Retirees are likely to experience both increases and decreases in spending levels, i.e., volatility or fluctuations, during their retirement years. By planning for and being prepared to adjust to such volatility in spending, retirees can increase their odds of success in retirement.

As the defined contribution retirement savings system matures and emphasis slowly shifts from the accumulation phase to the spending phase, the next big problem for the retirement industry to solve is retirement income. To design successful retirement income solutions, providers need a better understanding of spending patterns during retirement.

While data have shown that spending generally decreases in retirement, the reality is that many retirees experience meaningful ups and downs in their spending over time versus a continuous decline. These fluctuations in spending have significant implications for potential income solutions; particularly, in determining factors such as the optimal liquidity and accessibility characteristics, the level of equity exposure, etc.

For this study, we analyzed data from the Health and Retirement Study (HRS) and its supplement, the Consumption and Activities Mail Survey (CAMS).¹ We found that, on average, annual household spending declined by about 2% during retirement.² But this decrease is not uniform for all retirees.

As outlined in Figure 1, our research found that spending can fluctuate greatly in retirement. On average, about 1 in 4 retirees experienced at least a 17%–20% increase in annual spending over a two‑year period, while another 1 in 4 experienced at least a 20%–21% decrease in annual spending over a similar period. Moran, et al., reported similar levels of spending volatility across various groups during the transition into retirement.³ Farrell and Greig also show that older families can experience higher expense volatility.⁴

Based on our analysis, there is a considerable risk of experiencing large increases in spending at some point in retirement (Figure 2). Notably, 1 in 2 retirees (50.1%) experienced a spending increase of 0%–25% between ages 65 and 90. Further, over 1 in 4 (28%) households experienced a 25%–50% spending increase, and over 1 in 5 (21.5%) households experienced spending increases between 50% and100% during retirement.

Given the range of possible variations in spending increases, the amount of liquid assets retirees should hold in their portfolios to address any potential shortfall will vary. Generally, it will depend on personal factors such as income, expected expenses, health status, family situation, risk preference, etc.

Does Spending Volatility Change?

Although spending levels were highly correlated to investable assets or household income, spending volatility was not. While average spending increased as investable assets increased (Figure 3), the average change in household spending (measured in absolute value) did not fluctuate based on investable assets (Figure 4). The same held true for household income levels.

This means that retirees with any level of investable assets or income can face a high level of spending volatility, and they need to plan and prepare accordingly.

Are Spending Increases Temporary?

While an isolated one‑time spending increase may be absorbed with minimal impact, spending upticks that persist for a longer period can cause greater concern and may warrant additional planning. Our analysis showed that a significant number of retirees experienced sizable, long‑lasting spending increases (Figure 5). For instance, 15% of households that experienced spending increases of 25% or higher were still spending at the same elevated level (or even higher) after four years.

"...a significant number of retirees experienced long‑lasting spending increases of high magnitude."

If substantially increased spending levels persist, retirees may need to reevaluate their investment portfolio and adjust their withdrawal strategy. For example, if members of a retired household annuitize part of their assets to cover ongoing expenses and invest the rest in long‑term securities, they might need to liquidate those securities prematurely if a spending increase persists.

For some retirees, lifestyle changes could also be necessary to meet new spending needs and minimize the risk of depleting their nest egg.

Causes of Spending Volatility

Generally, some level of spending volatility is expected in retirement and may even be desirable. For example, if the variability is due to discretionary spending decisions—such as taking a long‑planned trip, making a dream purchase, or donating to a favorite charity—the spending increase would be less of a liquidity concern, although it could still have other important implications, as we will discuss later.

On the other hand, if volatility arises from unplanned increases in nondiscretionary or essential spending, then it could become a true liquidity event and a cause of concern.

In our study, a larger share of the variation in total spending for retirees was due to changes in nondiscretionary or essential spending (Figure 6). Overall, categories such as home and home‑related expenses accounted for the largest share of the variation, distantly followed by health‑related expenses and transportation.

A more interesting story emerged as we examined spending fluctuations across different income groups. While for most retirees—those with annual incomes of less than $150,000—overall spending volatility was largely due to changes in nondiscretionary spending, for retirees with income levels above $150,000, the lion’s share of volatility was due to changes in discretionary spending (Figure 7).

Increased spending on nondiscretionary expenses might require immediate cash. If retirees have insufficient liquid assets to address these needs, they may be forced to take untimely distributions from their longer‑term investment portfolios. This, in turn, could lower  their chances of enjoying a successful retirement.

As illustrated in Figure 7, the share of spending volatility attributable to changes in discretionary spending increased with income. This implies that the ability to generate higher incomes in retirement could translate into higher discretionary spending for these retirees—a desirable outcome.

Discretionary Spending Improves Financial Satisfaction

In 2019, CAMS asked respondents how their level of satisfaction with their financial situation had changed in the past six years. We estimated how retiree spending on discretionary and nondiscretionary items had changed between 2013 and 2019 and how those changes correlated with the reported change in financial satisfaction (Figure 8). We used simple linear probability to test the strength of this correlation.⁵

We found minimal correlation between nondiscretionary spending changes and financial satisfaction. Those who reported that they were “much less satisfied” and those who reported that they were “much more satisfied” both experienced a drop of 5%–7% in nondiscretionary spending between 2013 and 2019.

However, increased satisfaction was associated with higher discretionary spending. Those who were “much less satisfied” in 2019 reported a 32% drop in their discretionary spending, while those who were “much more satisfied” had a 7.4% increase. It can, therefore, be argued that increased discretionary spending is beneficial.

Spending Volatility and Retirement Income Solutions

Our findings show that retired households experienced variability in their spending, and many adjusted to these fluctuations. Therefore, non‑guaranteed retirement income products that can accommodate some level of volatility might be well aligned with retirees’ behavior if they fulfill other retirement income and savings goals.

Our research also shows that many retirees experience “liquidity events”—i.e., sudden large increases in spending—and need to have enough easily accessible liquid assets to meet these needs. In recent years, discussions about how to generate a steady cash flow from retirement savings have taken center stage, but in our view, not enough attention has been paid to the liquidity problem. To the extent that there are trade‑offs involved in choosing between seeking a guaranteed income stream and preserving liquidity, we need a better understanding of liquidity needs.

"If retirees have insufficient liquid assets...they may be forced to take untimely distributions from their longer‑term investment portfolios."

Our findings draw attention to the personal nature of spending. Based on our research, while overall spending volatility didn’t change across income levels, the source of that volatility tended to shift from nondiscretionary spending to discretionary spending as income increased. Because changes in discretionary spending were correlated with changes in financial satisfaction, it can be argued that volatility arising from increased discretionary spending is beneficial.

"...the source of volatility tended to shift from nondiscretionary spending to discretionary spending as income increased."

To this end, managing spending volatility could be a two‑pronged issue. On one hand, we should try to mitigate the effects of volatility arising from nondiscretionary spending, but on the other hand, we should also aim to generate higher income and/or investment returns, which can help boost discretionary spending. This might boil down to personal asset allocation strategies that can accommodate different levels of assets (or income requirements) and varying risk tolerance.

Appendix A

Change in spending across waves can be exaggerated in the presence of reporting or measurement error (the survey follows a group of households from 2005 to 2019 that were present in the 2005 CAMS, were at least 65 years old, and have been surveyed in at least three consecutive waves of the CAMS). The most recent wave was released in August 2021. We took several steps to mitigate the effect of measurement error in the analysis. First, we dropped all observations (household wave) that were in the top and bottom 1% of the spending distribution. Second, some spending categories were split in the early waves of the CAMS. The CAMS recorded 39 spending categories consistently between 2005 and 2019. So, we restrict our data between 2005 and 2019. Third, we restrict our sample to households that have been observed for at least three consecutive waves of the CAMS. Finally, we excluded observations with inter‑wave change in total household spending exceeding 150% (in absolute terms).

Appendix B

To estimate how much of the variation in total spending is explained by different components of spending (as shown in Figures 6 and 7), we used panel regressions with household fixed effects. In particular, the reported R2s are within the household variation across time explained by each component of spending. For each regression, the only explanatory variable used is the component of spending for which the R2s are reported. In other words, no other covariates such as demographic controls were used as the primary interest is only on the percentage of overall variation explained by each component of spending rather than any point estimates.

Appendix C

To estimate the correlation between change in financial satisfaction and change in different types of spending, first we created a binary variable. Those who reported that they were either “a little less satisfied” or “much less satisfied” were grouped as “unsatisfied.” Everyone else was put in the other (base) category. This binary variable was used as the outcome variable in a linear probability model to estimate the correlation between change in satisfaction and change in types of spending (discretionary and nondiscretionary). Additional demographic controls used were gender, race, age, change in marital status, and change in financial wealth between 2012 and 2018.

Actual outcomes may differ materially from estimates and probabilities provided in this Insights. Altering data inputs of the analysis shown in this Insights could yield different results.