How General is Control Balance Theory? Evidence from Ukraine

Abstract

This study tests control balance theory using interview data from a random sample of adults in a large city in Ukraine. This is the first empirical assessment of the theory to employ a random sample of adults in a nonwestern culture, and it is one of only two studies to incorporate Tittle’s theoretical revisions and measurement strategy for the control ratio. Although we found no evidence of a relationship between projected deviance and a dichotomous measure of control imbalance, respondents with a relatively large control imbalance were significantly more likely to project deviance than were other respondents. In addition, findings provide partial support for predicted contingent relationships involving constraint and self-control. We discuss possible ways in which the socio-cultural circumstances of Ukraine help to explain these findings.

Keywords:

• control balance theory
• cross-cultural
• Ukraine

Notes

¹ The theory also applies to aggregate units (Tittle, 1995).

² As is customary in household surveys, targeted respondents who could not be interviewed were replaced randomly using an initial oversampling of approximately 65%. A substantial number of replacements were made for targeted households in which a suitable respondent could not be located or scheduled for interview within the allocated call-back time, but the survey replacement rate is only slightly higher than is usual in surveys conducted in US cities (for example, see Lattimore, Tittle, & Grasmick, 2006 for a description of the Oklahoma City Survey procedures) and is comparable to surveys conducted in other countries (Couper & Leeuw, 2003; Vågerö et al., 2008; Kordos, 2005). Furthermore, overall survey completion rates are exceptionally high, with less than 1.5% missing on any given variable.

³ We repeated all analyses using a measure of respondents’ criminal activity in the past five years. The pattern of results was similar to findings based on projected deviance, although several relationships failed to reach statistical significance. This is not surprising given the fact that this measure of deviance extends over the past five years while measures of control ratios refer to the present.

⁴ We experimented with various clustering algorithms (e.g. k-means and k-medians) but opted against their use due to inconsistencies with the global control ratio (e.g., individuals with identical global control ratios sometimes were placed into separate zones). Additional research is needed to establish the relative importance of domain-specific control ratios.

⁵ The two measures are highly correlated (r = .96).

⁶ On average, domain-specific control ratios were largest for health and physical appearance (1.29) and conditions of living (1.27), followed by friends (1.17), strangers (1.17), neighbors (1.16), work (1.16) and family (1.15).

⁷ We used the mean for small theft as a proxy for intentionally avoiding paying for something.

⁸ Grasmick et al. (1993) showed that an additional item pertaining to physical activity (i.e. I have more energy and need for activity than other people) does not hang well with the other items and decreases overall reliability of the scale. Thus, it has become customary to construct the scale with the 23 items used here (see, e.g. Antonaccio & Tittle 2008; Botchkovar et al., 2009).

⁹ We conducted sensitivity analyses using additional control variables such as peer deviance and found similar substantive patterns of results, although, as expected, coefficients were attenuated.

¹⁰ Of the 700 adults interviewed, a total of 91 failed to provide data sufficient for computing control ratios and were eliminated from our analysis. Three additional cases were deleted due to missing data on one or more of the deviance items. Missing values on all remaining variables were imputed using the EM algorithm method (less than 1.5% of cases for any given variable).

¹¹ Given the small number of clusters, statistical power is limited. However, multilevel over dispersed Poisson models yielded substantively similar results (available upon request).

¹² Sensitivity analyses using quarters and fifths produced substantively similar findings (available upon request).

¹³ Ordinal logistic models predicting specific acts of deviance (results not presented) yielded substantively similar findings, except for a statistically significant positive effect of control imbalance on theft of something worth more than $5.

¹⁴ Due to reduced sample size and statistical power, we use a relaxed level of statistical significance.

¹⁵ Somewhat puzzling, estimates from supplementary ordinal logistic regression analyses of specific deviant acts (results not presented) revealed the strongest effects for physically harming or threatening to harm another person on purpose, suggesting that individuals with a relatively large control surplus are more likely to engage in violence than drink alcohol excessively or commit property offenses. Although this finding is compatible with earlier results showing a relationship between control surpluses and assaultive behaviors (e.g. Baron & Forde, 2007; Baron, 2009), it potentially challenges CBT in that physically harming or threatening to harm another person on purpose was rated by individuals with a large control imbalance as the least gratifying of all measured acts.

Additional information

Notes on contributors

Lorine A. Hughes

Lorine Hughes is Associate Professor of Criminology and Criminal Justice at the University of Nebraska at Omaha. Her research interests include youth street gangs, criminological theory testing, and social networks.

Olena Antonaccio

Olena Antonaccio is Assistant Professor of Sociology at the University of Miami. Her research interests include theory testing and development and comparative criminology.

Ekaterina V. Botchkovar

Ekaterina Botchkovar is Associate Professor of Criminology and Criminal Justice at Northeastern University. Her research interests include criminological theory development and comparative criminology.