Life Satisfaction Across the Lifespan: Findings from Two Nationally Representative Panel Studies

Because there are multiple ways that a person could evaluate his or her life, there are multiple variables that could be assessed to capture this overall evaluation. For instance, Diener et al. (1999) suggested that to gain a complete picture of a person’s well-being, researchers should assess the related but separable dimensions of positive affect, negative affect, and life satisfaction. Positive and negative affect are emotional dimensions that capture the affective feelings that people experience as they live their lives. Life satisfaction is a cognitive judgment that taps an individual’s reflective judgment about the conditions in his or her life. In the current paper, we use these terms (positive affect, negative affect, and life satisfaction) when referring to research that has assessed one of these more specific variables. We use the terms “happiness” or “subjective well-being” when discussing general findings or broad theoretical models that might reasonably apply in similar ways to the more distinct constructs. We should note, however, that the empirical research we present in these studies focuses specifically on life satisfaction, rather than on the affective dimensions.

Participants who were 88 or 89 reported slightly higher levels of satisfaction. However, the confidence intervals for each data point are quite large among the oldest groups, and this is the only age group that did not follow the general trend of lower life satisfaction among the very old.

Because there is an overall level-2 intercept in this model, there are only 20 dummy variables for the 21 calendar years and 20 dummy variables for the 21 waves. This contrasts with the instrumentation effect analyses from the previous model where there was no overall intercept. In this earlier model there were 38 dummy variables for the 38 age groups because we estimated separate intercepts and slopes for each group. Centered age was divided by 10 before calculating the higher-order powers.

Because life satisfaction was only added to the study in the fifth wave, we needed to decide whether to count the first wave of data collection or the first wave of life satisfaction data as Wave 0. To ensure that this variable had a similar meaning in all sub-samples, we decided upon the latter. However, results are the same regardless of the coding that is used.

We did run these analyses, and as we suspected, the wave and calendar-year effects were very small. More importantly, the estimated age trajectory did not change once these effects were controlled.