1 / 26

Interest and identity in network formation: Who do smallholders seek out for information in rural Ghana

Social Networks in Economics. There is a growing awareness within economics that social networks affect individual decisions and outcomes in a variety of domains: the organization of tradetechnology adoption and learningaccess to credit and to informal insurancefertility and health decisionsN

natane
Download Presentation

Interest and identity in network formation: Who do smallholders seek out for information in rural Ghana

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


    1. Interest and identity in network formation: Who do smallholders seek out for information in rural Ghana ? Paulo Santos and Christopher B. Barrett Cornell University Pew/Rockefeller Workshop on “Social Dynamics and the Microeconomics of Poverty” Bellagio, Italy April 1, 2005

    2. Social Networks in Economics There is a growing awareness within economics that social networks affect individual decisions and outcomes in a variety of domains: the organization of trade technology adoption and learning access to credit and to informal insurance fertility and health decisions Networks are usually defined in terms of exogenous variables such as ethnicity, gender or quasi-fixed variables such as location or religion. That may be true for the existence of networks. But given networks’ value, shouldn’t we expect agents to choose with whom they want to activate a tie? Social networks’ use likely endogenous.

    3. Our contribution to small but rapidly growing literature: methodologically, we propose a new way of measuring the potentially asymmetric effect of social distance … ordinal relations matter to cost/benefit of making a link. substantively, we find that controlling for relational attributes, individual identity plays little role in the process of network formation/activation. Network interlinkages appear modest. (albeit ID problem) The “strength of weak ties” hypothesis finds support in the data we use. Social Networks in Economics

    4. Describe Udry/Goldstein Ghana data Identifying social networks and their activation Explaining social contacts/acquaintance Measuring social distance Econometric strategy Estimation results: Who knows whom? Asking for information Network interlinkage? Strength of weak ties Conclusions Plan of Presentation

    5. Udry/Goldstein Ghana Data Rural communities in Southeastern Ghana: 4 clusters of villages 60 couples/triples per village interviewed 15 times each over 2 years Extensive data on identities (clan, gender, age, etc.), economic activities (farm and non-farm) and social interactions Available at http://www.econ.yale.edu/~cru2/ghanadata.html Used in series of papers by Goldstein and Udry, including with co-authors (e.g., Conley)

    6. Respondents were matched with 7 individuals randomly chosen from the sample and asked: “Do you know __?” “Could you go to ___ … … if you had a problem with unhealthy crops? … for advice on when to apply a new kind of fertilizer? …if you wanted to discuss changing your methods of planting? … if you wanted to find a buyer for any of your crops?” Questions reflect respondents’ potential networks, not actual networks. Identifying social networks

    8. Estimate probit model to explain this pattern of answers. Sequential structure (not knowing match implies not asking for advice): estimation of the models concerning information networks using the sub-sample of those who know the match. Explanatory variables determine costs and benefits of establishing a link: Gender, age, clan, land, non-land wealth, education Experience with different crops, importance of non-agricultural activities, occupation Note: physical distance problematic for these data and question

    9. Existing analyses of network formation typically express distance between two agents as the Euclidean norm of the difference of the agents’ characteristics: Dij = | Xi – Xj| Implicitly, this imposes symmetry on the effect of social distance upon the net returns to making contact. Yet ordinal relations (e.g., wealth) likely matter and socio-cultural convention may yield asymmetric cross-categorical matches (e.g., gender).

    10. An alternative approach: For non-categorical variables: I(Xi – Xj< 0) * |Xi – Xj| + I(Xi – Xj = 0) * |Xi – Xj| where I(?) is an indicator function. For categorical variables, use dummy variables describing match characteristics: e.g., mm = {1 if i=male and j=male, 0 otherwise) mf = {1 if i=male and j=female, 0 otherwise) etc.

    11. Estimate probit lijk* = X’? + uij for respondent i’s propensity to link with j on issue k assuming E(uij) = 0 , Var (uij) = 1 E(uij, uih) ? 0 if j ? h , E (uih, ujh) = 0 if i ? j E(X, uij)=0 Have to drop observations due to non-observation. But Kolmogorov-Smirnov tests do not reject null that distributions unchanged by sample reduction. Sequential choice process – estimate “know” then “link” conditional on knowing

    12. Results: - Individuals with more wealth and net business revenue more likely to know others, but it’s those who are wealthier still and less likely those who are poorer. - Common clan increases familiarity - Teachers more likely to be known - Highly asymmetric effects of wealth, net business revenue,gender (women less likely to know men), migrant status (migrants less likely to know non-migrants) - Tells richer story than simpler, Euclidean distance measure and statistically significantly different. Key result: Identity matters but poverty also fosters exclusion from social relations that are a pre-condition for potentially valuable social interactions (in this case, info).

    13. Results: - Individual’s own attributes (identity) generally not important to explain propensity to activate a link, conditional on knowing someone - Measures of social distance highly significant … relative social position matters enormously to link likelihood. - Co-residence has biggest effect (no surprise) - Those who work off-farm a lot ask those who do not - Asymmetric gender effects: women ask men, but men don’t ask women - conditional on knowing each other, more likely to ask those who are less wealthy (face saving with elites?) - ask questions of those with more experience than i, with stronger effect for new crop than established one.

    15. Differences in determinants of network linkages across topics raises the prospect that information networks are problem-specific (which suggests interest key) rather than common to all questions (which suggests identity plays the main role). Descriptive stats: 82% all or nothing – suggests strong interlinkages. Yet marketing differs from crop production, so perhaps apparent interlinkage due to similarity of purposes?

    17. Testing strategies: after including the fitted values of the probability to ask for advice on any other question as an additional explanatory variable and test the null that the explains anything further.

    18. Testing strategies: (ii) Multinomial logit on four coarser options (ask about nothing, production only, marketing only, or ask about all) – and test IIA hypothesis. In each case, overwhelming reject the null that we can combine options (p-values = 0.00) and fail to reject the null of IIA (p-values = 1.00).

    19. Granovetter’s (1973) hypothesis: one’s most valuable connections are those exercised infrequently. People with whom one interacts frequently are not key sources of new information (even if their willingness to help is greater). This suggests a nonlinear relation between the strength of a link and its effectiveness. Because strong ties are associated with similar identities (homophily), support for this hypothesis reinforces the importance of interest in shaping information networks. We test this hypothesis using the answer to the frequency of contact question “In a normal month, how often do you talk with ___?”

    22. Testing strategy: Jointly estimate zij = Y’ t + vij vij ~ N(0, s2) (14) lijk* = az + ?z2 + X’ ß + uij uij ~ N(0, 1) (15) where z= frequency of talk, and we identify t using physical distance. People speak more frequently with those who are most similar to them and who are most proximate to them (homophilous pattern of communications).

    25. Identity matters to knowing someone and to the frequency with which one speaks with a match. The existence of a network thus strongly influenced by identity/location. But in functional terms of asking for information (i.e., activating network), identity matters mainly in relational/ordinal terms, not in terms of one’s attributes in isolation from others’. People are selective as to who they ask for particular sorts of information, building problem-specific networks from among a subset of contacts. This suggests self-interest (and not only identity) plays a key role in the process of network activation, and (less so) in network formation.

More Related