DIVORCE EFFECTS'AND CAUSALITY IN THE SOCIAL SCIENCES 43 interaction,and situational aspects of the family are unclassified with respect to this key variable-a environment.Thus these studies by and large lack major shortcoming in a study of divorce effects:14 information on:parental personality and attitudes, Evidently the variable is not a reliable indicator of parents'own early demographic experience such as pre-existing conflict in a family,whether sub- premarital pregnancy,parental difficulties such as sequently divorcing or not. alcoholism,mental illness,criminal record,quality Just 12 per cent of couples separating between of the marital relationship,parental relationship 1965(age 7)and 1974 (age 16)in the NCDS were posi- with the children,spousal abuse,child abuse,and tive on the binary domestic tension'indicator.This other forms of family stress,conflict,and dysfunc- figure contrasts sharply with the estimate ofover one tion,and employment/unemployment in the years in four women estimated to suffer from domestic immediately preceding separation.Where present,violence at some time in their lives(British Medical information on these features is usually of poor or Association,1998).Domestic violence is,of course, summary quality.All of these studies lack informa-a great deal more serious than domestic tension'and tion on the reason for the separation,from either so the record of tension in the household appears parent's point of view. particularly doubtful.This is especially so since one We recall that the aim of controlling for back- would expect domestic violence to be more com- ground factors is to try to ensure that,as at the mon among divorcing than stable couples.15 The point of separation,families that subsequently studies considered here do not exclude the possibi- remain stable,divorce,or suffer bereavement,and lity that family conflict rather than divorce is the are reconstituted or become lone-parent families,source of divorce-outcome associations.That are alike in every respect relevant to the outcomes family conflict may be a significant element of this of interest:that the control variables are sufciently compre-association is suggested by Jekielek's(1998)careful bensive and specific that families with each combination of their study.The interactions she reports between conflict levels immediately prior to disruption are so similar that they and later outcomes suggest that post-divorce experi- could bave been assigned randomly to the divorce,bereavement, ence may be positive when the family is particularly and stable family etc.categories (see Lieberson,1985: troubled prior to divorce.The interaction effect ch.6).The absence of family and individual-level found by Chase-Lansdale et al.,alluded to earlier, variables of the type noted here as missing makes it might be interpreted in similar fashion,if behaviour very unlikely that the regressions carried out in these problems at 7 were indicative of an adverse family studies can be considered,at present,to have come environment prior to divorce.16 close to this condition.As a result,there are a large Confict is not the only prima facie unmeasured number of alternative possible explanations for the variable and the conflict hypothesis not the only association between parental divorce and later out-alternative.Another possibility is that couples who comes that have not been ruled out.Indeed,the divorce are selected for personal characteristics or existence of a net association remains to be estab- parenting styles that are,in themselves,conducive lished. to adverse outcomes (Maughan and McCarthy, A key alternative to the proposition that parental 1997;McGue and Lykken,1992).Just two studies of divorce causes long-term harm is that family conflict the 11 examined here include any indicator of existing prior to divorce is at the root of the higher parent-child interaction,and indicators of parental frequency of adverse outcomes among the children personal attributes are uncommon in these analyses of divorce.In order to test that hypothesis,good also.Nevertheless,in western societies it is widely measures of marital conflict,preceding the separa-believed that divorce does indeed adversely affect tion,would be needed.Just four of the eleven the children and that couples should 'stay together studies include any measure of marital conflict,and for the sake of the children'.In this context,it in the two cases that use the NCDS the indicator is seems possible,perhaps even likely,that one or known to be of poor quality.Information is missing both members of a couple with children who do on the NCDS domestic tension'indicator for 17 per actually resort to divorce either value their own cent of couples divorcing when the study child was wishes more than those of their children,or do not aged between 7 and 16.Thus,nearly a fifth of families consider the children's welfare at all,or believe that             5        "   A         4              G                                           B        N          ?                    +     "              4    ?         "                                3            "            A        $       *                                    "  "      %  # J %&'-A  E* 5                "  "                        +                                        !               + "                    B                        !             B          H                 B          <)
     "         !      <)
 2  4    %.                . %E 5              "  >  /        2 3 4 %C              B              H  %9          %&E- # .* %&.C #%E*   <)
       2  4   5                        3           #7  6 +  %&&'*
             2  4                  5                        %- 5               B           >    5  B               H""4 #%&&'*     5         B                             5    3    )J                     .            %E )B             B         +                                 #6   6)  %&&.( 6K   J"" %&&9* H       %%          >                    <               3         2      "   4 !         "                                       4