That's the indomitable conclusion quoted from the summary of a 2013 PNAS paper. In a bit more detail:
We identified several important age-related patterns in decision making under uncertainty: First, we found that healthy elders between the ages of 65 and 90 were strikingly inconsistent in their choices compared with younger subjects. Just as elders show profound declines in cognitive function, they also show profound declines in choice rationality compared with their younger peers.
And it's all based on (laboratory) games coming directly from the pages of game theory:
We found that subjects do make dominated choices in 11.7% of cases. However, most importantly, we found that age plays a significant explanatory role (Table S1). Older adults violate dominance disturbingly often, 24.9% of the time; next are adolescents at 10.1%, followed by young and midlife adults, for whom such violations are rare (5.2% and 5.4%; Fig. 2B). It is important to note that these statistics are not driven by a minority of subjects. Overall, 57% of our participants violated dominance at least once in gain trials and 75.6% violated it at least once in loss trials. Importantly, all but one of our subjects over 65 y of age violated dominance at least once in the loss domain. Moreover, the number of dominance violations increased as a function of age within the midlife (regression coefficient = 0.228, P < 0.001) and older adult groups (regression coefficient = 0.399, P = 0.003). Startlingly, our older adults lost 46.1% of expected possible earnings in these trials, which is significantly more than young adults (9.4%), midlife adults (9.6%), or adolescents (19.3%; t test, two-sided P < 0.001 for all comparisons).
As for such lab-based studies, the sample was fairly small, around 135 people from which about 36 were older than 65; all of the latter group were screened for dementia (with negative results). Also they say they've controlled for IQ, numeracy, and financial status, so they "demonstrate that the age effects we document are not driven by other differences between groups".
The authors also cite a number of prior works as confirming their general finding:
The population above 65 y of age continues to grow (10), and a growing literature indicates that older adults make decisions detrimental to their wealth, health, and general well-being. Elders borrow at higher interest rates, use credit balance transfers suboptimally, misestimate property value, and pay more fees to financial institutions (11). Most older adults even fail to choose health plans correctly (12). Older adults are also more likely to make errors when voting (13).
So, it this study representative of the research in this area, particularly respect to its harsh conclusion? And since Schmuddi asked me to detail further why I'm skeptical about this... the finding of this one study is that the elders are horribly inept on a certain task... As in all science, particular social science, one needs to ask if this finding is replicated. Also the implication (given the related work mentioned) is that a lot of elders' real-life decision errors/problems can explained by sub-optimality (dominated choice) in a fairly narrow task, as it's basically taken to imply "profound declines in choice rationality"... which given the large percentage involved... is about as good as saying that a basic tenet of economics (choice rationality) is only half-right for elders. These are pretty strong findings. (Just imagine what it means in the context of electing a president older than 65.) It's not unfathomable that there could be other studies, using related but different tests that moderate this conclusion. (And after bit more searching, the paper is famous enough, it was the subject of a TIME article when it came out... and the TIME article is less cautions about positing explanatory power from the paper's findings to real-life scenarios.)