report that removing such women does not much change aggregate estimates of unmet need for contraception.)

For this report, we rely most on the direct measure, for two reasons: the direct items allow consideration of mistiming as well as a desire for complete cessation of childbearing; and the major objection to the direct measure of wantedness is that it produces underreporting. But if significant percentages of pregnancies appear unwanted using a measure that is probably biased then the argument that unwantedness is a big problem is strengthened.

We need to distinguish between the wantedness of conceptions and of births. It is likely that many women change their minds about the impending birth during the course of the pregnancy, either becoming reconciled to the birth or regretting an initially wanted conception. Rosenzweig and Wolpin (1993), using data from a survey in the United States for which women were interviewed by random assignment either before or after a birth, found an 8 percent decrease in wantedness after the birth, which suggests that at least in this population regret may be more common than rationalization. In an extreme case, a woman may not have intended the sexual intercourse that produced the conception, or the conception, but report as pregnancy goes on that she wants the birth, perhaps not seeing any acceptable alternative. Such an "intended birth" would not be regarded as an indicator of good reproductive health. Conversely, a change of intentions about an initially wanted pregnancy could come about because of a change in circumstances during a pregnancy—abandonment or abuse by the father, for example. It has been argued that conceptions are intended even when births are not; this may be the case, for example, in cultures where a new or potential wife's proof of fecundity is highly valued.

Data on abortions provide evidence both of the extent of unintended pregnancy and of one of its major potentially harmful consequences in developing countries. But existing data are very incomplete. Direct estimates based on household surveys produce implausibly low estimates of the prevalence of induced abortion. In the United States, for example, where most states had liberalized abortion laws even before laws against first-trimester abortions were ruled unconstitutional in 1973, confidential surveys of providers suggest that induced abortions are more than twice as common as is reported in household surveys (Jones and Forrest, 1992). There has been some recent experimentation with survey methods in both developed and developing countries, and survey researchers may have given up too easily on the prospect of measuring abortion with direct questions (see Huntington, Mensch, and Miller, 1996; Laumann et al., 1994:457). The estimates produced by the Alan Guttmacher Institute and the World Health Organization (which we use in the text) are based