Estimates. These estimates are appropriate if one is interested in the treatment effect taking into account that some individuals fail to actually participate. However, as discussed above, 19% of the individuals assigned to the asset treatment did not actually receive the assets and it is useful to measure the treatment effect on those that did. One advantage of our research design is that we were able to anonymously survey even those that did not receive the assets on their votes and attitudes towards peace, as well as using the responses they separately provided to the survey company on their 2015 votes. This allows us to measure the treatment effect 21

22 on the treated, using assignment to treatment as an instrument for treated status. We present these results for our main outcomes in Table 6. Along with vote for the left and the right, we create an index support for peace deals by simply summing over the four responses to each of the four peace principles above. This creates a scale from 4 to 16. Columns 1 and 2 of Table 6 presents the familiar Intent to Treat (OLS) and the Treatment Effect on the Treated (IV-TOT) effects for the full sample. Notice that, as one might expect, the effect sizes are higher for the latter. The treatment effect on those treated with assets rises to an increase of 7.4 percentage points on the probability of voting for the left, a decrease of 5.4 percentage points on the vote for the right, and step change towards agreement on the potential peace deals. As Columns 3 and 4 show, once again, these effects are larger for those without prior trading experience. A separate issue with the data is that, as discussed above, we over-sampled individuals who voted in 2013 for non-orthodox center parties (Yesh Atid, Hatnuah and Kadimah). These voters could be considered more likely to move in response to treatment than their more ideological counterparts. In Column 5, we reweigh the sample so as to reflect the actual (official) vote share of Jewish parties in The size of the estimated effect is lower (a 4.4 percentage point increase) for the probability of voting for the left, but is similar (a 5.1 percentage point decrease) for the vote for the right, and for support for peace deals (0.505). This seems consistent with a pattern in which the treatment moves individuals from the center to the left but from the right to the center. This would naturally result in a slight weakening in the estimated effect on the vote for the left when we reduce the weight on center voters. To summarize: we find that exposure to incentives to trade financial assets has a strong effect both on individuals voting decisions and upon their attitudes towards making potential concessions for peace. Being assigned to financial asset exposure raises an individual s likelihood to vote for left parties considered more supportive of the peace process by 5-6 percentage points, and reduces an individual s propensity to vote for the 22

23 right by around 4 percentage points. The effect of financial exposure on those that were actually exposed to treatment is 7.4 percentage points. Further, exposure to financial assets increases an individual s support for a two state solution and willingness to make specific concessions for peace. These effects are particularly accentuated among those individuals who lacked prior experience in the financial markets. In the next section, we examine why. 5 Mechanisms The effects of exposure to financial markets on votes and political attitudes may operate through a number of potential mechanisms. We assigned individuals incentives to trade in financial assets that exposed inexperienced investors to a novel set of risks: the risks faced by firms within the national economy and of the other nation that is party to the conflict. Though the value of the financial assets that we provide are themselves small relative to the wealth of the average Israeli, trading in these assets could make individuals learn about and take more account of the risks to the broader national economy when making decisions that also relate to the nation including their voting decisions and attitudes towards the peace process. There are three particularly interesting aspects of this potential mechanism. First, if the mechanism is through exposure generating human capital then this would explain why the treatment effect makes the newly exposed more like existing traders in their votes and attitudes. The human capital channel is particularly likely given that the sums we assign of $50 or $100 (the latter equivalent to around 1.5 days of extra wages on average) is unlikely to be large enough on its own to equalize other ex ante differences between existing investors and non-traders. Second, existing investors tend to be selected on a number of dimensions, including their preferences for risk, while those that did not trade prior to the experiment tend to be more risk averse. But precisely because exposure to risk has more influence on the risk averse, experimentally assigned exposure through 23

24 financial assets to the political risks of firms and the broader economy that stem from the unresolved conflict under the status quo is also likely to have greater effect on risk averse individuals than for those that have greater tolerance for risk. In fact, in the absence of financial market exposure, the risk averse may be more supportive of the status quo than the uncertainty that stems from attempts to make deals for peace. Third, if the main effect is coming from learning due to incentives to trade in the financial market, and if learning is more effective through tangible feedback, then the effects should be larger with early divestment which provides realized gains and losses just prior to the elections. Beyond these exposure effects, a second possible mechanism is that the effect is coming from the actual skin-in-the-game we provide among those who were assigned to hold financial assets on the day of the election. Insofar as individuals believe that their assets will benefit from a particular election outcome, this may provide a small additional incentive to vote for that party. Indeed, Zussman, Zussman, and Nielsen (2008) show that both the Israeli and Palestinian stock markets respond favorably to political events that foster peace like the Oslo Accords, and negatively to those that foster conflict. 23 To study whether even this small skin-in-the-game has such an effect, we exploit two aspects of the research design. First, we can compare those individuals who were randomly endowed with stocks initially to those that were assigned cash to trade for the Tel Aviv Index. Both were exposed to the stock market. But given that we limited trades to 10% of each individual s portfolio every week, those initially assigned stocks still were constrained to have more skin-in-the-game on election day than those assigned cash. Second, we can compare those participants who were assigned to be divested of their portfolios on March 12, prior to the March 17 elections, to the remaining two-thirds who were divested on April 2. Naturally, the early divesters were exposed to the financial markets but lack any experimentally assigned skin-in-the-game on the day of the elections, while the late 23 Indeed Appendix Table 2 shows that even between January 30 and March 18, Israeli stock prices responded negatively to opinion polls that predicted an increase in the vote share of the right-of-center Likud, and right parties more generally, with a (lesser) fall in Palestinian stock prices. 24

25 divesters still held some experimentally assigned assets on that day. A further important question we can examine is whether being assigned Palestinian assets has stronger effects than Israeli assets. Being put into a position where one benefits when members of an out-group also benefit may generate empathy [quote social psychology on common fate (Kramer and Brewer, 1986). We will test whether individuals begin to identify more with the outgroup s welfare in Table X. Further, if Palestinian stocks are perceived to be more sensitive to political developments, and moves towards peace in particular, then people assigned Palestinian stocks may both learn more about the risks of conflict and face enhanced skin the game incentives. Figure 3 shows the views on this issue among those assigned to the different assets in the study. Specifically, every week, individuals were asked the question: Various factors affect the success of a particular company. When you think about the performance of [asset], which of the following is most important? 1. Quality of the management of the [ company or stocks included in the index ] 2. Quality of the workers in the [ company or stocks included in the index ] and the terms of their employment 3. Conditions in the economy: price stability, the level of the national debt, the quality of regulation 4. Internal political conditions: the quality of the government and lack of corruption 5. Regional political conditions: normal relations with the neighbors and the lack of conflicts and wars. Notice that holders of Palestinian stocks are relatively more likely to view interstate relations and the threat of conflict as the most important determinants of their assets value (Panel A). 24 Internal political conditions are also considered relatively more important for Palestinian assets (Panel B). Notice that national economic conditions in particular are also seen as the most important determinant by a large majority of Israeli assetholders, and close to half of Palestinian assetholders as well (Panel C). Relatively 24 Holders of the index funds, both Palestinian and Israeli, also tend to do so more than holders of individual company stock. 25

26 few chose the other options: the quality of management of the firm(s) and the quality of workers and their employment conditions. [Figure 3 ] [ Table 7] Table 7 unpacks the effects of different sub-treatments, examining how they affect the vote for the left (Columns 1-4) and the vote for the right (Columns 5-8). Recall that the average effect of the asset treatment raised the probability of voting left by 6 percentage points and decreased the vote for the right by 4.4 percentage points. Columns 1 and 5 compare the effects of being initially assigned stock or being assigned cash to trade for the Tel Aviv 25. Those assigned stock were on average 5.4 percentage points more likely to vote for the left and 3.7 percentage points less likely to vote for the right. In contrast, those that were initially assigned cash were 8.5 percentage points more likely to vote left and 7.2 percentage points less likely to vote right. This already suggests that the exposure effect seems to be greater than the skin-in-the-game channel. Columns 2 and 6 compare the effects of being initially assigned stock and cash endowments depending on whether the individual was divested of their portfolio before or after the elections. Observe that the stock treatment has similar effects on average âe of around 5 percentage points on vote for the left whether or not an individual actually held stocks on election day (Column 2). This suggests that there is little incremental skin-in-the-game effect on average, and much of the effect is coming from exposure. In fact, those endowed with stock were 7.1 percentage points less likely to vote for the right if they were divested before the elections and only 2 percentages points less if they were divested afterwards (Column 6). Further, individuals assigned cash were 15.2 percentage points more likely to vote for the left and 13 percentage points less likely to vote for the right if they were divested early, compared to 5.4 and 4.5 respectively among those who retained skin-in-the-game. These results again favor the exposure mechanism, and in fact suggest that there may be 26

27 an added benefit to learning from having received realized gains from one s investments prior to election day. Columns 3 and 7 further separate these effects by the nationality of the assets. Palestinian asset exposure raises the probability of voting left by 8.6 percentage points and lowers that of voting right by 8.5 percentage points among early divesters. The effect is substantially smaller among those divested after the elections, however. 25 This suggests that, for Palestinian stocks, the effect is primarily due to exposure rather than skin-inthe-game, and in fact, the realized gains prior to the elections appear to have had a major impact on voting. In contrast, those endowed with Israeli stock, the effect on vote for the left is greater for those that held assets on election day compared to early divesters (9 percentage points versus 2 percentage points, though not significantly different). The reduction in voting for the right appears to be similar (close to 5 percentage points) across groups. Taken together, these results suggest that the effect comes primarily from exposure than from the skin-in-the-game, with the effects tending to be larger for realized gains. As seen in Figure 2, Israeli stocks, including the Tel Aviv Index out-performed Palestinian stocks on the eve of the elections. Thus both those assigned to Israeli stock endowments and to cash that they could use to buy the TA-25 were exposed to greater price increases. The price changes themselves may have also had an independent effect on the intensity of the treatment. Columns 4 and 8 add a control for the price change in each individual s asset on the day of the election. Though the price change itself seems to have little direct effect on voting, controlling for the price change slightly strengthens the effect of the Palestinian treatment and slightly attenuates the effects of the Israeli and cash treatments. So far we have found limited evidence for the effect of skin-in-the-game. However this may be because the comparison we have been making has been between holding no experimentally assigned assets on the day of the election and some. Instead, it is plausible 25 the difference is significant at the 7% and 2% levels on left and right respectively 27

28 that there will be more of an effect among those that hold more assets on the day of the election. Naturally, given the small size of our intervention, the greatest exposure individuals would have is also limited, but it is informative to see even within our study, whether we can detect such effects. One issue is, of course, that an individual s holdings on the day of the election is endogenous to their trading decisions. We can, however, construct an instrument using the experimentally assigned and exogenous components of the individual s portfolios. Specifically, we can calculate for each individual the stock value on election day if that individual made no trades and kept the initial allocation (i.e. the stock value of a purely passive investor). The instrument is calculated based on the asset allocation, the redemption date (pre- or post- elections), the initial value (high or low) and the price change of the specific asset by election day. [Table 8] Columns (1-2) and (5-6) of Table 8 examine whether there is a separate effect from the asset treatment of the actual value of stock holdings on election day on the decision to vote left and right, both using the actual stock value directly (Columns 1 and 5) and instrumenting for this value using the passive investor instrument. Notice that there does not seem to be a separate effect of the actual asset holdings beyond the initial exposure. This is also true breaking the asset treatment down into Palestinian, Israeli and Cash treatments (Columns 3-4, 7-8). 26 [ Table 9] 5.1 A Closer Look at the Exposure Effect The results in Tables 7 and 8 suggest a limited role for a skin-in-the-game effect. We now focus on the channels through which exposure may have had an influence. Providing 26 Note that a passive investor assigned cash would not have any holdings at all, so their portfolio is equivalent to their initial assignment. Thus we cannot measure both separately. 28

29 individuals incentives to trade in the market also provides them incentives to learn about the determinants of the performance of those markets and thus to follow financial news. At the same time, if stock market exposure leads individuals to be more cognizant of economic risks, they will have further incentive to follow financial news even after the study. In the follow up survey, four months after the elections, we asked individuals about which news sources they regularly read, giving them a list of major Israeli newspapers. These included both financial and general news outlets. Table 9, Panel A reports the coefficients on the asset treatment in the full specification on the number of financial and non-financial outlets that individuals claim to read regularly. As the Panel A reveals, individuals assigned the asset treatment indeed read more financial newspapers even threefour months after the experiment. They did not significantly change the extent to which they read other news sources. Nor does it appear to be the case that, despite treated individuals voting more the left in the elections, that they switched to reading less rightwing newspapers (such as Sheldon Adelson s Israel Hayom) or more left wing sources (like Haaretz). We will come back to this when we discuss alternative mechanisms. We have argued that exposure to financial markets lead individuals to take more account of risks to the national economy. But although conflict generally tends to increase economic risks, parties on both sides of the spectrum may claim to be supportive of economic growth and reduced security risks. Thus, it is not obvious that taking more account of national risks would lead individuals to change their support for the peace process or to vote for the left. We therefore examine whether individuals exposed to financial markets not only increase their consumption of financial media but also change their views on the relative costs, benefits and risks of continued conflict under the status quo against those from a peace settlement. We test this directly by asking individuals how a peace settlement would affect Israel s economy, Israel s security, their personal safety and their personal economic condition. Panel B shows the ordered probit effects of asset exposure on these outcomes, based upon a five point scale (will worsen substantially, will worsen, 29

30 have no effect, improve, improve substantially). Notice that on average, though treated individuals do not expect to benefit personally from a two state solution, they are more likely to think that Israel s economy will benefit from a peace settlement. [Table 10] As discussed above, another implication of the the role of financial market exposure resulting in individuals taking more account of national risks is that the effects should be larger for the risk averse, for whom these risks become more influential. Table 10 shows OLS regressions of vote, support for potential peace deals and predictions of the potential effects of a peace settlement, including a term for whether individuals reported that they were at the median or lower in their willingness to take risks at baseline as well as an interaction term between this measure of risk aversion and asset market exposure. Observe that, risk averse individuals are significantly more likely to oppose peace deals (second row, Columns 3-7) and tend to have a less optimistic view of the outcome of a peace settlement (Columns 8-11). However, across the board, the effects of the asset treatment on the risk averse are greater. Risk averse individuals that were experimentally exposed to the financial markets are relatively more likely to support concessions for peace, and are more likely to predict that a peace settlement will improve Israel s security, Israel s economy, their own safety and, to some extent, their own economic situation. The differential effects of the risk averse are consistent with the role of the effect of exposure to financial markets operating through making individuals take more account of broader risks. 5.2 Alternative Mechanisms In this section, we evaluate a number of alternative channels that might explain parts of our results, including political information, a wealth effect and changes in happiness or social well-being. It is useful to note that for each of these alternatives The first possibility is that individuals are incentivized to trade and and in learning about financial 30

31 markets, they become more exposed to new sources of information on politics and platforms which might change their vote directly. Note that, even if this was the case, it is not obvious that the proposed channel would lead to increased propensities to vote for the left rather the right. However, to evaluate this, one month after the elections, we gave treatment and control subjects a battery of questions to test their political and economic knowledge. There were 13 multiple choice political knowledge questions. These included questions about the positions of the candidates (What is Herzog s position concerning the establishment of a Palestinian state as part of a political settlement?), events during the run-up to the elections (What was the main subject of Netanyahu s Congress speech?), and simple factual questions (Who was Minister of Defense in the previous government (until December 2014)?). We also included five questions on economic knowledge (e.g. What is the official unemployment rate today? What was the inflation rate the past year? Did the stock market rise or fall in March? By what percent?). We score these questions to obtain a political knowledge score and an economic knowledge score. 27 Table 9 Panel C shows the effects of the asset treatment on individuals political and economic scores. As the results show, there appears to have been no effect on their knowledge in these dimensions. The one exception, perhaps not surprisingly, is that individuals exposed to the financial markets through our experiment had more accurate knowledge of the performance of the Israeli stock market, with treated individuals without prior trading experience being 9 percentage points more likely than inexperienced members of the control provide an accurate answer to this question. These results indicate that the channel was not through improving broader economic or political knowledge but through those specific to the financial markets. Another alternative channel through which our treatment might have effected behaviour is the direct effect of having giving individuals up to $100, making them a little bit more wealthy. The argument might then be that wealthy people prefer peace and this alters their votes and attitudes in favor of the left. 27 For multiple choice questions, there is one correct answer. For numerical questions, we code an answer as correct if it is within 3 percentage points of the correct statistic. 31

32 On the other hand, increased wealth might make individuals more concerned about the threat of redistribution, leading them to support the right. 28 [Table 11] Observe that our treatment is a combination of the incentives to trade and the allocation of assets. To completely separate out a wealth effect from our, we would have had to allocated funds to a pure wealth treatment without the trading intervention. We did not do this, both for budgetary considerations and because it is unlikely that the small initial amounts we provide 200 NIS ($50) for the low treatment and 400 NIS ($100) for the high treatment have a substantial enough effect of change the wealth of our individuals enough to influence their votes and attitudes (for reference, the average daily wage in Israel was NIS 312 in December 2014). However, we are able to test for aspects of a potential wealth channel in a number of ways. First, recall that the effects that we find influence non-traders rather than those already involved in trading in stock, even comparing individuals with the same initial family incomes. If it was merely the effect of endowing individuals with the amounts that provided, it should have affected everyone treated more uniformly. Second, if the effect of the asset treatment is coming from a wealth effect, we should expect the effect to be larger for those that are relatively poor. Table 11, Columns 1 and 5 shows the coefficient of the interaction of the asset treatment with an individual reporting below average household income prior to the experiment. Notice that there is no significant difference in the effect for the relatively poor, and the point estimates suggest that, if anything the effects are slightly smaller for this group. Another way to test for a wealth effect is to exploit the fact that in our design, we randomly assigned half our treated participants to a high initial value of $50. Naturally, the high allocation might also may magnify other effects, beyond the wealth effect, by 28 Indeed, a number of studies find in non-experimental settings that the provision of assets leads individuals to become more conservative and right wing, not less (see eg Kaustia, Knüpfer, and Torstila (forthcoming)). A similar logic may also lead right-wing governments to favour privatization (Biais and Perotti, 2002). 32

33 raising skin-in-the-game and potentially focusing individuals attention more financial market risks. However, if our results purely reflect a wealth effect it would suggest that the effects should always be greater for those in the high treatment condition. Columns 2 and 6 show, however, that the effects of being assigned to the high treatment are at best, mixed. Individuals randomly assigned to the high treatment actually are slightly less likely (by 1 percentage point) to vote for the left than those assigned to the low treatment, but they do respond more on the right reducing their likelihood of a right vote by 2.3 percentage points more than those in the low asset treatment. Columns 3 and 7 show that there is limited additional effect for individuals receiving the high stock or high cash treatment relative to their low counterparts, though those differences that do emerge that parallel the direction of our main effects come from a slightly greater reduction of the probability of voting right on average of those initially endowed with a high allocation of stocks rather than cash. Columns 4 and 8 unpack this further and show that individuals assigned to a high allocation of Palestinian stocks, in particular, responded significantly more in lowering their probability of voting for the right (by 8 percentage points relative to the control). On the probability of voting for the left however, those with a high allocation of Palestinian stocks show about the same increases as those with the low allocation. Individuals assigned the high allocation of Israeli stocks, in contrast, tend to respond less than those with the low allocation. Taken together, these results do not appear consistent with an overall wealth effect and suggest instead that a high allocation appears to be magnifying other effects, particularly that of exposure to Palestinian stocks, rather than a pure wealth effect per se. Another way to examine to test for a wealth effect is to test whether it influences individuals well-being in a measureable way. It may be that gaining the assets we provide, even if it is a small amount, makes people happy and happy people favor peace. To address this question, we asked individuals not only about their overall well-being but also a battery of the top 10 main predictors of individual well-being based upon Benjamin, 33

34 Heffetz, Kimball, and Szembrot (2014, Table 2). 29 As Table 9 Panel D suggests, however, the asset treatment not only failed to affect individual s overall life satisfaction but also did not significantly affect any of these other indicators. 6 Discussion This is the first paper to measure the effects of providing incentives for individuals to trade in the stock market on their attitudes towards peace and individuals electoral choices. We find that exposure to incentives to hold and trade stock market assets raises individuals support for peaceful compromise, changing their evaluation of the effects of national platforms and even their votes. These effects are accentuated for those who lacked prior exposure to the stock market and those that are risk-averse. These results are, we argue, consistent with the role of exposure to financial markets in helping people take more and more precise account of national risks in making decisions on national policy issues. As with any novel study, replications and extensions of this work are highly desirable to flesh out the mechanisms more fully. For example, we find limited effects of the direct skin-in-the-game incentive provided by the stock market in our intervention, except for participants endowed with Israeli stocks. This may reflect the relative familiarity that individuals already possess with the Israeli stock market relative to Palestinian assets, making those with skin-in-the-game relatively more interested in the Israeli stock market than those without. However, we cannot rule out that our intervention simply provided too small a direct incentive, a topic we explore in future research. We also disallow short sales to avoid potentially fanning the flames in an already fissile situation, but measuring the political effects of such design elements is also a topic we intend to explore. In companion research we also measure the impact of financial market exposure 29 We excluded the mental health question as this would have been considered intrusive in the cultural context. 34

35 on subsequent financial literacy, willingness to invest in the stock market, and social attitudes, as well as behavioural differences that arise from being exposed to specific company stocks versus more-diversified national indices. The grave costs of the Israel-Palestinian conflict mean that there is avowed support among moderate policymakers on both the left and the right of the ideological spectrum for securing peace. 30 Yet, there remains substantial disagreement on the feasibility of and risks involved in maintaining the status quo versus making peace concessions. Our paper suggests that part of this may stem from risk aversion and limited awareness of the relative effects of status quo conflict on national risks. Our study therefore underlines a largely under-explored mechanism exposure to financial markets as a path for policymakers from both sides to align people s incentives with the risk and returns of the broader economy in a way that may allow increased political space for peaceful compromise. More generally, the main thrust of contemporary policy suggestions in areas of persistent ethnic conflict like that between Israelis and Palestinians have focused upon international peacekeeping. 31 Our results suggest, however, a fruitful role for financial instruments as well. One intriguing possibility, for example, is that rather than focusing on providing aid in kind to populations in conflict zones, donors could also provide individuals from groups in conflict resources earmarked to invest in stock in their national exchanges. This may lead them to internalize and take more account of the gains and risks of peacemaking and conflict to society more generally. In so doing, financial exposure may be conducive to building lasting peace. References Alesina, A., and E. La Ferrara (2000): Participation in heterogeneous communities, Quarterly Journal of Economics, 115(3), As the Prime Minister of Israel, Benjamin Netanyahu, has asserted: We want peace. I want peace. We crave for peace. We greet each other with the word peace Shalom.... I think we can get it. It s hard but we have to pursue it. Interview with Peter Greenberg, 2014, Israel: the Royal Tour, 8:59-9:20, PBS. 31 At the local level, several initiatives also focus on building civil society organizations and complementarity in tourism and other industries. 35

36 (2005): Ethnic diversity and economic performance, Journal of Economic Literature, 43(3), Benjamin, D. J., J. J. Choi, and J. Strickland (2010): Social Identity and Preferences, American Economic Review, 100, Benjamin, D. J., O. Heffetz, M. S. Kimball, and N. Szembrot (2014): Beyond Happiness and Satisfaction: Toward Well-Being Indices Based on Stated Preference, American Economic Review, 104(9), Biais, B., and E. Perotti (2002): Machiavellian Privatization, American Economic Review, 92(1), Bisin, A., and T. Verdier (2011): The Economics of Cultural Transmission and Socialization, in Handbook of Social Economics, ed. by J. Benhabib, A. Bisin, and M. O. Jackson, chap. 9. North Holland. Blattman, C., and E. Miguel (2010): Civil War, Journal of Economic Literature. Bonaparte, Y., and A. Kumar (2013): Political activism, information costs and stock market participation, Journal of Financial Economics, 107, Boycko, M., A. Shleifer, and R. Vishny (1994): Voucher privatization, Journal of Financial Economics, 35(2), Brav, A., W. Jiang, F. Partnoy, and R. Thomas (2008): Hedge Fund Activism, Corporate Governance and Firm Performance, Journal of Finance, 63(4), Bursztyn, L., F. Ederer, B. Ferman, and N. Yuchtman (2014): Understanding mechanisms underlying peer effects: evidence from a field experiment on financial decisions, Econometrica, 82(4), Coval, J., and T. Moskowitz (1999): Home bias at home: local equity preference in domestic portfolios, Journal of Finance, 65(6), Dohmen, T., A. Falk, D. Huffman, U. Sunde, J. Chupp, and G. G. Wagner (2011): Individual Risk Attitudes: Measurement, Determinants and Behavioral Consequences, 9(3), Eckstein, Z., and D. Tsiddon (2004): Macroeconomic consequences of terror: theory and the case of Israel, Journal of Monetary Economics, 51, Fearon, J. D. (1996): Bargaining over objects that influence future bargaining power., Paper presented APSA general meetings. Grinblatt, M., and M. Keloharju (2001): How Distance, Language and Culture Influence Stockholdings and Trades, Journal of Finance, 56(3), Guidolin, M., and E. L. Ferrara (2007): Diamonds are Forever, Wars are Not: Is Conflict Bad for Private Firms?, American Economic Review, 97(5),

37 Haber, S., and E. Perotti (2008): The political economy of financial systems, Discussion paper, Tinbergen Institute Discussion Paper. Hirschman, A. O. (1977): The passions and the interests: political arguments for capitalism before its triumph. Princeton University Press, Princeton, NJ. Hong, H., and L. Kostovetsky (2012): Red and Blue Investing: Value and Finance, Journal of Financial Economics, 103, Huberman, G. (2001): Familiarity Breeds Investment, Review of Financial Studies, 14(3), Jha, S. (2012): Sharing the Future: Financial Innovation and Innovators in Solving the Political Economy Challenges of Development, in Institutions and Comparative Economic Development, ed. by M. Aoki, T. Kuran, and G. Roland. IEA Conference Proceedings 150: Palgrave Macmillan. (2013a): Can Financial Innovations Mitigate Ethnic and Civil Conflict?, World Financial Review. (2013b): Trade, Institutions and Ethnic Tolerance: Evidence from South Asia, American Political Science Review, 107(4), (2014): Unfinished Business : Historic Complementarities, Political Competition and Ethnic Violence in Gujarat, Journal of Economic Behaviour and Organisation, 104, (2015): Financial Asset Holdings and Political Attitudes: Evidence from Revolutionary England, Quarterly Journal of Economics, 103(3), Jha, S., K. Mitchener, and M. Takashima (in progress): Swords into Bank Shares: Finance, Conflict and Political Reform in Meiji Japan,. Jones, S. L., W. L. Megginson, R. C. Nash, and J. M. Netter (1999): Share Issue Privatizations as Financial Means to Political and Economic Ends, Journal of Financial Economics, 53, Kaustia, M., S. Knüpfer, and S. Torstila (forthcoming): Stock Ownership and Political Behavior: Evidence from Demutualizations, Management Science. Kramer, R. M., and M. B. Brewer (1986): Social group identity and the emergence of cooperation in resource conservation dilemmas, in Psychology of Decisions and Conflict, Vol 3: Experimental Social Dilemmas, ed. by M. D. Wilke, H., and C. Rutte, pp Verlag Peter Lang, Frankfurt. Levine, R. (2005): Finance and Growth: Theory and Evidence, in Handbook of Economic Growth, ed. by P. Aghion, and S. Durlauf, chap. 12. Elsevier. Manekin, D., G. Grossman, and T. Mitts (2015): Symbolism or Materialism? A Public Opinion Approach to Territorial Conflict, mimeo, Penn. 37

38 Megginson, W. L., and J. M. Netter (2001): From State to Market: A survey of empirical studies on Privatization, Journal of Economic Literature, 39(2), Merton, R. (1987): A simple model of capital market equilibrium with incomplete information, Journal of Finance, (42), Montalvo, J. G., and M. Reynal-Querol (2005): Ethnic polarization, potential conflict and civil wars, American Economic Review, 95(3), Montesquieu, C. d. S. (1748): The Spirit of the Laws. Cambridge University Press, 1989 edn. Nunn, N., and L. Wantchekon (2011): The Slave Trade and the Origins of Mistrust in Africa, American Economic Review, 101(7), Rajan, R. G., and L. Zingales (2003): Saving capitalism from the capitalists: unleashing the power of financial markets to create wealth and spread opportunity. Crown business, New York. Reuters (2015): Netanyahu says no Palestinian state as long as he s prime minister, Rousseau, P. L., and R. Sylla (2008): Financial Systems, Economic Growth and Globalization, in Globalization in Historical Perspective, ed. by M. Bordo, A. Taylor, and J. Williamson. University of Chicago Press, Chicago. Sambanis, N., and M. Shayo (2013): Social Identification and Ethnic Conflict, American Political Science Review, 107(2), Shayo, M., and A. Zussman (2014): Conflict and the Persistence of Ethnic Bias, mimeo, Hebrew University of Jerusalem. Smooha, S. (2015): Still Playing by the Rules: 2013 Index of Arab-Jewish Relations in Israel. University of Haifa. Van Rooij, M., A. Lusardi, and R. Alessie (2011): Financial literacy and stock market participation, Journal of Financial Economics, 101, Voigtländer, N., and J. Voth (2012): Persecution Perpetuated: The Medieval Origins of Anti-Semitic Violence in Nazi Germany, Quarterly Journal of Economics, 127(3), World Bank (2011): World Development Report: Conflict, Security and Development. World Bank, Washington, DC. Zussman, A., N. Zussman, and M. O. Nielsen (2008): Asset Market Perspectives on the Israeli-Palestinian Conflict, Economica, 75,

39 Table 1: Summary Statistics and Covariate Balance Variable Control Asset Treatment Palest. Stock Israeli Stock Cash Late Divest N Voted Right ' (0.429) (0.428) (0.425) (0.432) (0.427) (0.431) Voted Left ' (0.340) (0.344) (0.347) (0.341) (0.344) (0.342) Bought/Sold Shares in Last 6 Mths [0/1] (0.480) (0.479) (0.481) (0.473) (0.486) (0.478) Willingness to Take Risks [1-10] ** 4.744** 4.715** *** (2.262) (2.265) (2.317) (2.183) (2.328) (2.287) Time preference median or above (0.476) (0.475) (0.470) (0.483) (0.467) (0.479) Financial Literacy Test: Correct Answers [0-7] (1.644) (1.636) (1.665) (1.583) (1.688) (1.654) Male (0.500) (0.500) (0.499) (0.501) (0.499) (0.500) Age [Yrs] ** * *** ** (13.634) (13.389) (13.422) (13.610) (12.783) (13.365) Married (0.489) (0.491) (0.495) (0.486) (0.491) (0.493) Education [Yrs] (2.032) (1.978) (2.006) (1.929) (2.024) (2.040) Religiosity: Secular (0.483) (0.484) (0.486) (0.483) (0.482) (0.488) Traditional (0.372) (0.370) (0.380) (0.368) (0.354) (0.383) Religious (0.329) (0.330) (0.334) (0.318) (0.344) (0.339) Ultra-Orthodox (0.274) (0.278) (0.269) (0.289) (0.273) (0.270) Region: Jerusalem (0.289) (0.287) (0.284) (0.286) (0.295) (0.298) North (0.294) (0.296) (0.281) (0.289) (0.339) (0.292) Haifa (0.345) (0.349) (0.347) (0.346) () (0.351) Center (0.455) (0.454) (0.454) (0.461) (0.441) (0.454) Tel Aviv (0.399) (0.396) (0.405) (0.384) (0.403) (0.399) South (0.309) (0.306) (0.326) (0.306) (0.257) (0.302) West Bank (0.268) (0.274) (0.253) (0.289) (0.281) (0.265) Monthly Family Income [NIS] (Mid-Point of SES categories) (5510) (5564) (5443) (5556) (5818) (5486) This table provides the mean and standard deviations in parentheses by treatment arm for the 1319 individuals for whom we observe their 2015 vote. We also indicate whether a univariate regression coefficient shows significant difference vs control at 10%, **:5%, ***1%; For attitudinal questions, 1 indicates "agree or strongly agree". + : We only have the income categories for 1319 individuals, but use a categorical variable, including one for ` missing', in the regressions. 39

40 Table 2: Financial Assets Raise Vote for the Left in 2015 (1) (2) (3) (4) (5) (6) Sample Full Full Full Full Experienced Inexperienced Asset Treatment 0.064** 0.056** 0.054** 0.060** *** (0.029) (0.024) (0.023) (0.023) (0.045) (0.029) Voted Right ' *** * *** * (0.018) (0.105) (0.091) () (0.105) Voted Left ' *** 0.695*** 0.596*** 0.533** 0.659*** (0.032) (0.094) (0.091) (0.243) (0.100) Bought/Sold Shares in Last 6 Mths [0/1] 0.099*** (0.023) (0.041) (0.040) Traditional *** ** *** (0.032) (0.059) (0.039) Religious *** *** *** (0.032) (0.059) (0.040) Ultra-Orthodox *** *** *** (0.039) (0.088) (0.046) Post Secondary Education 0.067** ** (0.033) (0.081) (0.037) BA Student 0.088** *** (0.038) (0.086) (0.043) BA Graduate and Above 0.062** ** (0.030) (0.068) (0.036) Willing to Take Risks [1-10] (0.005) (0.010) (0.006) Time preference above median (0.022) (0.041) (0.026) Financial Literacy Test: Correct Answers [0-7] * (0.007) (0.016) (0.008) Control for Vote in 2013 No Yes Yes Yes Yes Yes Strata Fixed Effects No No Yes Yes Yes Yes Demographic Controls No No No Yes Yes Yes Observations 1,309 1,309 1,309 1, R-squared Notes : This table provides OLS estimates of the effect of exposure to the asset treatment on the probability that an individual voted for a left party in Regressions are estimated by OLS. Robust standard errors in parentheses ***1%, **5%, *10% Left parties include the Zionist Union, Meretz & the Arab Joint List. Center parties:yesh Atid, Kulanu, Shas and Yahadut HaTorah. Right parties: Likud, Habayit Hayehudi, Israel Beytenu & Yachad. In Column 3, we include 104 stratification block intercepts constructed to stratify sequentially on: 2013 vote, sex, traded stocks, geographical region, discrepancies in 2013 vote across surveys and risk aversion. Column 4 adds controls for demographics and ex ante preferences, that include sex, age, education categories, six regional dummies, income categories, marital status, religiosity categories, time preference above the median, financial literacy score and subjective willingness to take risks.. Columns 5 and 6 show the differential effects on the subsample of those that had traded 6 months before the experiment (Column 5) and those that had not (Column 6). 40

41 Table 3: Financial Assets (More Weakly) Lower Vote for the Right in 2015 (1) (2) (3) (4) (5) (6) Sample Full Full Full Full Experienced Inexperienced Asset Treatment * * * (0.031) (0.025) (0.024) (0.024) (0.041) (0.031) Voted Right ' *** 0.525*** 0.492*** (0.025) (0.113) (0.122) (0.357) (0.174) Voted Left ' *** *** ** *** (0.022) (0.096) (0.088) (0.198) () Bought/Sold Shares in Last 6 Mths [0/1] ** (0.021) (0.040) (0.040) Traditional 0.102*** 0.149*** 0.085** (0.032) (0.055) (0.042) Religious *** 0.354*** 0.174*** (0.049) (0.089) (0.061) Ultra-Orthodox (0.086) (0.201) (0.101) Post Secondary Education * (0.034) (0.072) (0.040) BA Student (0.039) (0.079) (0.046) BA Graduate and Above ** (0.032) (0.062) (0.038) Willing to Take Risks [1-10] *** (0.005) (0.009) (0.006) Time preference above median (0.021) (0.040) (0.026) Financial Literacy Test: Correct Answers [0-7] * (0.007) (0.014) (0.009) Control for Vote in 2013 No Yes Yes Yes Yes Yes Strata Fixed Effects No No Yes Yes Yes Yes Demographic Controls No No No Yes Yes Yes Observations 1,309 1,309 1,309 1, R-squared Notes : This table provides OLS estimates of the effect of exposure to the asset treatment on the probability that an individual voted for a right party in Regressions are estimated by OLS. Robust standard errors in parentheses ***1%, **5%, *10% Left parties include the Zionist Union, Meretz & the Arab Joint List. Center parties:yesh Atid, Kulanu, Shas and Yahadut HaTorah. Right parties: Likud, Habayit Hayehudi, Israel Beytenu & Yachad. In Column 3, we include 104 stratification block intercepts constructed to stratify sequentiallyon: 2013 vote, sex, traded stocks, geographical region, discrepancies in 2013 vote across surveys and risk aversion. Column 4 adds controls for demographics and ex ante preferences, that include sex, age, education categories, six regional dummies, income categories, marital status, religiosity categories, time preference above the median, financial literacy score and subjective willingness to take risks.. Columns 5 and 6 show the differential effects on the subsample of those that had traded 6 months before the experiment (Column 6) and those that had not (Column 7). 41

42 Table 4: Difference in Difference Estimates on 2015 Vote Share Voted for Left Voted for Right Sample All All Experienced Inexperienced All All Experienced Inexperienced (1) (2) (3) (4) (5) (6) (7) (8) Indicator for 2015 *** *** 0.225*** 0.063** 0.113*** 0.113*** 0.090** 0.126*** (0.023) (0.023) (0.040) (0.026) (0.024) (0.024) (0.041) (0.028) Asset Treatment (0.022) (0.028) Asset Treatment x * 0.050* ** (0.026) (0.026) (0.047) (0.031) (0.026) (0.026) (0.046) (0.032) Individual FE No Yes Yes Yes No Yes Yes Yes Number of Individuals 1,309 1, ,309 1, Observations 2,618 2, ,680 2,618 2, ,680 R-squared Notes : This table provides Difference- in - Difference estimates of the effect of exposure to the asset treatment on the probability that an individual voted for a left party in 2015 (Columns 1-4) or a right party (Columns 5-8) in Regressions are estimated by OLS. Robust standard errors, clustered at the individual level, are in parentheses ***1%, **5%, *10% Left parties include the Zionist Union, Meretz & the Arab Joint List. Center parties:yesh Atid, Kulanu, Shas and Yahadut HaTorah. Right parties: Likud, Habayit Hayehudi, Israel Beytenu & Yachad. In Columns 1 and 4, we provide an intercept for 2015, which shows that both votes for the left and right increase relative to 2013, and a placebo comparison: those treated with assets in their pre-treatment vote in Columns 2 and 6 control for individual fixed effects. Columns 3-4 and 7-8 compare the effects in the subsamples of those that had and had not traded assets within 6 months before the experiment. 42

43 Table 5: Financial Assets Raise Support for Peace Concessions Estimator Ordered Probit OLS: 1 if Agree / Tend to Agree OLS: 1 if Disagree Sample All Experienced Inexperienced Inexperienced Inexperienced (1) (2) (3) (4) (5) A. Two states for two peoples (Mean = 2.522, SD = 1.14) Asset Treatment ** 0.070* ** (0.079) (0.142) (0.102) (0.036) (0.034) Observations 1, R 2 / Pseudo R B. The 1967 Borders will be the borders between the two countries with a possibility of land swaps (Mean = 2.164, SD = 1.083) Asset Treatment 0.164** *** ** (0.079) (0.141) (0.102) (0.037) (0.036) Observations 1, R 2 / Pseudo R C. Jerusalem will be divided into two separate cities: Arab and Jewish (Mean = 1.822, SD = 1.039) Asset Treatment 0.189** * (0.086) (0.148) (0.110) (0.033) (0.037) Observations 1, R 2 / Pseudo R D. Palestinian refugees will receive adequate compensation and be allowed to return to the State of Palestine only (Mean=2.135, SD=1.075) Asset Treatment 0.194** *** *** (0.077) (0.139) (0.099) (0.041) (0.042) Observations 1, R 2 / Pseudo R This table shows the effect of financial asset treatment on an individual's support for potential concessions for peace. Columns 1-3 show the coefficient of Ordered Probit regressions of whether individuals disagree (1), tend to disagree (2), tend to agree (3) and agree (4) with each potential peace concession on asset treatment. Columns 3-5 subset the data to those individuals who had not traded 6 months prior to the experiment. For interpretability Column 4 shows OLS effects on the probability of tend to agree or agree. Column 5 estimates the OLS effect on the probability of choosing disagree. All regressions control for the full set of demographics and randomization strata from Table 2, Col. 4. Robust standard errors in parentheses. ***1%, **5%, *10%. 43

44 Table 6: Treatment Effect on the Treated and Reweighted Estimates Estimator OLS IV (Treatment Effect On the Treated) OLS Sample All All Experienced Inexperienced Reweighted (1) (2) (3) (4) (5) A. Left Vote 2015 Asset Treatment 0.060** 0.074** *** (0.023) (0.029) (0.053) (0.037) (0.026) Observations 1,309 1, ,309 R-squared B. Right Vote 2015 Asset Treatment * * ** (0.024) (0.029) (0.048) (0.039) (0.023) Observations 1,309 1, ,309 R-squared C. Peace Deal Support Scale Asset Treatment 0.467** 0.573** *** 0.505** (0.190) (0.233) (0.422) (0.288) (0.198) Observations 1,277 1, ,277 R-squared This table compares the Intent To Treat estimates of the effect of asset exposure (Column 1) to estimates of the Treatment Effect on the Treated (Columns 2-4). This estimate uses assignment to treatment as an instrument for actually being treated with asset exposure. Right Vote and Left Vote are indicators for vote in the 2015 elections. The Peace Deal Support Scale is the sum of the answers to the four peace deal questions and takes values from 4 to 16. Column 5 reweighs the data by the actual vote share for each Jewish party in All regressions control for the full set of demographics and randomization strata. Robust standard errors in parentheses in columns 1-4. Standard errors are clustered by party vote in 2013 in Column 5. Significant at ***1%, **5%, *10%. 44

45 Table 7: Mechanisms: Differential Effects of Divestment Timing and National Origin of Asset N = 1309 individuals Voted for Left Voted for Right (1) (2) (3) (4) (5) (6) (7) (8) Stock Treatment 0.054** (0.024) (0.024) Stock Treat. x Divest Pre-Elections 0.056* ** (0.031) (0.029) Stock Treat. x Divest Post-Elections 0.053** (0.026) (0.026) Palestinian Stock x Divest Pre-Elections 0.086** 0.093** ** ** (0.037) (0.042) (0.037) (0.040) Palestinian Stock x Divest Post-Elections (0.030) (0.036) (0.031) (0.034) Israeli Stock x Divest Pre-Elections (0.039) (0.057) (0.035) (0.047) Israeli Stock x Divest Post-Elections 0.090*** (0.030) (0.053) (0.030) (0.050) Cash Treatment 0.085*** ** (0.032) (0.032) Cash Treat. x Divest Pre-Elections 0.151*** 0.152*** *** *** (0.049) (0.050) (0.079) (0.042) (0.042) (0.067) Cash Treat. x Divest Post-Elections (0.037) (0.037) (0.074) (0.038) (0.038) (0.066) % Price Change of Exposed Asset by Election Day (0.009) (0.008) R-squared This table provides OLS estimates on the probability that an individual voted for the left in 2015 (Columns 1-4) and the right (Columns 5-8). All regressions control for the full set of demographics and randomization strata from Table 2, Column 4. Robust standard errors in parentheses. Significant at ***1%, **5%, *10%. 45

46 Table 8: Measuring the Direct Effect of Skin in the Game on 2015 Vote N= 1309 individuals Voted for Left Voted for Right OLS IV OLS IV OLS IV OLS IV (1) (2) (3) (4) (5) (6) (7) (8) Asset Treatment 0.064** 0.060** ** ** (0.025) (0.025) (0.025) (0.025) Stock value- actual on election day (100s NIS (0.008) (0.010) (0.007) (0.009) Palestinian Stock Treatment 0.063** 0.056* (0.032) (0.034) (0.032) (0.034) Palestinian x Stock Value on Election Day (0.013) (0.017) (0.012) (0.016) Israeli Stock Treatment * (0.033) (0.034) (0.031) (0.031) Israeli Stock x Stock Value on Election Day (0.012) (0.014) (0.010) (0.012) Cash Treatment 0.086** 0.085*** ** ** (0.039) (0.031) (0.039) (0.031) Cash x Stock Value on Election Day (0.056) (0.050) R This table tests whether the stock value on election day has a direct effect on the probability an individual voted for the left (Columns 1-4) and right (5-8) in the 2015 elections. Columns 1 and 5 shows the OLS estimate, including the actual asset value held by an individual on election day. As this is endogenous, Columns 2 and 4 provide IV estimates, instrumenting for the stock value on election day using the stock value of a purely passive investor who made no trades. The instrument is calculated based on the asset allocation, the redemption date (pre- or postelections), the initial value (high or low) and the price change of the specific asset by election day. Columns 3-4 and 7-8 add separate interaction terms with Palestinian Stock, Israeli Stock and Cash. Observe that since a passive investor provided cash would hold no stocks on election day, it is collinear with assignment to cash itself and we do not have a separate instrument for cash in Column 4 and 8. All regressions control for the full set of demographics and randomization strata from Table 2, Column 4. Robust standard errors in parentheses. Significant at ***1%, **5%, *10%. 46

47 Table 9: Mechanisms: Knowledge, Well-Being, Predicted Benefits of Peace Sample A. Media Consumption (OLS) [Jul 5] Mean SD Asset Treatment SE Asset Treatment Which of the following newspapers/websites do you usually read? Number of non-financial outlets [0-5] (1.032) (0.075) (0.097) Haaretz [0/1] (0.358) (0.023) (0.029) Israel Hayom [0/1] (0.495) (0.035) (0.045) Number of financial outlets [0-3] (1.120) 0.203*** (0.074) 0.195** (0.093) Observations 1, B. Effects of a Two-State Agreement (Ordered Probits) [Mar 19] Suppose Israel reaches a permanent agreement with the Palestinians on the principle of two states for two peoples. How do you think this will affect... [1 (worsen a lot), 2 (worsen), 3 (no change), 4 (improve), 5(improve a lot)] Israel's security? (1.392) (0.076) (0.097) Israel's economic situation? (1.329) 0.126* (0.073) 0.223** (0.094) your own personal security? (1.237) (0.075) (0.094) your own economic situation? (1.047) (0.077) (0.101) Observations 1281 / C. Political and Economic Knowledge (OLS) [May 5] Political Knowledge Score [Prop Correct out of 13] (0.212) (0.013) (0.018) Economic Knowledge Score [Prop Correct out of 5] (0.276) (0.016) (0.021) Stock mkt performance answer within 3pp of actual (0.489) 0.066** (0.033) 0.091** (0.042) Observations D. Subjective Well Being (Ordered Probits) [Mar 19] All Inexperienced (1) (2) 1, Overall, how satisfied are you with your life? [1-4] (0.661) (0.079) (0.101) On a scale from 0 to 10, how would you rate The overall well-being of you and your family (2.100) (0.072) (0.091) The happiness of your family (1.885) (0.072) (0.094) Your health (1.895) (0.070) (0.093) The extent to which you are a good, moral person and (1.379) (0.071) (0.092) living according to your personal values The quality of your family relationships (1.765) (0.070) (0.092) Your financial security (2.304) (0.071) (0.088) Your sense of security about life and the future in general (2.229) (0.069) (0.089) The extent to which you have many options and possibilities in your life and the freedom to choose among (2.238) (0.071) (0.090) Your sense that your life is meaningful and has value (2.053) (0.071) (0.090) Observations 1, The table reports the coefficient of Asset treatment from a separate regression on the different outcomes mentioned in the first column. All regressions control for the full set of demographics and randomization strata. Robust standard errors in parentheses. On July 15, we asked individuals which newspapers they usually read from among the following: Globes, Marker, Haaretz, Vesti, Yediot Ahronoth, Israel Hayom, Kalkalist and Maariv. Of these, Globes, Marker and Kalkalist are financial outlets. (Panel A). On March 19, we asked individuals to predict the effects of a two state solution at two levels- - personal and national-- and on two dimensions: security and the economy (Panel B). In a follow-up survey on May , we asked individuals 13 political knowledge questions, of which 2 were questions on salient events in the run-up to elections, 6 were questions on the positions taken prior to the elections by the two leading candidates for the right and left-- Netanyahu and Herzog, and 5 were on political facts. Economic knowledge questions asked individuals to provide estimates on the unemployment rate, inflation rate, whether the stock market rose and fell and its change in value, and the change in housing prices. All answers were scored correct if they were within 3pp of the correct answer. On March 19, we also asked individuals about their Subjective Well Being (Panel D). These included the top ten aspects that predict personal wellbeing from Benjamin et al. (2014, Table 2), excluding mental health. ***1%, **5%, *10%. SE 47

48 Table 10: Financial Asset Exposure Has Greater Effects on the Risk Averse OLS (1) (2) (3) (4) A. Main Outcomes Voted Left Voted Right Peace Deals Index Asset Treatment (0.038) (0.042) (0.320) Risk Averse ** (0.045) (0.047) (0.367) Asset Treatment * Risk Averse *** (0.050) (0.052) (0.406) Joint F (Asset Treatment Vars) Prob>p Observations 1,309 1,309 1,277 R-squared B. Support for Peace Deals Two State Soln 1967 Borders Split Jerusalem Refugees Asset Treatment (0.104) (0.104) (0.101) (0.120) Risk Averse * ** (0.117) (0.117) (0.116) (0.136) Asset Treatment * Risk Averse *** 0.300** 0.253** 0.341** (0.130) (0.127) (0.128) (0.150) Joint F (Asset Treatment Vars) Prob>p Observations 1,277 1,277 1,277 1,277 R-squared C. Effect of a Two State Solution on: Israel's Security Israel's Economy Own Security Own Economic Asset Treatment (0.143) (0.130) (0.125) (0.111) Risk Averse (0.160) (0.146) (0.142) (0.125) Asset Treatment * Risk Averse 0.311* 0.282* 0.260* (0.176) (0.164) (0.158) (0.140) Joint F (Asset Treatment Vars) Prob>p Observations 1,282 1,282 1,281 1,281 R-squared This table shows the differential effects of asset treatment on risk averse individuals, defined as those with ex ante subjective risk aversion at the median or below. All regressions are OLS, and control for the full set of demographics and randomization strata from Table 2 Column 4, except that we replace the willingness to take risk measure with a dummy for being risk averse. Panel A shows the effects on the main outcomes: the vote choice and the peace deal index (on scale of 4-16). Panel B examines each component separately (on a 1 (disagree) 2 (tend to disagree) 3(tend to agree) 4 (agree) scale). These include: Two states for two peoples, The 1967 Borders will be the borders between the two countries with a possibility of land swaps, Jerusalem will be divided into two separate cities: Arab and Jewish, and Palestinian refugees will receive adequate compensation and be allowed to return to the State of Palestine only. Panel C. shows the interaction effect on responses to the question: Suppose Israel reaches a permanent agreement with the Palestinians on the principle of two states for two peoples. How do you think this will affect... [1 (worsen a lot), 2 (worsen), 3 (no change), 4 (improve), 5(improve a lot)]. Robust standard errors in parentheses. 1%***,5%**,10%. 48

49 Table 11: Interactions with Wealth and a High Allocation of Assets OLS, N=1309 Individuals Voted for Left Voted for Right (1) (2) (3) (4) (5) (6) (7) (8) Asset Treatment 0.061* 0.066** (0.032) (0.026) (0.032) (0.026) Below Avg Income (0.044) (0.045) Asset Treatment * Below Avg Income (0.050) (0.047) High Allocation ** (0.023) (0.022) Stock Treatment 0.057** (0.027) (0.028) High x Stock Treatment (0.026) (0.024) Palestinian Stock Treatment (0.033) (0.034) High x Palestinian Stock Treatment ** (0.038) (0.035) Israeli Stock Treatment 0.087*** (0.033) (0.032) High x Israeli Stock Treatment (0.037) (0.033) Cash Treatment 0.101** 0.101** (0.042) (0.042) (0.039) (0.040) High x Cash Treatment (0.050) (0.050) (0.047) (0.047) R This table provides OLS estimates on the probability that an individual voted for the left in 2015 (Columns 1-4) and the right (Columns 5-8). Columns 1 and 5 include a dummy variable for whether an individual reported ex ante that they had monthly household income below the national average, as well as an interaction with the asset treatment. Columns 2 and 6 separate the effect of being allocated a high value of stocks (NIS 400 ~ US$100) relative to NIS100 (~US$50). Columns 3-4, 7-8 include interactions with different endowments of stocks vs cash (Columns 3,7) and the national origin of the stock (Column 4,8). All regressions control for the full set of demographics and randomization strata from Table 2, Column 4. Robust standard errors in parentheses. Significant at ***1%, **5%, *10%. 49

50 Figure 1: What did the researchers learn from this study? Elections Politics Econ. Knowledge Interest in Mkts Investor Choices Risk Attitudes Capital Market Trust in Mkt Increase Fin. Access Sell Stocks Which Stocks to Invest Nothing A Lot Don't Know Trust in Others Foreign Fin. Mkts These are the results of an open-response question at the end of the trading period (eg March 12 or April 2) to the question What did the researchers learn from the study?. Respondents only include the 840 participants who actually received treatment. Notice that, despite the study being conducted around the time of the elections, only eight mentioned politics or elections in their responses. The modal responses (other than don t know ) were that the researchers learned about the subjects economic knowledge, and attitudes towards risk and the capital market, which are indeed the subject of a companion paper. 50

51 Figure 2: Vote in Treatment and Control Groups in 2013 and Left Center/Other Right Control Asset Treatment 2013 Elections Left Center/Other Right Control Asset Treatment 2015 Elections Vote Share N=1309. 'Other' includes 59 individuals in 2013 and 17 in individuals did not vote in These graphs only include the 1309 subjects (out of 1345 assigned to treatment) for whom we have the 2015 vote. Note that 27 individuals in the sample did not vote in We over-sampled center voters (based upon their choice in 2013) at twice their vote share. Notice that the treatment and control group are well-balanced on vote choice in the 2013 elections. However, following the treatment, there is a shift to the left, and away from the right, in the asset treatment group relative to the control during the 2015 elections. 51

52 Figure 3: What was the major determinant of your asset s value? Inter-State Relations Most Important PLE BOP PALTEL TA25 Cash-TA25 LUMI BEZQ Elections Domestic Politics Most Important State of National Economy/ Policies Most Important Elections 15feb mar201515mar apr feb mar201515mar apr feb mar201515mar apr2015 These graphs show how the 840 participants who received an asset treatment answered a weekly question Various factors affect the success of a particular company. When you think about the performance of [assigned stock or index], which of the following is most important?. Panel A shows the proportion choosing regional political conditions: normal relations with the neighbors and the lack of conflicts and wars.. Panel B shows the proportion choosing Internal political conditions: the quality of the government and lack of corruption. Panel C shows the proportion choosing Conditions in the economy: price stability, the level of the national debt, the quality of regulation. Notice that holders of Palestinian stocks are relatively more likely to view both inter-state relations, and domestic politics as the most important determinants of the assets value, and holders of the indices tend to do so more than holders of individual company stock. Further, large proportions in both groups also tend to view the state of the national economy and the policies that affect it as the most important determinant of their assets value. Relatively few chose the other options, which included: the quality of management of the firm(s) and the quality of workers and employee conditions. 52

53 Figure 4: Asset Prices during the Experiment and 2015 Elections. Asset Price Ratio (Initial Price = 1) baseline survey financial trading treatment begins intermediate survey early divestment (1/3) Elections post survey final financial survey (2/3) LUMI TA25 BEZQ BOP PLE PALTEL knowledge survey Jan 29, 2015 Feb 12, 2015 Feb 26, 2015 Mar 12, 2015 Mar 26, 2015 Apr 09, 2015 Apr 23, 2015 May 07, 2015 Israeli stocks (Bezeq Telecoms (BEZQ), Bank Leumi (LUMI) and the Tel Aviv 25 (TA25)) are dotted in blue, Palestinian stocks (Palestine Telecoms (PALTEL), Bank of Palestine (BOP) and the Palestinian General Market Index (PLE)) are green. Asset prices fluctuated over the course of the experiment, with greater volatility for Israeli stocks. Israeli stocks ended up increasing, while Palestinian stocks remained relatively stable until the eve of the elections. The elections, that resulted in gains for the right-wing Likud party, led to sharp gains for Israeli stocks and losses for Palestinian stocks. 53

54 RISKING PEACE: SUPPLEMENTAL APPENDIX (NOT FOR PUBLICATION) Saumitra Jha and Moses Shayo November 17,

55 Table 1: Comparison of the Sample and the Israeli Population Sample (Initial) N = 1345 Sample (2015 Vote) N = 1309 Israeli Population 1. Region: Jewish Population in District (%) Jerusalem District Northern District Haifa District Central District Tel Aviv District Southern District West Bank Gender (Jewish Population above age 18 (%)) Male Female Age (Jewish Population above age 18 (%)) Male Female Religiosity (Jewish Population aged 20 and over (%)) Not religious/secular Traditional Religious Ultra-orthodox Education (Jewish Population level of schooling (%)) Less than high school grad (0 to 10 yrs.) High school graduate (11 to 12 yrs.) Post-secondary/BA Student (13 to 15 yrs.) College grad and above (16+ yrs.) Net Monthly Income per Household (NIS) Mean 10,978 11,044 14,161 Median 12,000 12,000 12, Statistical Abstract of Israel, Table 2.15, 2014 Totals 2. Statistical Abstract of Israel, Table 8.72, 2014 Totals 3. Statistical Abstract of Israel, Table 8.72, 2014 Totals 4. Statistical Abstract of Israel, Table 7.6, 2013 Totals. Survey data for (4) includes all observations age 20 or over (8 excluded from total sample) 5. Statistical Abstract of Israel, Table 8.72, 2014 Totals Income Survey, Table 12. Percentages correspond to entire population, not just Jewish. Survey data represents midpoint of SES categories. Table 2: Low (and Non-Differential) Attrition in the Voting Outcome Number of Individuals in: Control Asset Treatment Palestinian Stock Israeli Stock Cash Endowment Late Divestment Initial assignment Observed vote in 2015 elections Proportion observed

56 Table 3: Election Polls and Asset Price Performance Closing Asset Price Each Day (% of Feb 12 price) (1) (2) (3) (4) (5) % Seats Predicted for the Right (0.528) (0.407) (0.381) % Seats Predicted for the Left (0.240) (0.247) (0.175) % Seats Right x Israeli Stock ** ** ** (0.605) (0.607) (0.613) % Seats Right x Palestinian Stock (0.530) (0.526) (0.531) % Seats Left x Israeli Stock (0.472) (0.474) (0.478) % Seats Left x Palestinian Stock (0.242) (0.234) (0.235) % Seats Predicted for the Likud (0.143) (0.144) % Seats Predicted for the Zionist Union (0.186) (0.162) % Seats Likud x Israeli Stock * * (0.276) (0.280) % Seats Likud x Palestinian Stock ** ** (0.145) (0.136) % Seats Zionist Union x Israeli Stock (0.383) (0.388) % Seats Zionist Union x Palestinian Stock (0.191) (0.200) Asset Ticker Fixed Effects Yes Yes Yes Yes Yes Quadratic Time Trends No Yes Yes No Yes Week Fixed Effects No No Yes No Yes Observations R-squared This is an OLS regression of the closing price of each asset each day that there was a poll between January 30 to March 18. Each asset's value is divided by its value on February 12. The assets include all those participating in the study: Israeli Stocks include LUMI, TA25, BEZQ. Palestinian Stocks include: PLE, PALTEL and BOP. Reference Stocks include: AMGNRLX (the Amman Stock Exchange General Index) EGX30 (the Cairo 30 Index), XU030 (the Istanbul Index), and CYFT (the Cyprus/FTSE 20). The % of Seats for Left and Right are based on the simple averages of all polls linked in "Opinion Polling for the Israeli Legislative Election 2015" in Wikipedia and supplemented by an aggregation website maintained by Haaretz ( All regressions include asset fixed effects and errors are clustered at the asset level. We sequentially add Quadratic Time Trends and Fixed Effects for each week. Notice that the reference stocks are largely unaffected by the polls. However, Israeli stocks lose value with increases in predicted shares for the right. Looking at the two main parties which were the focus of the election (and for whom an increase in seat share would reduce reliance on coalition partners) in Columns 4 and 5 reveals that an increase in seat share for Likud was associated with a fall in the value of both Israeli and Palestinian stocks in our study. In contrast, an increase in the Zionist Union's share weakly increased the value of the Israeli stocks in the study. 3

57 Figure 1: Initial Allocation Screen. Here is a list of all the assets participating Both company stocks and index funds (explained). Note the asset you won and the # of shares you own. If the price of your asset increases, the value of your assets will increase accordingly. If the price goes down total value in NIS total value in JOD # shares current price in JOD 4