Role of Time and Space in Social Science Research

Role of Time and Space in Role of Time and Space in Social Science ResearchSocial Science Research

Yu XieYu XieUniversity of MichiganUniversity of Michigan

Third PrincipleThird Principle

Patterns of population variability vary with Patterns of population variability vary with social context, which is often defined by social context, which is often defined by time and space. time and space.

““Social Context Principle”Social Context Principle”

Different “Regimes” of Different “Regimes” of VariabilityVariability Social contexts are different from social groups in that Social contexts are different from social groups in that

the former are self-contained social systems with natural the former are self-contained social systems with natural boundaries, for example by time and space. boundaries, for example by time and space.

Patterns of individual variability may be governed by Patterns of individual variability may be governed by “relationships” between individuals that are not reducible “relationships” between individuals that are not reducible to individuals’ attributes. to individuals’ attributes.

Patterns of individual variability may be governed by Patterns of individual variability may be governed by macro-level conditions such as “social structure,” macro-level conditions such as “social structure,” “political structure,” or “culture,” which may be “political structure,” or “culture,” which may be discontinuous and fixed. discontinuous and fixed.

Collective action may lead to changes of macro-level Collective action may lead to changes of macro-level conditions and human relationships –major sources of conditions and human relationships –major sources of social change. [Premise of Marxism.]social change. [Premise of Marxism.]

Accounting for Heterogeneous Accounting for Heterogeneous ResponsesResponses

More difficult to handle, a degree of freedom More difficult to handle, a degree of freedom problem.problem.

Possible with nested data, assuming that patterns Possible with nested data, assuming that patterns of relationships are homogeneous (or following a of relationships are homogeneous (or following a distribution) within social contexts (by time or distribution) within social contexts (by time or space).space).

kk is allowed to vary across k (k=1,…K), social is allowed to vary across k (k=1,…K), social context, but is fixed within k. For example:context, but is fixed within k. For example:

YYikik = = kk + + kkDDikik + + ikikkk = = ++zzkk++k k kk = = ++zzkk++kk

Application of the Social Context Principle. Application of the Social Context Principle.

CommentsComments Comment 1Comment 1: It is possible to impose a parametric : It is possible to impose a parametric

assumption on individual-level dassumption on individual-level d ii, but the results are , but the results are dependent on the assumption. (Bayesian approach.)dependent on the assumption. (Bayesian approach.)

Comment 2Comment 2: Nested (or hierarchical) structure could be : Nested (or hierarchical) structure could be used for time-variation or spatial variation. A key used for time-variation or spatial variation. A key assumption: there are common features that are shared by assumption: there are common features that are shared by different elements in a common social context. different elements in a common social context.

Comment 3Comment 3: If variations across social contexts are : If variations across social contexts are systematic, they can be modeled (i.e., multi-level models, or systematic, they can be modeled (i.e., multi-level models, or hierarchical linear models, random-coefficient models, and hierarchical linear models, random-coefficient models, and growth-curve models). If such variations are left as growth-curve models). If such variations are left as observed (or saturated), we have the fixed-effects model. observed (or saturated), we have the fixed-effects model.

Two Important Features of Time Two Important Features of Time and Spaceand Space

1. 1. CommonalityCommonality: sameness or similarity : sameness or similarity across different units of analysis within a across different units of analysis within a common domain of time/space. This is common domain of time/space. This is the basis for multi-level modeling. the basis for multi-level modeling.

2. 2. Continuity/contiguityContinuity/contiguity: social contexts are : social contexts are more different if they are further apart in more different if they are further apart in time/space. This is the basis for time/space. This is the basis for temporal/spatial autocorrelation. temporal/spatial autocorrelation.

Why Study Variation Why Study Variation by Time and Space? by Time and Space?

Accounting for temporal and spatial Accounting for temporal and spatial variation is of intrinsic scientific value.variation is of intrinsic scientific value.

Temporal variation allows us to study Temporal variation allows us to study dynamic processes and test theoretical dynamic processes and test theoretical hypotheses.hypotheses.

If we cannot observe dynamic processes, If we cannot observe dynamic processes, under strong assumptions, spatial variation under strong assumptions, spatial variation provides an alternative (“reading history provides an alternative (“reading history sideways”)sideways”)

This is the basis for comparative sociology. This is the basis for comparative sociology.

Research Designs for the Study of Research Designs for the Study of Temporal VariationTemporal Variation

Time series study: N=1, t =1, …T. Almost Time series study: N=1, t =1, …T. Almost always the aggregate for a population. always the aggregate for a population.

Trend study: iTrend study: itt=1…N=1…Ntt, t=1,..T, not same i’s , t=1,..T, not same i’s observed over time. observed over time.

Panel study, i=1…N, t=1,..T, N >>T.Panel study, i=1…N, t=1,..T, N >>T.Same i’s observed over time. Same i’s observed over time.

Event history study, n=1…N, t=1,..T,T >>NEvent history study, n=1…N, t=1,..T,T >>N Same i’s observed over time. Same i’s observed over time.

Spatial AnalysisSpatial Analysis

Is similar to trend study. Is similar to trend study. Cannot observe the same persons in Cannot observe the same persons in

multiple times. multiple times.

Focus of the Classical Chicago Focus of the Classical Chicago School of SociologySchool of Sociology

Importance of residential location Importance of residential location “locational attainment”.“locational attainment”.

Effects of neighborhood on socioeconomic Effects of neighborhood on socioeconomic attainment.attainment.

Earlier work on assimilation experiences of Earlier work on assimilation experiences of European immigrants.European immigrants.

Controlling Heterogeneity Via Fixed Controlling Heterogeneity Via Fixed Effects ModelEffects Model

Is done at the individual level in a panel Is done at the individual level in a panel study == assuming that there are study == assuming that there are unobserved, fixed characteristics at the unobserved, fixed characteristics at the individual level.individual level.

Is done at the contextual level in a multi-Is done at the contextual level in a multi-level or trend study == assuming that there level or trend study == assuming that there are unobserved, fixed characteristics at are unobserved, fixed characteristics at the contextual level.the contextual level.

Time versus Space regarding Time versus Space regarding Commonality and ContinuityCommonality and Continuity

Commonality is usually a stronger assumption Commonality is usually a stronger assumption for a spatial entity than for a unit observed over for a spatial entity than for a unit observed over time. time. Boundary could be arbitraryBoundary could be arbitrary Assumed commonality may not be actually trueAssumed commonality may not be actually true

Temporal continuity is not symmetrical. Past Temporal continuity is not symmetrical. Past affects future. affects future.

““11stst Law of geography” Law of geography” distance decay distance decay function in space.function in space.

Space could be social rather than geographic. Space could be social rather than geographic.

Two Primary Demographic Sources Two Primary Demographic Sources of Social Changeof Social Change

1. Cohort replacement.1. Cohort replacement. Cohort is no more than one way to define social Cohort is no more than one way to define social

grouping. Significance depends on social consequence. grouping. Significance depends on social consequence. There are within-cohort differences: between and within There are within-cohort differences: between and within

other forms of social grouping.other forms of social grouping. It intersects with period: If there is a non-repeatable It intersects with period: If there is a non-repeatable

significant juncture early in life (such as education), it’s significant juncture early in life (such as education), it’s social consequences are likely to be cohort effects, social consequences are likely to be cohort effects,

2. Life-course.2. Life-course. Within-cohort change.Within-cohort change. Primarily a function of age.Primarily a function of age. Serves also as a principle of social grouping. There is Serves also as a principle of social grouping. There is

variability between and within other forms of social variability between and within other forms of social grouping. grouping.

Two Sources of “Autocorrelation”Two Sources of “Autocorrelation” 1. For different individuals (i=1,..N) in a common 1. For different individuals (i=1,..N) in a common

environment (k), they may share a common error environment (k), they may share a common error (variance component model). E.g, (variance component model). E.g,

YYikik = = + + DDikik + + kk + + ikik

2. Individuals in adjacent social contexts (say k and 2. Individuals in adjacent social contexts (say k and k+1) are more similar to each other than individuals k+1) are more similar to each other than individuals located in distant social contexts (social contagion located in distant social contexts (social contagion effects). That is,effects). That is,

k k andand k+1k+1 may be correlated. may be correlated. We usually only consider autocorrelation of order 1 We usually only consider autocorrelation of order 1 across contexts (immediate neighbor effect). across contexts (immediate neighbor effect).

Problem of Age, Period, and CohortProblem of Age, Period, and Cohort

Linear dependency: Linear dependency: Cohort + age = period.Cohort + age = period.Cohort = period - ageCohort = period - age

Various devices have been made. They Various devices have been made. They all require untestable assumptions and all require untestable assumptions and thus debatable.thus debatable.

Problem is often a substantive one. Problem is often a substantive one. Requires knowledge of the subject matter. Requires knowledge of the subject matter.

Reading History Sideways?Reading History Sideways?

Arland Thornton’s 2001 PAA presidential Arland Thornton’s 2001 PAA presidential address and 2005 book.address and 2005 book.

Spatial variation should not be mistaken as Spatial variation should not be mistaken as temporal variation.temporal variation.

Major reason for the failure of social Major reason for the failure of social Darwinism.Darwinism.

Reasons: contextual variability, such as Reasons: contextual variability, such as culture, and path-dependency. culture, and path-dependency.

Example 1 (Xie and Hannum 1996, AJS)Example 1 (Xie and Hannum 1996, AJS)

Motivated by Nee’s market transition theory.Motivated by Nee’s market transition theory. Research question 1: are earnings returns Research question 1: are earnings returns

to education higher in Chinese cities that to education higher in Chinese cities that have experienced more rapid economic have experienced more rapid economic growth than in other cities? growth than in other cities?

Research question 2: are earnings returns Research question 2: are earnings returns to political capital (CCP) lower in Chinese to political capital (CCP) lower in Chinese cities that have experienced more rapid cities that have experienced more rapid economic growth than in other cities?economic growth than in other cities?

Xie and Hannum’s ResultsXie and Hannum’s Results

The study is based on a multi-level model.The study is based on a multi-level model. Two main findings:Two main findings:

No regional variation in the influences of No regional variation in the influences of political capital and gender on earnings.political capital and gender on earnings.

Education returns are lower, not higher, in Education returns are lower, not higher, in cities that have experienced more economic cities that have experienced more economic development (but with higher intercepts) than development (but with higher intercepts) than in other cities. in other cities.

Example 2 (Hauser and Xie 2005, Example 2 (Hauser and Xie 2005, SSR)SSR)

An extension of Xie and Hannum, using An extension of Xie and Hannum, using trend data between 1988 and 1995 in trend data between 1988 and 1995 in 35 35 Chinese cities.Chinese cities.

Temporal-spatial design.Temporal-spatial design. Capitalizing on differential rates of Capitalizing on differential rates of

economic growth by region. economic growth by region.

Hauser and Xie’s FindingsHauser and Xie’s Findings Net returns to schooling almost doubled for both Net returns to schooling almost doubled for both

men and women during this period. Returns to men and women during this period. Returns to party membership, net of other factors, more party membership, net of other factors, more than doubled. than doubled.

These increases in returns to human capital and These increases in returns to human capital and political capital do not account for the political capital do not account for the remarkable rise in the overall level of inequality.remarkable rise in the overall level of inequality.

Increases in returns to schooling were Increases in returns to schooling were depressed in cities experiencing greater levels depressed in cities experiencing greater levels of economic growth in the intervening period. of economic growth in the intervening period.

Example 3 (Raymo and Xie 2000, Example 3 (Raymo and Xie 2000, ASR)ASR)

Research Questions: Research Questions: How does educational homogamy change in How does educational homogamy change in

response to economic development?response to economic development? Is there a unique East Asian regime in Is there a unique East Asian regime in

educational homogamy? educational homogamy? The study is again temporal-spatial, The study is again temporal-spatial,

although relying only on data from four although relying only on data from four geographic units. geographic units.

Raymo and Xie’s FindingsRaymo and Xie’s Findings

There has been declines in the strength of There has been declines in the strength of educational assortative mating.educational assortative mating.

But there appears a convergence across But there appears a convergence across US. China, and Japan. US. China, and Japan.

Thus, no unique Confucian pattern of very Thus, no unique Confucian pattern of very high levels of educational assortative high levels of educational assortative mating. mating.