Citation

Greenberg DF. J. Quant. Criminol. 2010; 26(4): 437-443.

Copyright

(Copyright © 2010, Holtzbrinck Springer Nature Publishing Group)

DOI

10.1007/s10940-010-9111-9

PMID

unavailable

Abstract

Virtually every textbook devoted to longitudinal statistical methods begins by touting their advantages over cross-sectional methods: for purposes of studying change it is helpful to have actual data on change. Also, information about temporal ordering helps to establish a relationship between variables as causal. Each of these alleged advantages calls for elaboration.In principle, change can be studied with cross-sectional data--provided one has a properly specified equation representing the relationships among a set of variables and is able to estimate its parameters without bias. However, the researcher cannot always be confident that a posited relationship accurately describes the causal relationships at work, or that its parameters can be estimated consistently. For example, if x and y are mutually dependent, an OLS regression of y on x will mischaracterize the influence of x on y. While an instrumental variable estimation can help, it may require stringent assumptions if implement...

Language: en