PLEASE SCROLL DOWN FOR ARTICLE

Transcription

1 This article was downloaded by:[princeton University] On: 19 May 2008 Access Details: [subscription number ] Publisher: Routledge Informa Ltd Registered in England and Wales Registered Number: Registered office: Mortimer House, Mortimer Street, London W1T 3JH, UK Journal of Ethnic and Migration Studies Publication details, including instructions for authors and subscription information: Herd and Network Effects in Migration Decision-Making Gil S. Epstein Online Publication Date: 01 May 2008 To cite this Article: Epstein, Gil S. (2008) 'Herd and Network Effects in Migration Decision-Making', Journal of Ethnic and Migration Studies, 34:4, To link to this article: DOI: / URL: PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: This article maybe used for research, teaching and private study purposes. Any substantial or systematic reproduction, re-distribution, re-selling, loan or sub-licensing, systematic supply or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

2 Journal of Ethnic and Migration Studies Vol. 34, No. 4, May 2008, pp Herd and Network Effects in Migration Decision-Making Gil S. Epstein This paper examines the impact of informational cascades (herd behaviour) and networks in influencing migrants location choices. It argues that herd behaviour and networks complement each other in explaining foreign location decisions. The paper develops both theories, emphasising the differences and the similarities between them. Empirical evidence shows that both herd behaviour and network externalities are important components in the decision-making of migrants. Keywords: Migration; Herd Behaviour; Networks Introduction Immigrants from one foreign source are often found concentrated together in host locations. While the clustering can reflect language proficiency due to former colonial relations, often the knowledge of the host language is not a characteristic of clustered immigrants. For example, we find concentrations of Turks in Germany, Tamils in Switzerland, Moroccans in the Netherlands and Belgium, Italians in Argentina, Greeks in Australia, and Ukrainians in Canada. In the United States, noticeable clusters of Mexican immigrants exist in California, Texas, Florida and Chicago. Fiftyeight per cent of migrants from Guanajuato, the Mexican state with the highest emigration rate to the US, go to California; 23 per cent to Texas. There are also more specific instances where emigrants from a town or region are concentrated in one foreign town or region. To name just one example of many, Macedonians from Skopje make up a sizeable part of the population of Gothenburg, Sweden. The prevailing explanation for such emigrant concentration is the existence of beneficial network externalities (see Stark 1991). Network externalities arise when previous emigrants provide shelter and work, or assistance in obtaining credit, or generally reduce the stress of relocation to a foreign culture (see Chiswick and Miller 1996; Church and King 1993; Gottlieb 1987; Grossman 1989; Marks 1989). Gil Epstein is Professor in the Department of Economics at Bar-Ilan University, Israel. Correspondence to: Prof. G. Epstein, Dept of Economics, Bar-Ilan University, Ramat Gan 52900, Israel. epsteig@mail.biu.ac.il ISSN X print/issn online/08/ # 2008 Taylor & Francis DOI: /

3 568 G.S. Epstein Economic models of migration have been criticised for reducing the explanation of migration decision-making to economic or labour market variables, omitting the role of social ties or interaction. This paper shows that economics can model such effects. In this context, it is able to account for the concentration of immigrant groups from the same place in particular destinations. Moreover, the models set out in this paper offer the possibility of distinguishing between two types of social interaction that are often blurred in non-economic literature: networks and herds. Network externalities imply I will go where my people are, since they will help me. An alternative explanation for clustering of immigrants from the one location is herd behaviour. Herd behaviour implies: I will go where I have observed others go, because all those who went before me cannot be wrong, even though I would have chosen to go elsewhere. Herd behaviour thus entails discounting private information to follow others (Epstein 2002). There is a substantial literature on herd effects. The theory of information cascades, or herd effects, has been applied to the explanation of behaviour in a number of contexts (see Banerjee 1992; Bikhchandani et al. 1992; Scharfstein and Stein 1990). There is also substantial literature on network externalities in migration. Beneficial network externalities arise when the stock of previous immigrants is sufficiently large to provide accommodation and work, other economic assistance and/or reduce the stress of adapting to a foreign culture (see also Carrington et al. 1996; Chiswick and Miller 1996; Munshi 2003; Zahniser 1999). Disadvantageous network externalities may arise if immigration is subject to adverse selection (high-productive immigrants do not want low-productive people to immigrate; see Stark 1991, 1995). Or, if the number of foreigners increases, there is more competition for jobs, which lowers immigrants wages. Negative network externalities limit the number of immigrants who can benefit from such externalities. Bauer and Gang (1999) have examined network effects in a model of return migration. Bauer et al. (2007) examine the relative importance and effects of these types of network externalities. The purpose of this paper is to describe how both herd behaviour and network externalities affect the decision-making of migrants. In the herd model, emigrants may have some private information, but are imperfectly informed about the attributes of alternative foreign locations, and pay attention to previous emigrants decisions. Behaviour is rational, on the supposition by new emigrants that previous emigrants enjoyed information that they did not have. The outcome is that individuals discount private information and migrate to the location that previous emigrants have been observed to choose. With herd behaviour, people can come to realise that following the herd has been a mistake, and they could change their minds about where to locate. The source of inefficiency of herd behaviour is discounting private information that was accurate. On the other hand, network externalities are beneficial to the new immigrant arriving in the host country and help provide different ethnic goods to the migrants already there. The networks help the migrant find a job, housing, consumption of ethnic goods and so on. In many cases, all this is done even in the language of the home country. In the case of network externalities,

4 Journal of Ethnic and Migration Studies 569 the size of the network determines its value in providing appropriate beneficial services to migrants on their arrival and while living in the new destination. Consider yourself living in a small town in a low-income country, and you decide to emigrate. Where would you go? You might prefer a certain country because of your familiarity with its language. That may leave you with a number of alternatives. Or you may choose a foreign location because of the presence of people there from your home community. You may have a relative or a friend of the family in the foreign location, or at least the name and address of somebody who knows your family and who will treat you sympathetically, assisting you with housing and finding a job, and perhaps in explaining the rules of the neighbourhood. Help in adjusting to a new environment, such as finding jobs and accommodation, is provided by network externalities. The ability to speak the language in a foreign country is in itself helpful. The network allows the immigrant to preserve his traditions and history in the new environment. Network externalities are not always positive. An increase in the number of foreigners in the host country inflates competition for jobs available to immigrants, thus decreasing the immigrants wages. Moreover, as the number of immigrants increases, the local population may become xenophobic. If full information were available about local conditions, migrants would choose the location where there are net benefits from network externalities. If such full information is not available, a choice is made under conditions of uncertainty. If you have imperfect information, which decision-making rule should you adopt? In the face of uncertainty, a common resolution is to diversify, but here you confront an indivisible location judgment. You may not know all that much about life in a particular location. You observe, however, that other people, who are like you, have recently been favouring this location. That is, you may take the position that so many other people cannot be wrong. If you behave in this way and discount your own information or feelings in order to follow the decisions of others, you are adopting a decisive rule that gives rise to herd behaviour. In order for a population of immigrants to produce network externalities that will attract other migrants, they must be sufficiently large in number. In many situations, it is not clear how this critical mass of people arrived at a certain location. Informational cascades*herd behaviour*help us understand the creation of the critical mass that creates network externalities. In order to create a herd in a certain location, the number of immigrants needed is relatively small. Thus, herd behaviour may be an explanation for the creation of the mass of immigrants that is sufficient to attract others to join and enjoy the positive externalities of the network. The immigrants will decide to follow the flow (informational cascade or herd behaviour) rather than the stock (network externalities) of previous immigrants destinations. This paper sets out a formal framework introducing informational cascades and compares them to a model of network externalities. I begin by presenting a theoretical section that distinguishes between network externalities and herd behaviours and then consider empirical evidence for these phenomena.

5 570 G.S. Epstein Theory: Herd Behaviour and Network Externalities Network Externalities Let us first consider network behaviour. Consider an individual who obtains benefits from migrating to a certain country. We denote the benefits individual j obtains by the function U j (.). In our simplified environment, the benefits the individual receives, U j (.), depend on two main components that are present in the host country: the first is the wage, w j, that the migrant will receive by migrating to the new location; and the second is the stock of immigrants from the same origin who have previously migrated to the new location, N. From the above discussion, the migrant s advantage from migrating to a certain location (his or her utility) increases the more s/he earns (increases with wages), and the benefits increase as the size of the network externalities increase. In other words, as the stock of migrants who have already immigrated to the same host location is higher, the migrant will obtain higher benefits when s/he arrives at the new location. We could write this in the following mathematical j (w j ; N) 0 j(w j ; N) 0 For a given utility, there is a trade-off between changes in wages and changes in the size of the network. In other words, we could think of the way the benefits/utility of the migrant change as a result of a decrease in the wages of the migrant. If wages drop, how could we compensate the migrant in terms of the size of the network in order to make sure s/he obtains the same total benefits or utility s/he obtained before the change in wages? 1 In order to get the same benefits from migrating to a host country, there is a trade-off between the wage level in the host country and the stock of immigrants who migrated before immigrant j from the same country: namely, in order to keep the same utility level one could increase one and decrease the other, such that the same utility would be obtained. Putting it differently we may find different countries with alternative compositions of wages and sizes of network that will give a migrants the same benefit level. Mathematically it may be presented in the following j (w j ; N) dw j B0 j (w j ; j Moreover, as the migrant s wage and/or the stock of previous migrants from the same origin rises, the benefits/utility the new migrant obtains will increase. Let us denote the wages of the migrants by w f. It is not clear if the migrants and the local population s wages are the same. It may well be that the migrants receive a wage lower than that of the local population. This, of course, depends on the skill levels of

6 Journal of Ethnic and Migration Studies 571 migrants and their willingness to work at lower wages than those of the local population, even doing the same job. Moreover, the migrants may complement the local labour market or may act as substitutes. This will affect the equilibrium wages determined. Assume a normal downward sloping demand function for migrant labour where, as wages increase, the demand for workers decreases. Denote this by q d (w f,n L )so qd (w f ; N L ) B0: In addition to depending on the migrant wage, f expression implies that demand also depends on the relationship between the migrants and the non-migrant population, N L. Assume that the supply of migrants willing to work is constant at N in the short run and the migrants are willing to work for any wage with the supply of non-migrants fixed at a level of N L. Equilibrium occurs when demand for migrant workers, q d (w f,n L ) equals supply N: q d (w f,n L ) N. Let us note that all migrants will be employed and it could be that the migrants and the local population are often in different labour markets. The wages that satisfy the equilibrium constraint q d (w f,n L )N are denoted by w f (N L; N): In other words, the wage in equilibrium is a function of the stock of immigrants in the country. More specifically we can easily show f (N L; N) B 0; namely, as the stock immigrants increases, the equilibrium wage decreases. Of course, the immigrants wages are also a function of the local population size, which we assume does not change. Let us now consider the total effect of an increase in the stock of immigrants on the migrant s benefits/utility level. Remember that the migrant has two components that affect his or her benefits/utility after migration: the wages s/he will earn and the size of the network. Increasing the size of the network, i.e. the size of the population of migrants in the host country, will affect the migrant in two ways: 1. The increase in the number of migrants will decrease wages ð dw f B0Þ: The dn decrease in wages will decrease the benefits the migrant (w f ; N) B0 f 2. Increasing the network will increase the benefits the migrant can (w f ; N) 0 We have therefore obtained two opposing effects resulting from the increase in the size of the network: a direct effect, which increases the migrant s benefits; and a negative effect via the decrease in wages, which decreases the migrant s benefits. Finally, we must look at the full derivative of the change in the size of the stock of immigrants on the migrant s utility, as follows:

7 572 G.S. Epstein du j (w f ; j(w f ; j(w f ; f The stock of migrants (the network effect) affects utility in two ways: directly via the positive externalities that the network provides to the migrants, and indirectly via negative externalities on the wages, which decrease. The first component on the righthand side of (4) is positive, while the second component is negative. The old migrants (the stock of immigrants) who are already in the host country prefer that the maximum number of migrants coming to the country will be such that their utility is maximised. This occurs when (4) equals zero. 2 That is, the increase in the benefit from one extra migrant which increases the positive network effect will equal the decrease in the benefit on the side of the wages. This occurs at a point N * as presented by Figure 1. Put another way, the marginal increase of migrants utility from externalities equals the marginal effect of the decrease in wages as a result of the additional j (w f ; j(w f f Thus if the stock of immigrants exceeds N +, increasing the stock of immigrants increases the externalities, although wages also decrease. The left-hand side of the equation tells us the benefit accruing from an additional migrant; that is, how dw f dn (4) (5) Utility / 3 Probability N* Number of immigrants Figure 1. Impact of network effects

8 Journal of Ethnic and Migration Studies 573 growth in migrant stock resulting from one more person increases the desirability of a location. The right-hand side indicates the costs of an additional migrant, i.e., how the impact of an individual on wages tends to make the location less desirable. If the benefits of an additional person exceed the costs, it is worthwhile adding an additional person, so existing migrants will wish to have more migrants joining them. However, if the costs exceed the benefits, then the existing migrants do not wish any more migrants to join them. A new migrant who is considering migrating to a host country may take into account the stock of immigrants already in the host country, N, and s/he may also take into account the expected number of migrants who will decide to go to this host country after s/he arrives there. This means that the migrants may also take into account what will happen in the future and thus consider the expected number of migrants, not only the actual number at the present time. Therefore, the migrant may be looking at an expected value E(N) and not N. The migrant would then compare the expected utility of going to this country with the expected utility obtained from migrating to a different county. As argued above, given that the immigrant is already in the host country, s/he prefers that the stock of immigrants be equal to N*. However, when this individual makes the decision whether or not to migrate to this county, s/he compares the expected utility from different countries and chooses the one with the highest value. We may therefore see migrants deciding to migrate to a country in which the stock of migrants has already exceeded N *. Thus, the probability that an individual chooses to migrate to a country where the stock of immigrants already exceeds N * is positive. This probability, however, decreases as the stock of immigrants already in the host country increases. We conclude (see Figure 1) as follows: Given network externalities, the probability of an individual migrating to a certain country has an inverse U-shape relationship, with regard to the stock of immigrants already in the host country. Herd Behaviour Let us now consider herd behaviour. Following Epstein (2002), the model specifies that migration decisions are made sequentially, with people contemplating emigration at a given stage in their lives. When making a decision such as migration, individuals must consider all the different alternatives. Since potential migrants do not know for certain how living in any of the new host countries will turn out, they try to obtain as much information as possible in order to help them make the best possible decision. In the absence of any real information, the individual believes it is likely that each of the locations is equally good. However, when an individual receives information, s/he will calculate the probability that a given location is the best one for him or her. Not every individual receives the same material or the same type of report. Information comes in many different types and forms. One may receive facts by searching the internet, reading newspapers, watching the TV, and so on. Not everyone receives information, though it is probable that most do. In the case where

9 574 G.S. Epstein individuals receive a number of different informative items, we calculate the total output. Does it tell us where to migrate to or not? We call this dimension a signal. Individuals respond (or do not) to signals or information packages about hostcountry possibilities. As stated above, not all individuals receive these signals (for example, the individual did not watch TV on the day a programme on the country s/he was considering was broadcast). Thus an individual receives a signal with probability p. The information received may not be correct. Moreover, one place may be good for one type of person while it may not be suitable for another, so even information that is provided in good faith may not be correct for the person receiving it. We assume that with probability q the signal is true for the prospective migrant. In the case where there are k options the individual can follow, without information s/he would choose randomly between them, and thus where the individual does not have any information, the probability of choosing randomly between the different options equals to 1 : It is assumed of course that the k probability of the signal identifying the correct location is greater than the probability for one chosen randomly, otherwise it would be better to choose randomly than to use the information obtained by the signal: q 1 k : The individual does not only confine him/herself to receiving signals, but s/he also observes the behaviour of previous migrants. Potential migrants cannot, however, observe the information signal that was the basis for the previous migrants decisions. Given the data available, each individual chooses a country to which to migrate. The structure of the game and Bayesian rationality are common knowledge. This means that, given the information each individual has, the signal s/he received or did not receive, and what s/he saw previous migrants do, s/he updates the probability about which location is appropriate for him/her. At this point we assume that the individuals are rational, and that they only follow the herd if it is strictly preferred to any other alternative. 3 The decisions migrants take under different situations where they have to choose between two different locations are presented in Figure 2. Let us consider the case of the first individual (individual 1) considering migration, and let us assume that neither of the two previous individuals chose to emigrate. In this case s/he knows that neither received a signal. Individual 1 will copy them if, and only if, s/he does not receive a signal; otherwise s/he will follow the signal s/he receives, since s/he rates more highly the chances of obtaining a greater benefit when s/he migrates following the signal. If one of the previous two individuals chose not to migrate and the other chose to migrate, we can infer that the first individual did not receive a signal and the second individual did. If individual 3 then receives a signal that indicates migration to the country to which the second individual has migrated, individual 3 will join the

10 Journal of Ethnic and Migration Studies 575 Ind. 1 Ind. 2 S1 S2 M1 S1 S2 M1 M2 S1 SN S2 M2 M1 SN S1 SN S1 SN MN Ind. 3 M1 M1 M1 M1 M1 M1 S Signal S1 signal to migrate to country 1 S2 signal to migrate to country 2 SN did not receive any signal M migration decision M1 migrate to country 1 M2 migrate to country 2 MN not to migrate at all. Figure 2. Migrant decision-making with herd effects second migrant. Otherwise, if a signal different from that of individual 2 is received, individual 3 will follow his own signal. If individuals 1 and 2 have chosen to migrate to the same country, and individual 3 receives a signal to migrate to a different country, s/he will still migrate to the same country as individuals 1 and 2. In order to explain this let us look at Figure 2.

11 576 G.S. Epstein The first individual may receive a signal, S, to go to country 1, S1, country 2 S2, or not receive a signal at all, SN. If he did not receive a signal he will not migrate. If he received a signal to migrate to country j (j1,2) he will follow his signal and migrate to country Mj since this is better than staying in his home country. Now let us consider the case where the two first migrants choose to migrate to country number 1, M1. Since individual 3 did not see the signals the first two received and can only see where they migrated to, s/he can conclude that the first migrant received a signal (see discussion above) and the second either received a signal also for country 1, S1, or did not receive a signal at all, SN. Moreover it is clear that migrant number 2 did not receive a signal to migrate to country 2. Therefore individual 3 does not know on which branch s/he is standing, on the right or left, in Figure 2. If the third individual receives a signal to migrate to country 1, S1, s/he will follow this signal as it does not contradict where the other migrants have already gone. If the migrant does not receive a signal, SN, and s/he sees that both individuals migrated already to country 1, showing that at least one had a signal to go to that country and not to country number 2, s/he will then choose to migrate to country 1, M1. Therefore, s/he considers the signals received by others as if s/he received them him/herself. On the other hand, if s/he received a signal to migrate to country 2, s/he definitely knows that at least two of three individuals (including her/him) received signals, two of which are opposing with equal probability of being true. It is not clear if the second migrant received a signal or not. If s/he did, it was to country 1 and definitely not to country 2. Therefore, on average, the number of signals indicating to migrate to country 1 is higher than that for country 2 (1p signals to country 1 while only 1 to country 2). Individual 3 will choose to migrate against the signal received and follow the herd. Based on the fact that there is a positive probability of receiving a signal (or not receiving a signal), the individuals use this information to understand why others have decided to migrate to wherever they went, and conclude from this, at one of the levels, what type of signal each individual received. 4 This is the basis of herd behaviour. Individuals will migrate following the herd (flow) while disregarding their own private information. It is clear that every subsequent individual would follow the herd. The moment the first two have decided to go to the same place, all the succeeding ones, without any connection to what signal they receive or do not receive, will decide to follow the herd and migrate to the same place. Moreover, it can be shown that if v people have migrated to country a, and v-2 to country b, there will be herd behaviour whereby all later individuals will migrate to country a, disregarding any information they obtain themselves (see Epstein 2002 for a full proof). Herd and Network Effects Together Consider the case where one individual has chosen to migrate to a country. A second individual receives a positive signal indicating emigration to a different country. If

12 Journal of Ethnic and Migration Studies 577 this later individual chooses to follow the first migrant, then s/he knows that all successors will follow for informational and payoff reasons (herd behaviour and positive externalities). If s/he chooses the other country, there is a positive probability that s/he will end up alone. So, while s/he may think that the basic payoff or utility from moving to the alternative country is as good as for the first country, the awareness of the positive network payoff will induce him/her to choose the location chosen by this first emigrant. Herd behaviour is, therefore, more pronounced when migrants believe that the early migrants will be followed by others. In the presence of beneficial networks, where the networks help the migrants, the utility from migrating to a country depends on: (1) the stock of immigrants who have previously immigrated and (2) how many people will migrate in the future. So even if the wage in a country is relatively low, the positive effects of the network may make that country an attractive location. For example, suppose n people have migrated to country j and one individual has migrated to country k, and that the utility of a migrant in country j is higher than that of a migrant in country k. In this case it could be that, if n migrants had instead migrated to country k and one migrant to country j, utility in country k would have been higher than in country j. One could think of this as a self-fulfilling prophecy where each thinks this is best and by doing so makes it best, even though, under different circumstances, it really would not have been the best location. With herd behaviour, the probability that a signal received by an individual to migrate to a particular country is true is a function of both the number of previous migrants who have migrated to the same country and the stock of immigrants who have chosen other countries. We can define the probability in the following way. Suppose an individual has received a signal indicating that country j is best, and has to choose between countries j and k. The benefits from the network influence the probability that a signal is true via the relative number of migrants who previously immigrated to the different countries. Where there are positive network effects, herd effects are more pronounced. On the other hand, a migrant may move to a country and find out that the marginal positive effect of the network, namely the change in the benefits from migrating to that country, is lower than the marginal negative effect of the wage. In other words, the stock of immigrants who have migrated to this host country has exceeded N *. In Figure 1 we see that, as a result of networks, and where there is a stock of fewer than N * migrants, the desirability of migrating to that country increases with the increase in the stock of migrants; but when the stock of migrants exceeds N * the desirability decreases. A migrant who is living in a host country where the stock of immigrants is higher than N * will now send negative signals to the home country advising against migrating to that country. The local population in the home country receives these negative signals. However, that population knows that a large number of individuals have migrated to that particular country, and may even receive other information that this is the place for them. An individual who has to make a decision will weigh the information received: the

13 578 G.S. Epstein stock of previous individuals who migrated to that country (and to other countries), the general information received while observing the flow of migrants, and the negative information received from the migrants who have already migrated to that country. This individual also knows that there is a likelihood that the information s/he received from the migrants in the host country may reflect their interests in discouraging other migrants from joining them (as this will decrease their utility when the stock of migrants in the host country increases to N at point 3 on Figure 1). However, it may be optimal for the migrant to join them even if there are negative signals. Epstein (2002) shows that, under such conditions, in order for the individual to follow the flow (herd), the proportion of negative signals relative to the stock of migrants must fall. In other words, even though the migrants who are already in the host country are sending signals not to come, the potential migrant looks both at how many migrants have recently joined that specific host country and at how many negative signals are being received. It is clear to the migrant that there is a probability that the signals are incorrect. Thus, if the number of migrants who joined recently is sufficiently large, relative to the number of signals received, the potential migrant will decide to migrate despite the negative signals. In our case, if (at point 3 in Figure 1) the number of migrants is sufficiently high, the immigrants will continue migrating to that host country. There are two circumstances in which network and herd effects have qualitatively different impacts. Where migration to a particular location is just beginning, the network may not be sufficiently large to attract migrants if there was full information, but the herd effect may cause large flows even before networks are present. It will also be the case that, where the size of the population of migrants grows to be very large, migrants with full information would not choose the location, whereas herd effects may continue to draw migrants. Empirical Evidence Several empirical studies investigate the determinants of location choice of immigrants in the United States. Bartel (1989) finds that post-1964 US immigrants tend to locate in cities with a high concentration of immigrants of similar ethnicity. She further shows that more highly skilled migrants are less geographically concentrated and rely less on the location of fellow-country nationals. Dunlevy (1991), focusing on Caribbean and Latino immigrants to the US, and Jaeger (2000), who differentiates between immigrants of different status, find that immigrants tend to locate where former immigrants of the same ethnicity are concentrated. Jaeger (2000) further shows that migrants in the employment admission category are more likely to locate in areas with low unemployment. Bauer et al. (2007) distinguish two types of network and herd effects, using data from the Mexican Migration Project. 5 One of their network variables captures the general characteristics that describe the type of origin-specific consumption products that migrants wish to consume in a US location. This variable is generally called

14 Journal of Ethnic and Migration Studies 579 ethnic goods, which describes the availability of products that are unique to specific groups from specific origins. Their other network variables capture origin, village connections, and the history of a village in US locations. Using these two variables helps to distinguish a generalised network effect from the village-specific links. The herd variable describes the flow of migrants during the year. Their empirical results show that both network externalities and herds have significant effects on the migrant s decision about where to migrate. Moreover, the significance and size of the effects vary according to the legal status of the migrant and whether the migrant is a new or a repeat migrant. The treatment above (Figure 1) shows that the ethnic goods network effect has an inverse U shape, not a simple positive effect, as often presented in the literature. In other words, as the network increases, the benefits to the migrant increase. However, if the network becomes too large, the benefits to the migrants decrease. In contrast, the flow (herd effect) significantly increases the size of the flow of migrants. How does one calculate networks? Bauer et al. (2000), based on the same data, use three different measures of ethnic networks in order to investigate different channels through which ethnic networks affect the location choice of migrants. The empirical evidence shows that the availability of ethnic goods, the information provided by return migrants to potential migrants in their origin village, and the number of current migrants from an origin village living in the host location compared to other locations are significant and important for the location choice of migrants. Their empirical results further show that the effect of the size of a Mexican migration network in a US location on the probability that individuals will immigrate to this location follows an inverted U. That is, increasing the network size will increase the benefits to the migrants up to a certain point, but, after that point, increasing the networks will decrease the benefits to the migrants. This behaviour is robust with respect to the type of network effect examined. However, when all three network variables are included simultaneously, the effect of the share of the cumulative migration experience of a village is U-shaped (increasing the size decreases the benefits at the beginning and only afterwards increases the benefits to the migrants), while the other network variables maintain their inverted-u shape. These patterns are also found in examining legal and illegal migrants, with networks being more important for illegal migrants. Networks are also more important for first-time migrants compared to repeat migrants. Epstein and Cohen (2006) investigated the case where migrants from the Soviet Union decided to live in Israel. They show that both network and herd behaviour determined the location choice of the migrants: the network effect has an inverted U shape and the herd effect is linear. It was also discovered that there is an inverted U shape effect of the size of the network on language proficiency and the probability of finding a job. Moreover, it was shown that those who used a higher component of herd behaviour in their decision obtained higher earnings than those using the higher network component.

15 580 G.S. Epstein In general, we can conclude from the above that, where networks are of relatively small or moderate size, individuals may frequently choose a location with lower income but which provides the benefits of a network. However, as networks increase in size, the determining factor for migrant decisions will more often be wages, since when many large networks are already in place, difference in the benefits from network externalities will be overwhelmed by differences in the wages. Therefore, although migrants will be willing to give up wages to get a larger network initially, as networks become sufficiently large the migrant will prefer locations with higher wages. Concluding Remarks The purpose of this paper has been to draw attention to herd behaviour and network externalities in influencing migrant location. Herd behaviour complements network externalities in explaining foreign location decisions. Network externalities may not be sufficient to explain the choice of foreign locations, since a number of alternative locations may all offer network externalities. Herd effects can explain which of the alternatives offering network externalities is chosen. There are a number of additional dimensions to the differences between network externalities and herd effects in explaining the choice of emigrants destinations. Positive network externalities tell a story of efficiency through the internalised benefits provided by the externalities. No mistakes are possible. In contrast, herd behaviour introduces the possibility of economic inefficiency through the discounting of accurate private information. Also, a prior critical presence of emigrants with the same cultural background or from the same location is required for network externalities. This is not a prerequisite for a herd effect to explain choice of foreign location. When the population of prior immigrants in a foreign location is small, network externalities are, of course, not present. Nonetheless, emigration decisions are made, generally under conditions of uncertainty. In such cases, we can only look to herd effects to explain initial immigrant clustering. Once the immigrant population reaches a particular size, relations can become more impersonal, and the arrival of someone from back home may not evoke the same feeling of responsibility and benevolence. Network externalities can, therefore, be subject to diseconomies of size of the immigrant population. After a sufficiently long presence, a local individualistic culture can take hold* Let the new arrival work hard and succeed on his own merits, like I did. Thus, after a certain number of immigrants have arrived, it may be beneficial for the emigrant to join a different network. Herd behaviour may lead immigrants to continue coming to the same location when network externalities no longer justify this decision. Often host countries want migrants (Epstein and Hillman 2003). Trying to influence migration, even a limited migration, may result in a herd of migrants who are not wanted by the host country or by the local population. This scenario

16 Journal of Ethnic and Migration Studies 581 illustrates the complexity of national preferences and xenophobia regarding immigrant composition. The host country may have a clear objective of what it wants, but as a result of informational cascades, a herd of migrants may arise and bring unexpected results. In the third section of the paper I discussed empirical evidence for herd and network externality behaviour of immigrants. Empirically we can distinguish between network externalities and herd effects in stock and flow terms. If the flow of emigrants to different locations is related to the prior flow or stock of emigrants, we can infer network externalities are important. If the flow of emigrants is related to prior flows, herd effects are important. The evidence, as presented above, shows that both network externalities and herd behaviour influence the decisions of migrants both for legal and illegal migration. Therefore, in analysing the decision-making of migrants and their impact on the local economy, one should consider both the network and the herd behaviour of the migrants. Notes [1] Formally, for a given utility level, an iso-utility locus (indifference curve) is described by the following: du j (w j ; j (w j ; N) dw j (w j ; N) dn 0 [2] The second order condition must satisfy: d2 U j (w f ; N) B0: dn 2 [3] The rationality of the individuals could be presented by three main assumptions that govern the individuals actions (these assumptions minimise the likelihood of herd behaviour. Firstly, an individual who does not receive a signal and observes that everybody else has chosen to stay home will also choose not to migrate. Secondly, an individual who is indifferent about following her/his own signal or copying someone else s choice will follow her/his own signal. Thirdly, an individual who is indifferent about following more than one of the previous migrants decisions will choose to randomise her/his decision with equal probabilities assigned to the different alternatives. [4] This can be shown formally in the following way. Assume that individuals 1 and 2 emigrated to country j and individual 3 has a signal to migrate to country k. Denote by Pr(j½j,j,k) the probability that country j is the correct country to migrate to, given that the first individual has migrated to country j, the second person has migrated to country j and the third person receives a signal to migrate to country k. This probability will be calculated by the third individual in order to determine if he should migrate to country j. In a similar way, Pr(k½j,j,k) is the probability given the same information that country k is the correct country to migrate to. Using Bayes rule, individual 3 can calculate the probability that the true signal is j out of Pr(j; j; kj j) Pr(j) m possible countries. Bayes rule implies that Pr(jj j; j; k) where Pr(j; j; k) Pr(j,j,k) is the probability that individuals 1 and 2 migrate to country j while individual 3 obtains a signal to migrate to country k and since without any other information all countries have the same probability of being the correct place to migrate to: Pr(j) 1 m thus

17 582 G.S. Epstein Pr(jjj; j; k) p 3 q 2 (1 q) 1=m p 2 (1 p)q(1 q) 1=m Pr(j; j; k) In the same way, individual 3 can calculate the probability that the true signal is k: Pr(kjj; j; k) p 3 q(1 q) 2 1=m p 2 (1 p)q(1 q) 1=m Pr(j; j; k) As explained above q 1. Since in this case k2 and we are comparing two possibilities so k q0.5, Pr(jj j; j; k)pr(kj j; j; k) (8) Thus, individual 3 will migrate to country j even though a signal was received to migrate to country k. [5] Massey and Zenteno (1999) show that the Mexican Migration Project is a good source of reasonably representative retrospective data on documented and undocumented Mexican migration to the United States. The data comprise more than 7,000 households in 52 communities. Each year since 1987, two to five communities are surveyed. Each community is surveyed only once. In general, 200 households in each community are selected through random sampling. The data include information on the socio-economic characteristics of the household head, such as age, education and marital status, their migration histories, including information on the year of migration, costs of border crossing, documentation and location in the United States. (6) (7) References Banerjee, A.V. (1992) A simple model of herd behaviour, Quarterly Journal of Economics, 107(3): Bartel, A.P. (1989) Where do the new US immigrants live?, Journal of Labour Economics, 7(4): Bauer, T. and Gang I.N. (1999) Temporary migrants from Egypt: how long do they stay abroad?. Bonn: IZA, Discussion Paper 3. Bauer, T. Epstein, G.S. and Gang, I.N. (2000) What are migration networks? Bonn: IZA, Discussion Paper 200. Bauer, T., Epstein, G.S and Gang, I.N (2007) The influence of stocks and flows on migrants location choices, Research in Labor Economics, 26: Bikhchandani, S., Hirshleifer, D. and Welch, I. (1992) A theory of fads, fashion, custom, and culture change as informational cascade, Journal of Political Economy, 100(5): Carrington, W.J., Detragiache, E. and Vishwanath, T. (1996) Migration with endogenous moving costs, American Economic Review, 86(4): Chiswick, B.R. and Miller, P.M (1996) Ethnic networks and language proficiency among immigrants, Journal of Population Economics, 9(1): Church, J. and King, I. (1993) Bilingualism and network externalities, Canadian Journal of Economics, 26(2): Dunlevy, J.A. On the settlement patterns of recent Caribbean and Latin immigrants to the United States, Growth Change, 22(1): Epstein, G.S. (2002) Informational cascades and decision to migrate. Bonn: IZA, Working Paper 445 (original version 1998). Epstein, G.S. and Cohen, O. (2006) Immigrants during the 1990s from the former Soviet Union: herd effect and network externalities, The Economic Quarterly, 53(1): (in Hebrew).

18 Journal of Ethnic and Migration Studies 583 Epstein, G.S. and Hillman, A.L. (2003) Unemployed immigrants and voter sentiment in the welfare state, Journal of Public Economics, 87(78): Gottlieb, P. (1987) Making Their Own Way: Shorthorn Blacks Migration to Pittsburgh, Urbana: University of Illinois Press. Grossman, J.R. (1989) Land and Hope: Chicago, Black Southerners, and the Great Migration. Chicago: University of Chicago Press. Jaeger, D.A. (2000) Local labour markets, admission categories, and immigrant location choice, online at Marks, C. (1989) Farewell*We re Good and Gone: The Great Black Migration. Bloomington: Indiana University Press. Massey, D.S. and Zenteno, R. (1999) A validation of the ethnosurvey: the case of MexicoUS migration, International Migration Review, 34(3): Munshi, K. (2003) Networks in the modern economy: Mexican migrants in the US labor market, Quarterly Journal of Economics, 118(2): Scharfstein, D.S. and Stein, J.C. (1990) Herd behaviour and investment, American Economic Review, 80(3): Stark, O. (1991) The Migration of Labour. Oxford: Basil Blackwell. Stark, O. (1995) Altruism and Beyond: An Economic Analysis of Transfers and Exchange Within Families and Groups. Cambridge: Cambridge University Press. Zahniser, S.S. (1999) Mexican Migration to the United States: The Role of Migration Networks and Human Capital Accumulation. New York and London: Garland.