Cross-lagged panel analysis (CLPA) is one of a variety of quasi-experimentaI forms of analysis used by an increasing number of social scientists for making causal inferences (Chafee, et al., 1970; Crano, et al., 1972; DuvalI and Welfling, 1973; Shingles, 1975). Because quasi-experimentaI research is less rigorous than true experimentation, its results are typically subject to plausible rival interpretations. This is true of CLPA and has led to a number of attempts to refine the approach. However, the lack of consensus and the contradic tory solutions stemming from these efforts discourages use of the analysis. The confusion stems partly from the fact that the contributors to the debate represent independent disci plines, characterized by different research traditions and terminologies (Goldberger, 1971), and partly from the failure of those employing the procedure to develop and articulate formal causal models or an explicit theory of CLPA. This paper reviews, compares and evaluates the major approaches to CLPA. Section one presents the basic design for a two-variable, two-wave case in the context of a causal model, and section two uses the model to discuss the major cross lagged panel derivations. Sections three and four present these derivations in a four-fold typology. The final section considers the relaxation of several of the principal assump tions of the model. The paper is intended primarily for the potential users of CLPA who find the current literature bewildering. It also serves as a warning to those who would use the design based on their familiarity with any one of the principal derivations. It will be shown that the various approaches differ in impor tant ways from one another and that their relative value de pends upon the research context in which one is working. Those who have already mastered CLPA should find the discus sion useful for organizing their own thinking. I have tried to avoid an excessively mathematical discussion. A reading knowledge of basic scalar algebra suffices for a complete understanding of the topic.
[1]
F. Fisher.
Causation and specification in economic theory and econometrics
,
1969,
Synthese.
[2]
H. Blalock,et al.
Applied multivariate analysis and experimental designs
,
1975
.
[3]
R. Darlington,et al.
Multiple regression in psychological research and practice.
,
1968,
Psychological bulletin.
[4]
H. Simon,et al.
Spurious Correlation: A Causal Interpretation*
,
1954
.
[5]
D. Campbell,et al.
EXPERIMENTAL AND QUASI-EXPERIMENT Al DESIGNS FOR RESEARCH
,
2012
.
[6]
R. Gordon.
Issues in Multiple Regression
,
1968,
American Journal of Sociology.
[7]
Anthony P. M. Coxon,et al.
Contemporary Problems in Statistics
,
1985
.
[8]
J. Kemeny.
A Philosopher Looks At Science
,
1960
.
[9]
Some comments on "Techniques for estimating the source and direction of influence in panel data."
,
1970
.
[10]
D. Campbell,et al.
More plausible rival hypotheses in the cross-lagged panel correlation technique.
,
1969
.
[11]
F. M. Andrews,et al.
Detecting Causal Priorities in Panel Study Data
,
1964
.
[12]
F. Fisher.
A Correspondence Principle for Simultaneous Equation Models
,
1970
.
[13]
Douglas A. Hibbs,et al.
Problems of Statistical Estimation and Causal Inference in Time-Series Regression Models
,
1973
.
[14]
F. F. Stephan,et al.
Statistical Procedures and their Mathematical Bases.
,
1936
.
[15]
Donald T. Campbell,et al.
Does Intelligence Cause Achievement?: A Cross-Lagged Panel Analysis
,
1972
.
[16]
Arthur S. Goldberger,et al.
Econometrics and psychometrics: A survey of communalities
,
1971
.
[17]
Otis Dudley Duncan,et al.
Some linear models for two-wave, two-variable panel analysis.
,
1969
.
[18]
D. R. Heise.
Problems in Path Analysis and Causal Inference
,
1969
.
[19]
Steven H. Chaffee,et al.
Mass Communication and Political Socialization
,
1970
.
[20]
K. Land.
Principles of Path Analysis
,
1969
.
[21]
Donald C. Pelz,et al.
Heise's Causal Model Applied
,
1970
.