Newlyweds’ Optimistic Forecasts of their Marriage: For Better or For Worse?

Abstract

Newlywed spouses routinely hope and believe that their relationships will thrive, but theoretical accounts differ on whether optimistic projections such as believing that one’s marriage will improve are sources of strength, random forecasting errors, or self-protective mechanisms. To test these opposing perspectives, we asked 502 newlywed spouses in 251 marriages to predict how their overall feelings about their relationships would change over the following four years, and we then compared these reports to their prospective marital satisfaction trajectories. Nearly all spouses predicted their marital satisfaction would remain stable or improve over the following four years. Marital satisfaction declined on average despite this high overall level of optimism. Wives with the most optimistic forecasts showed the steepest declines in marital satisfaction. These wives also had lower self-esteem and higher levels of stress and physical aggression toward their partners initially. Thus, believing that one’s marriage will improve does not make it so and instead may paradoxically mask risky relationships among women. These findings may be important in helping to understand low rates of premarital counseling utilization by showing that nearly all couples overestimate the durability of their existing satisfied feelings at the start of their marriage. Future research is needed to understand the psychological processes allowing couples to commit to and stay in risky relationships.

Despite ample evidence that divorce and relationship distress are common (e.g., Whisman, Beach, & Snyder, 2008), married and unmarried individuals alike believe their own odds of divorce are low, even after they are reminded of the actual divorce rate (Baker & Emery, 1993; Fowers, Lyons, Montel, & Shaked, 2001). Does believing that one’s relationship is immune from adverse outcomes increase the likelihood of marital success years later? Can positive projections such as believing that one’s relationship will improve actually promote more satisfying relationships, or are these optimistic forecasts simply random affective forecasting errors – or perhaps even marks of troubled relationships? This study aimed to address these questions, examining newlywed spouses’ predictions for how their marriage would change, the extent to which spouses’ initial projections about their marital trajectories corresponded to their actual four-year marital trajectories, and the initial characteristics of spouses with different types of marital forecasts.

Romantic partners frequently engage in a variety of cognitive processes that allow them to see each other and their relationships in a positive light and maintain their commitments to each other. When focused on the immediate present, spouses make benign attributions for irritating behaviors (e.g., Bradbury & Fincham, 1990), attend more to information that supports and strengthens their relationship (Miller, 1997), and believe that their partners meet their ideals more than their partners report they actually do (e.g., Murray, Holmes, & Griffin, 1996). Spouses also enhance their retrospective views of their relationships, positively distorting the extent to which their relationships have recently improved even in the context of declining satisfaction (Karney & Frye, 2002). These types of positive cognitive processes are generally seen to promote marital well-being (cf. McNulty & Fincham, 2012).

Emerging evidence supports the presence of forward-looking cognitive biases in relationships as well. Consistent with people’s tendency to hold positive beliefs about their futures (e.g., Weinstein, 1980), dating partners are more optimistic in their predictions about how likely their relationships are to last than are their roommates and parents, and they are more confident in these predictions (MacDonald & Ross, 1999). Romantic partners also exhibit forward-looking biases that enhance their odds of staying in their relationships (Arriaga, Capezza, Goodfriend, Rayl, & Sands, 2013), such as being overly pessimistic about how affected they would be by negative relationship events (e.g., their partner being disrespectful, rude, or emotionally distant; Green et al., 2013) or if their relationship were to end (Eastwick, Finkel, Krishnamurti, & Loewenstein, 2008; Gilbert, Pinel, Wilson, Blumberg, & Wheatley, 1998).

Nothing is known, however, about how individuals believe their feelings about their relationships will change over time, or whether believing the relationship will improve enhances subsequent marital well-being. Individuals generally expect that the future will be better than the present (which is better than the past; e.g., Brickman, Coates, & Janoff-Bulman, 1978); this “unrealistic optimism” is hypothesized to promote mental health and greater well-being by creating a cognitive frame that focuses attention on positives, minimizes negatives, and fosters benevolent interpretations of ambiguous events (Taylor & Brown, 1988; Taylor & Brown, 1994). Individuals who want their circumstances to improve may also be more likely to implement the steps needed to achieve that goal, ultimately making them more likely to reach it (e.g., Gollwitzer, 1993, 1999). Accordingly, in the context of marriage, individuals who enter marriage believing that their relationships will only get better should experience the most satisfying trajectories compared to individuals with less optimistic views of how their relationship will unfold (e.g., expecting their feelings will decline or remain stable).

Competing theoretical perspectives make different predictions regarding the nature of optimistic marital forecasts. Harold Kelley (1983) argued that partners’ estimates of their relationship stability are “determined by the interpretations made of the past…and by the intelligence with which the future is imagined, for example, by whether account is taken of external factors that are not presently salient but will come into play when disruption threatens. These considerations should make us wary of taking at face value respondents’ predictions of the future of their relations” (p. 310). Consistent with this view, broader research on affective forecasting (see Wilson & Gilbert, 2003) suggests that forecasts are likely to be highly biased and thus should not predict what actually unfolds. Predictions for the future tend to correspond more closely to what individuals want to see happen than what is actually likely (e.g., Babad & Katz, 1991; Epley & Dunning, 2000). For example, in the study of dating couples’ forecasts of the stability of their relationships described earlier, outside observers’ predictions of relationship longevity were more accurate than partners’ own predictions (MacDonald & Ross, 1999), likely due to partners’ tendency to minimize the negative aspects of their relationships. Individuals also routinely err when predicting how long future emotional states will last (e.g., Gilbert et al., 1998), so it would not be surprising for newlywed spouses to overestimate the durability of their existing, highly satisfied feelings. We would also expect errors even among spouses who grounded their predictions in past histories of their relationships (e.g., “I’ve been feeling more satisfied, so my relationship must be getting better”) given that retrospective recall is positively biased (Karney & Frye, 2002). Thus, according to this view, if individuals’ forecasts about their future feelings often prove untrue due to a variety of cognitive biases, then newlywed spouses’ forecasts about how their marital satisfaction will change are unlikely to be meaningful.

Taking this argument one step further, it is possible that marital forecasts are inversely associated with actual marital trajectories. Believing that the relationship will get better may set spouses up for disappointment when it does not (as is true on average; see Aron, Norman, Aron, & Lewandowski, 2002), and this disappointment may be more potent than if spouses had not expected improvements (e.g., Geers & Lassiter, 1999; Huston, Caughlin, Houts, Smith, & George, 2001). Moreover, spouses may only report believing that things will get better when there are concerns with the status quo: individuals in risky relationships may compensate for their current (negative) circumstances by adopting a more extended evaluative frame of their relationships that includes positive views of the future (Martz et al., 1998). More generally, hoping that circumstances will improve is a common response to couple distress, particularly when investments in the relationship are already high (Rusbult, Zembrodt, & Gunn, 1982), as would be the case among newlyweds. Paradoxically, then, positive forecasts (e.g., believing that one’s marriage will improve) may mark risky relationships, which then go on to struggle.

To test these opposing views, we asked newlywed spouses to predict how their feelings about their relationships would change over the next four years and subsequently examined these reports in relation to their four-year, eight-wave marital satisfaction trajectories. If predicting that things will get better is beneficial for marital well-being, we would expect less severe declines in marital satisfaction among individuals with these optimistic forecasts. If, however, forecasting improvement is a random affective forecasting error, we would expect no association between initial forecast and subsequent marital satisfaction. Alternatively, if forecasting improvement creates unrealistic expectations and/or marks problematic relationships, we would expect steeper declines in marital satisfaction among spouses with more positive forecasts.

Given that this question was explicitly about change, we focused mainly on how forecasts corresponded with the slope of change in marital satisfaction, controlling for intercept, in order to compare predictions of change with prospective, observed change. Nonetheless, we also examined initial differences in satisfaction by forecast type to explore whether predictions of change covaried in a systematic way with how things were going at the time. Here we examined whether individuals in satisfying relationships were especially likely to make optimistic forecasts because they believed things could only get better, for example, or whether individuals in relatively dissatisfying relationships made more optimistic forecasts because they had more room for improvement.

We further tested these conceptual models by examining the initial characteristics of spouses with different marital forecasts. To identify an inclusive set of risk factors, we drew on the Vulnerability-Stress-Adaptation (VSA) model (Karney & Bradbury, 1995). The VSA model posits that relationship satisfaction (and changes in satisfaction over time) are governed by the quality of couple interaction, the stresses couples encounter, and the traits partners bring to marriage. To assess the quality of couple interaction, we included spouses’ self-reports of physical aggression as well as their observed negative communication. Although physical aggression predicts poor marital outcomes (e.g., Lawrence & Bradbury, 2001), many couples who exhibit physical aggression in their relationships either do not view it as a problem or view it as being relatively unstable (Ehrensaft & Vivian, 1996), suggesting that they may downplay this risk by believing that their relationships will improve. We also included observational measures of negative communication to allow for the possibility that marital forecasts might reflect couples’ reactions to less severe negative interaction patterns (as compared to the self-report measure of physical aggression). To assess external stress, we included spouses’ reports of recent negative life events (i.e., acute stress). We focused on acute rather than chronic stress (see Randall & Bodenmann, 2009 for extended discussion of this distinction) to test the possibility that spouses may report believing that the relationship will get better simply because they have been under recent strain which is likely to change over time (as compared to chronic stress, which by definition is more constant). To assess partners’ personality, we included self-esteem and trait optimism. Self-esteem is particularly relevant to marital forecasting because it has been hypothesized that individuals may adopt unrealistically positive views for the future as part of a motivated process to protect their self-esteem and reduce fears about the future (e.g., Hoorens, 1993; Weinstein & Klein, 1996), suggesting that individuals with low self-esteem would be especially likely to hope that the relationship will improve. Lastly, we included trait optimism to test whether individuals who believe that their relationships will improve are simply more optimistic in general or whether marital forecasts represent a distinct construct.

The different theoretical models we have outlined provide competing predictions regarding the pattern of initial differences that might be found between individuals with different forecasts. If more optimistic forecasts are beneficial, they should be associated with characteristics that would promote such success, such as greater optimism, less negative communication, and lower levels of stress. If marital forecasts were random cognitive biases or they exert their influence solely through contrast effects (e.g., spouses who expected more were more disappointed), we would expect no differences in initial characteristics. If, however, more optimistic forecasts reflect existing concerns, they should be associated with risks such as lower self-esteem, poorer communication, and/or higher levels of external stress.

Method

Participants

Two studies were conducted in a Central Florida community surrounding a major state university (Ns = 82 couples and 169 couples). In both studies, couples were recruited with (a) advertisements in community newspapers and bridal shops and (b) invitations sent to eligible couples who had completed marriage license applications in the county. All couples were screened for eligibility in a telephone interview. Inclusion required that this was the first marriage for each partner; the couple had been married less than 6 months; each partner was at least 18 years of age; each partner spoke English and had completed at least 10 years of education (to ensure comprehension of the questionnaires); couples did not have children; and wives were not older than 35. Eligible couples, after providing oral consent, were scheduled for an initial laboratory session.

Participants were of comparable age across samples, with spouses in their mid-20s and husbands being slightly older than wives on average (Table 1). Most participants were Caucasian (> 80%) and Christian (> 60%). Sixty-five percent of couples reported living together before marriage, and 31% became parents at some point during the study. We combined the samples (total N = 251 couples) because all couples met identical selection criteria; the studies used highly similar procedures, measures, and designs; and doing so afforded more power.

Table 1.

Demographics

Spouse	Age		Years of education		Full-time employed	Full-time student	Yearly income		Caucasian	Christian
	M	SD	M	SD	(%)	(%)	Median	SD	(%)	(%)
Sample 1 (N = 82 Couples)
Husband	25.12	3.32	16.43	2.22	40%	54%	$5K–$10K	$4.83K	83%	59%
Wife	23.67	2.77	16.35	1.77	39%	50%	$5K–$10K	$4.41K	89%	59%
Sample 2 (N = 169 Couples)
Husband	25.53	4.13	16.48	2.33	59%	35%	$5K–$10K	$7.21K	94%	66%
Wife	23.84	3.60	16.32	2.01	45%	43%	$0K–$5K	$5.41K	86%	63%

Note. The relatively low income level of participants reflects the fact that many were full-time students at the baseline assessment.

Procedure

Before their laboratory session, participants were mailed questionnaires to complete at home and bring with them to their appointment, with a letter instructing partners to complete all questionnaires independently. Upon arriving to the session, spouses completed a written consent form approved by the local human subjects review board and then participated in problem-solving discussions and completed additional measures. Couples were then paid for participating (Sample 1 = $50, Sample 2 = $70).

At approximately 6-month intervals subsequent to the initial assessment, couples were re-contacted by telephone and mailed questionnaires, along with postage-paid return envelopes and a letter of instruction reminding partners to complete forms independently. This procedure was used at all follow-up procedures except at Time 5. At the Time 5 assessment, couples completed questionnaires at home and brought them to the laboratory where they engaged in a variety of tasks beyond the scope of the present study. After completing each phase, couples were mailed a check for participating (Study 1 = $40, Study 2 = $40–$50).

Measures

Marital satisfaction

Marital satisfaction was assessed eight times over the 4 years of each study, once every 6 months. To ensure that results were not idiosyncratic to a specific instrument, two measures of satisfaction were used. The first was the Quality of Marriage Index (QMI; Norton, 1983), a six-item scale asking spouses to report the extent to which they agree or disagree with general statements about their marriage (e.g., “We have a good marriage”). Five items ask spouses to respond according to a 7-point scale, whereas one item asks spouses to respond according to a 10-point scale, yielding scores from 6 to 45. High scores reflected greater satisfaction with the relationship.

The second measure of satisfaction was a version of the Semantic Differential (SMD; Osgood, Suci, & Tannenbaum, 1957) that asked spouses to rate their perceptions of their relationship on 7-point scales between 15 pairs of opposing adjectives (e.g., bad–good, dissatisfied–satisfied). The SMD yields scores from 15 to 105 such that higher scores reflect greater satisfaction with the relationship. For both samples, and for both measures, coefficient alpha was > .90 for husbands and for wives across all phases of the study. The two measures were highly correlated (mean correlation = .89 across all phases of the study for husbands and wives).

Marital forecasting

At the initial assessment (6 months into marriage) and again at the final assessment (4 years into marriage), spouses were asked to predict how their satisfaction would change over the following four years. Spouses responded to the question “Over the next four years, do you expect that your overall feelings about your marriage will become…” on a five-point scale, with 1 = Much worse, 2 = A little worse, 3 = Stable, 4 = A little better, and 5 = Much better.

Observed negative communication

To assess marital behavior independent of spouses’ perceptions, we videotaped spouses engaging in two 10-minute discussions about an area of difficulty in the relationship (e.g., dividing household chores). Each partner chose a topic for one of the interactions. Using the Verbal Tactics Coding Scheme (Sillars, Coletti, Parry, & Rogers, 1982), we coded each speaking turn as negative if it included sarcasm, criticism, blaming, avoiding responsibility, or hostile questions. A total proportion of negative behavior exhibited by each husband and each wife was computed for each conversation by dividing the number of codes for each spouse by the total number of speaking turns for that spouse in that conversation. Accordingly, scores on each conversation could range from 0.0, indicating that no speaking turns were negative, to 1.0, indicating that every speaking turn was negative. We calculated an index of negative behavior by averaging across both conversations for each couple. Due to technical difficulties, behavioral data were lost from 6 couples, leaving data from 245 of the 251 (98%) couples available for analysis.

To determine the reliability of our coding, a subset of the discussions (30% in Study 1 and 25% in Study 2) were randomly chosen to be coded by a second rater, and agreement between coders was assessed by calculating intraclass correlation coefficients (ICC) between the proportions of speaking turns coded as negative by each coder. Reliability was adequate in both studies (Study 1 ICC = .75; Study 2 ICC = .89).

Physical aggression

Physical aggression (e.g., threw something at spouse, pushed/ shoved spouse) in the past year was assessed using the eight-item Violence subscale of the Conflict Tactics Scale (Straus, 1979). Items were rated on 3-point scales (0 = never, 1 = once, and 2 = twice or more) and summed to create a total physical aggression variable.

Trait optimism

We assessed spouses’ trait optimism using eight items from the Life Orientation Test-Revised (Scheier, Carver, & Bridges, 1994; sample item: “In uncertain times, I usually expect the best”). Spouses were asked to rate the extent to which each item described them on a scale ranging from 0 (“I disagree a lot”) to 4 (“I agree a lot”), such that higher scores indicated greater optimism. Internal consistency was high for husbands and wives (coefficient alpha > .80).

Self-esteem

We assessed spouses’ self-esteem using the 10-item Rosenberg Self-Esteem Questionnaire (Rosenberg, 1965). Scores on the measure can range from 4 to 40, with higher scores indicating higher self-esteem (sample item: “I take a positive attitude toward myself”). Internal consistency was high for husbands and wives (coefficient alpha > .80).

External stress

We assessed external stress at the first assessment by having couples complete a version of the Life Experiences Survey (Sarason, Johnson, & Siegel, 1978), designed to assess life events in the previous 6 months. Sixty-five negative, stressful events were selected, with an emphasis on concrete events likely to occur in a young, married population. Events were grouped to represent nine domains: marriage, work, school, family and friends, finances, health, personal events, living conditions, and legal problems. For each event, spouses were asked to indicate whether the event occurred (0 = no, 1 = yes). To be included in the final composite score, however, the event could not represent a likely consequence of marital satisfaction or marital distress, excluding 14 items (e.g., sexual difficulties). Thus, the measure tapped only those stressors external to (i.e., unlikely to be caused by) the marriage. The final stress score, which could range from 0 to 51, was computed by adding together the number of events that the spouse reported had occurred.

Results

Descriptive statistics

Newlywed spouses reported very positive marital forecasts. Among husbands, 16% predicted that their overall feelings about their marriage would remain stable (n = 41), 39% thought their feelings would get a “little better” (n = 98), and 43% thought their feelings would get “much better” (n = 108). Few husbands predicted their feelings would get “much worse” (n = 2) or a “little worse” (n = 2). Among wives, 19% predicted that their feelings would remain stable (n = 47), 33% thought their feelings would get a “little better” (n = 82), and 48% thought their feelings would get “much better” (n = 121). Less than 1% thought their feelings would get “much worse” (n = 1). Given the qualitatively distinct nature of predicting that one’s feelings about the marriage would worsen and the small numbers of spouses who made such a prediction (thus precluding detailed comparisons with the more positive forecast types), we excluded these five couples from the remaining analyses, resulting in a total sample of 492 spouses from 246 marriages.

We examined marital satisfaction trajectories using growth curve analytic techniques and the HLM 7.0 computer program (Raudenbush, Bryk, & Congdon, 2010)¹. Husbands’ and wives’ data were estimated simultaneously within the same equations (as opposed to nesting spouses within couples) in order to allow for sex-specific intercepts, slopes, and random effects, thus providing greater overall flexibility in modeling (Atkins, 2005). Time was estimated as number of months since the couple’s wedding date and was uncentered so that the intercept terms (B_f00 and B_m00) could be interpreted as the initial value six months into marriage. We used the following equations:

$$\text{Level}1:Y_{\text{ti}}={(\mathit{\text{female}})}_{\text{ti}}[{\pi }_{\mathrm{f}0\mathrm{i}}+{\pi }_{\mathrm{f}1\mathrm{i}}{(\mathit{\text{Time}})}_{\text{ti}}]+{(\mathit{\text{male}})}_{\text{ti}}[{\pi }_{\mathrm{m}0\mathrm{i}}+{\pi }_{\mathrm{m}1\mathrm{i}}{(\mathit{\text{Time}})}_{\text{ti}}]+{\mathrm{e}}_{\text{tij}}$$

$$\text{Level}2:{\pi }_{\mathrm{f}0\mathrm{i}}(\text{wife intercept})={\beta }_{\mathrm{f}00}+{\mu }_{\mathrm{f}0\mathrm{i}}$$

$${\pi }_{\mathrm{f}1\mathrm{i}}(\text{wife slope})={\beta }_{\mathrm{f}10}+{\mu }_{\mathrm{f}1\mathrm{i}}$$

$${\pi }_{\mathrm{m}0\mathrm{i}}(\text{husband intercept})={\beta }_{\mathrm{m}00}+{\mu }_{\mathrm{m}0\mathrm{i}}$$

$${\pi }_{\mathrm{m}1\mathrm{i}}(\text{husband slope})={\beta }_{\mathrm{m}10}+{\mu }_{\mathrm{m}1\mathrm{i}}$$

These equations include separate intercepts and slopes for men and women, and sex-specific variance components at Level 2.

Results indicated a significant decline in marital satisfaction on average using the QMI and SMD (p < .001). Thus, despite the fact that no one predicted that their overall feelings toward the marriage would decline (and most predicted they would improve), their marital satisfaction did just that over the first four years of marriage.

Marital satisfaction trajectories by initial forecast

To test the association between forecasts and marital satisfaction, we examined possible differences in the marital satisfaction trajectories of spouses with relatively more favorable forecasts versus more moderate forecasts. Given the group sizes, we collapsed the “stable” and “little better” groups into a “moderate” group in order to have two relatively equally sized groups for comparison and maximize power. Thus, the husband groups included a moderate group (56%) compared to a positive group (44%), with similar groups for wives (moderate = 51%, positive = 49%).

To examine differences in the initial intercept and linear slope of satisfaction by forecast type, sex-specific forecasts were included at Level 2 in the equations described above as a predictor of intercepts and slopes (e.g., husbands’ forecasts predicted their own intercepts and slopes), and were coded such that the reference group (coded as 0) was moderate forecasts (stable/little better) and positive forecasts (much better) were coded as 1.

Results, shown in Table 2, indicated that wives with more positive forecasts exhibited steeper declines compared to wives with more moderate forecasts (p < .05) ². Results were similar for the QMI and the SMD. There were mixed findings when comparing the initial level of satisfaction for wives with more positive forecasts versus wives with moderate forecasts. The intercept by forecast interaction (differences in estimated initial levels of satisfaction) was not significant (Table 2). Actual initial levels (Table 3) revealed a nonsignificant finding for the QMI and a marginal (p < .10) finding for the SMD in which wives with more positive forecasts had lower initial satisfaction, suggesting a trend but not a conclusive effect. Husbands with different forecasts did not differ in their initial levels of satisfaction or in their change in marital satisfaction over time.

Table 2.

Summary of multilevel models comparing marital satisfaction trajectories among spouses with moderate versus positive forecasts (N = 246 couples)

Fixed effect	Coefficient (SE)	t ratio	df	Effect size r
Husbands
QMI
Intercept	41.74 (0.37)	---	---	---
Intercept × forecast	−0.40 (0.49)	−0.81	244	0.05
Linear slope	−0.51 (0.08)	−6.11***	244	0.36
Linear slope × forecast	0.06 (0.09)	0.66	244	0.04
SMD
Intercept	95.01 (0.79)	---	---	---
Intercept × forecast	−1.03 (1.14)	−0.91	244	0.06
Linear slope	−0.87 (0.16)	−5.39***	244	0.33
Linear slope × forecast	0.04 (0.20)	0.19	244	0.01
Wives
QMI
Intercept	42.09 (0.30)	---	---	---
Intercept × forecast	−0.22 (0.46)	−0.48	244	0.03
Linear slope	−0.39 (0.07)	−5.38***	244	0.33
Linear slope × forecast	−0.21 (0.10)	−2.10*	244	0.13
SMD
Intercept	97.03 (0.75)	---	---	---
Intercept × forecast	−0.55 (1.10)	−0.50	244	0.03
Linear slope	−0.76 (0.17)	−4.50***	244	0.28
Linear slope × forecast	−0.46 (0.23)	−2.02*	244	0.13

^*p < .05.

^**p < .01.

^***p < .001

Notes: All intercepts were significant p < .001 because the lowest possible score was higher than zero, so these statistics are not reported. Forecast was coded 0 for moderate forecast and 1 for positive forecast, so the interaction term represents the difference for individuals with moderate versus positive forecasts. Effect size r = sqrt [t²/(t² + df)].

Table 3.

Initial differences between spouses with moderate versus positive marital forecasts (N = 246 couples)

	Moderate Forecasts			Positive Forecasts
	N	M	SD	N	M	SD	t ratio	d
Husbands
QMI	138	41.72	4.72	107	42.24	3.87	−0.93	0.12
SMD	138	95.44	10.05	107	96.51	9.61	−0.84	0.11
Negative communication	134	0.07	0.05	106	−0.08	0.95	1.15	0.15
Physical aggression	136	0.31	1.04	108	0.18	1.45	0.83	0.10
Trait optimism	138	21.45	5.40	108	22.55	4.87	−1.65	0.21
Self-esteem	138	34.35	4.89	108	34.81	4.71	−0.76	0.10
Acute stress	138	4.28	3.41	108	4.81	3.38	−1.20	0.16
Wives
QMI	126	42.58	4.37	120	42.00	4.71	1.00	0.13
SMD	126	98.88	9.02	120	96.80	9.90	1.73+	0.22
Negative communication	123	−0.02	1.04	117	0.02	0.97	−0.32	0.04
Physical aggression	125	0.55	1.53	120	1.15	2.95	−1.98*	0.26
Trait optimism	126	22.26	5.47	120	22.16	5.53	0.15	0.02
Self-esteem	126	34.41	4.76	120	32.69	5.60	2.60*	0.33
Acute stress	126	4.48	3.71	120	6.09	4.30	−3.16**	0.40

⁺p < .10.

^*p < .05.

^**p < .01.

Notes: Initial differences in spouses’ own (versus partner) characteristics were examined. Negative communication behaviors were coded by two separate coding teams, thus resulting in significantly different means between samples. To ensure that this did not affect the results, we re-analyzed the data after first standardizing within sample and then collapsing across the two samples. Doing so did not change the pattern of results. We report the standardized means here.

We examined four-year divorce rates among wives with moderate versus positive forecasts to examine whether these results were due to differential rates of marital dissolution (e.g., the moderate forecast group may have experienced higher divorce, artificially inflating their marital trajectories). Seventeen percent of wives with moderate forecasts divorced compared to 12% of wives with positive forecasts; divorce rates did not differ significantly between groups, χ²(1, N = 246) = 1.65, p > .10. Further, results obtained with the QMI and SMD did not change when controlling for marital dissolution.

Initial differences by forecast type

To better understand the initial characteristics of wives who predicted moderate versus positive forecasts, we conducted a series of t tests. Results, shown in Table 3, indicated that wives making positive forecasts were characterized by significantly higher levels of physical aggression, lower self-esteem, and higher levels of external stress at the first assessment compared to wives with more moderate forecasts, with effect sizes in the small-to-moderate range. Controlling for these characteristics (entered simultaneously at Level 2) did not change the direction of the results but the effect of initial forecast became marginally significant (p = .06 and .07 for the QMI and SMD, respectively)³. No differences were found in observed negative communication or in trait optimism⁴ (p > .10). Husbands with moderate versus positive forecasts did not differ in any of these domains.

We also examined whether individuals with different forecasts differed in their rates of premarital cohabitation or their rates of becoming parents over the course of the study. Husbands with more moderate forecasts were significantly more likely to cohabit before marriage (72%) compared to husbands with more positive forecasts (58%), χ²(1, N = 246) = 4.84, p = .03, but wives with different forecasts did not differ in their rates of premarital cohabitation, χ²(1, N = 246) = 0.01, p > .10. No differences were found between forecast types in the rate of becoming parents over the course of the study for husbands or wives, χ²(1, N = 218) = 1.18 and χ²(1, N = 217) = 0.07, respectively, p > .10. These findings indicate that wives’ different marital trajectories by forecast type were not accounted for by demographic differences.

Forecasting at four years

Last, we examined spouses’ forecasting four years into marriage, when they were again asked how they thought their overall feelings about the marriage would evolve. For husbands (N = 135 at the final assessment) and wives (N = 137 at the final assessment), initial forecasting was a significant predictor of subsequent forecasting (Husbands: B = 0.80, Wald test = 4.46, p < .05; Wives: B = 1.11, Wald test = 7.44, p < .01). Moreover, despite the fact that wives who had initially made more positive forecasts experienced larger 4-year declines in satisfaction than those with moderate forecasts, they remained more likely than wives with moderate forecasts to predict that the following four years would get much better (40% versus 17%), χ²(1, N = 139) = 8.64, p < .01.

Discussion

Using data from 251 newlywed couples, this study examined how newlywed spouses’ predictions for how their marriages would change corresponded with observed change in marital satisfaction over the first four years of marriage. We first confirmed that newlyweds are optimistic regarding how their marriages would change. Although marital satisfaction has consistently been shown to decline on average over the newlywed years (Aron et al., 2002; Karney & Bradbury, 1997; Kurdek, 1998), practically all couples believed their overall feelings about their relationship would stay the same or improve over the next four years, with the most common prediction for husbands and for wives being that their overall feelings about their relationships would “get much better” over the following four years. Thus, much in the same way that couples routinely make positively interpretations of their present circumstances (e.g., Murray et al., 1996) and reconstruct the past in more positive ways (e.g., Karney & Frye, 2002), newlywed spouses also tend to view their futures in an optimistic way. This nearly-uniform tendency to believe that the marriage will be better than–or at least as good as– the present may help explain why many couples proceed to get married despite experiencing premarital doubts (e.g., Lavner, Karney & Bradbury, 2012): they likely believe their circumstances will improve.

Despite this strong tendency toward optimism, newlyweds’ initial forecasts did not map onto prospective reality: marital satisfaction declined on average. Moreover, wives who predicted the greatest increases in satisfaction actually had the greatest declines in satisfaction. As such, forecasts for the future appear to be highly biased (Kelley, 1983) and seem to be based more on what individuals want to happen than on what is actually likely (e.g., Epley & Dunning, 2000; Weinstein, 1980).

Although we cannot make causal claims on the basis of our correlational design, examining the initial characteristics of wives with different forecasts provided some insights into why these forecasts proved incorrect. There were inconclusive findings regarding whether wives with more positive forecasts had lower initial marital satisfaction, but wives who believed their relationships would get much better did report significantly lower self-esteem, more stressful life events, and higher levels of physical aggression toward their partners compared to wives with more moderate forecasts. This pattern of results indicates that these women were able to separate their more general negative characteristics from their predictions of what the future will hold (i.e., their forecast), but their ability to keep their negative attributes from affecting their evaluations of their marriage (i.e., their satisfaction) declines over time. Even so, this strategy may still prove adaptive in promoting relationship stability: these women were no more likely to divorce than women with more moderate forecasts and were more likely to continue making positive forecasts for their future relationship satisfaction four years later.

Before discussing the implications of these findings, we first outline several caveats. First, as with much of the research examining newlywed marriage, the sample as a whole was disproportionately Caucasian, middle-class, and well-educated, suggesting that they were relatively low-risk (cf. Karney et al., 1995). The relative percentages of spouses with negative, moderate, and positive forecasts may differ in higher-risk samples. Further research is needed to compare the marital trajectories and initial characteristics of individuals who believe their marriages will decline with individuals who believe their marriages will remain stable or improve. Second, we only included four-year follow-up data on satisfaction and dissolution, raising the possibility that additional differences (particularly with regard to dissolution) might have emerged had the couples been studied for a longer period of time. Third, although we drew on the vulnerability-stress-adaptation model to identify a range of initial risk factors, there were many other characteristics in each of these domains that we did not assess (e.g., neuroticism, attachment security, positive communication, chronic stress) that could also characterize individuals with different forecast types and account for differences in the degree of change in satisfaction over time. Fourth, our measure of forecasting only assessed how spouses believed their overall feelings about their relationships would change over time. Although this measure was ideal for comparing predicted trajectories and actual trajectories, we caution that there is a difference between being optimistic about how things will change (i.e., feelings becoming better) and being optimistic about the overall state itself (i.e., feelings being good). Further research is needed to compare spouses’ predictions about how satisfied they will be in their marriages in the years ahead with their actual satisfaction. Fifth, the finding that individuals with higher levels of risk tend to make more optimistic forecasts is tentative and needs to be interpreted with caution: it held only for women; the differences in initial risk were small-to-medium in size; and the higher risk did not fully explain the association between more optimistic forecasts and more negative marital satisfaction trajectories. We view this finding as an important first step into understanding the associations between risk, forward-looking cognitive biases, and marital satisfaction over time, but follow-up work is needed to more fully understand this phenomenon.

Notwithstanding these limitations, these findings add to a growing body of work suggesting that seemingly positive processes are not always beneficial for well-being (McNulty & Fincham, 2012). Although partners’ general levels of trait optimism did predict more positive marital trajectories, believing that one’s overall feelings in their marriage would get “much better” was not similarly predictive of later marital success; these results build on previous findings suggesting that global (i.e., trait optimism) and specific (i.e., marital forecasts) attributes may operate differently in predicting marital outcomes (Neff & Karney, 2002; 2005). The fact that newlywed spouses’ optimistic forecasts about how their marriages would unfold did not buffer them from declining satisfaction also highlights a conceptual distinction between positive cognitive processes focused on the here-and-now and those processes focused on the future. Like benign attributions for problematic behaviors (e.g., Bradbury & Fincham,1990), trait optimism is beneficial in shaping the meaning partners make of the present (e.g., helping them see conflicts as resolved, Srivastava et al., 2006; promoting cooperative problem-solving, Assad et al., 2007). In contrast, unrealistic optimism (believing the relationship will get better; expecting that the partner will rarely make mistakes, McNulty & Karney, 2004) is forward-looking, encompassing a set of expectations about what has yet to occur and reflecting, at least in part, the hope that perceived risks will not derail or undermine the relationship. In this manner, these forward-looking beliefs are less of a characteristic of healthy marriages (or healthy partners) and more of a barometer of one’s own risk for future distress.

The present findings also raise the possibility that this “risk barometer” may be calibrated differently for men and women. Husbands’ forecasts appear to be random affective forecasting errors, unrelated to how their marriages actually unfold or to their level of initial risk (as assessed here), suggesting a general process by which newlywed husbands have optimistic beliefs about the future of their relationships (i.e., predicting their feelings will remain stable or improve). Wives also report optimistic predictions regarding the future of their relationships, but the extent to which they did so was affected by their own level of risk, consistent with the view that more positive forecasts arise in the context of difficult circumstances (Martz et al., 1998). Women’s greater attunement to threat signals (e.g., Olff, Langeland, Draijer, & Gersons, 2007) and to dyadic relationships (e.g., Baumeister & Sommer, 1997) may render them more sensitive to threats to their romantic relationships and, in turn, more likely than men to regulate this risk through cognitive processes that reduce their perception of risk going forward (i.e., believing their relationships will improve). Future research should test this idea.

Further research is also needed to understand the processes by which satisfaction declines among women with more optimistic forecasts and the consequences of this decline. Were wives with optimistic beliefs especially unprepared for the emergence of additional difficulties and stressors later on (e.g., Kelley, 1983), and less likely to actively attempt to resolve these difficulties because they believed things would get better? Four years later, do these women recognize and express disappointment that their marriages were less satisfying than they originally hoped they would be? If so, do they conceptualize this disappointment as resulting from forces outside their control or as the fault of their husbands (i.e., outcome- versus person-related disappointment; see van Dijk & Zeelenberg, 2002)? These different beliefs would likely be reflected in different behavioral responses as well (e.g., trying harder in order to achieve the desired outcome versus disapproving of the other person and wanting to escape) and may have different clinical implications. More work is needed on the conceptual overlap between forecasting and these disappointment-related constructs and how they might interact to affect relationship satisfaction over time.

More generally, these findings may be important in helping to understand low rates of premarital counseling utilization, particularly among high-risk couples (e.g., Sullivan & Bradbury, 1997). In the Building Strong Families (BSF) Project, for example, a recent large-scale federal initiative to provide premarital interventions to unmarried low-income couples, only 55% of couples assigned to the intervention condition attended at least one session, despite the program including financial incentives and a variety of supports such as childcare, transportation, and meals (Dion, Avellar, & Clary, 2010). Models of health behavior argue that people are likely to engage in preventive behaviors to the extent that they perceive they are susceptible to the problem (e.g., Sheeran & Abraham, 1996; Strecher, Champion, & Rosenstock, 1997), which in the context of premarital counseling means that couples must believe that marital problems could happen to them (Sullivan, Pasch, Cornelius, & Cirigliano, 2004). Nonetheless, the results reported here indicate the opposite trend—nearly all couples overestimate the durability of their existing satisfied feelings. Moreover, this tendency persists over time, particularly among the riskiest partners: unrealistically optimistic beliefs remained relatively stable for many high-risk wives who, despite undergoing the steepest declines in marital satisfaction over the first four years of marriage, continued to believe the following four years would get much better. Accordingly, practitioners working with these families may first need to help them evaluate their circumstances in a more realistic manner in order to motivate them to change, and then help them explore, critically assess, and adjust their expectations for how their relationships will unfold.

In conclusion, the data reported here demonstrate that believing one’s marriage will improve does not make it so, and that for women in particular, very optimistic forecasts may paradoxically mark risky relationships. These findings indicate that positively-biased views do not uniformly promote increased marital satisfaction over time and suggest that newlywed couples routinely overestimate their likelihood of marital success. Further study of partners’ predictions about the future of their relationships and the psychological processes that allow partners to commit to and stay in risky relationships is warranted.