New evidence of strong but slow portfolio responses to equity premium changes

Standard portfolio theory predicts strong investor responses to changes in the equity premium, but empirical evidence has found much lower sensitivity. This column studies the effects of changes in the Norwegian wealth tax, which generated variation in the after-tax equity premium. It finds that investors adjusted strongly, but slowly, with adjustment continuing for several years. This is consistent with the presence of adjustment frictions, such as attention or information costs. The findings can also help reconcile household portfolio behaviour with standard levels of risk aversion.

How strongly investors respond to changes in expected returns is a fundamental question in financial economics. Standard portfolio theory predicts that when the expected return on equities relative to safe assets – the equity premium – rises, investors should increase their exposure to stocks, with the size of the response increasing with their willingness to bear risk. This response is central not only to theories of portfolio choice and asset prices, but also to current policy debates over the taxation of capital and wealth, where tax systems can substantially alter the relative after-tax returns on different assets (Bastani and Waldenström 2024).

Empirically, however, investors appear surprisingly unresponsive to changes in expected returns. Survey evidence finds only modest relationships between measures of investors’ expected stock returns and their portfolio allocations (Giglio et al. 2019). This could imply that investors are implausibly risk averse relative to conventional estimates. Alternatively, the small measured response may be because investors respond gradually. A slow response in turn may be due to information, attention, or other adjustment costs which prevent investors from immediately moving to their desired portfolios. Recent evidence on retirement portfolios points to an important role for such frictions (Choukhmane and de Silva 2024).

In Fagereng et al. (2023), we distinguish between these explanations by exploiting changes in the Norwegian wealth tax that generated exogenous variation in the after-tax equity premium. During the 1990s, preferential wealth-tax treatment of equities was first introduced and years later removed, providing changes in the equity premium in both directions. Furthermore, because households below a given wealth threshold are exempt from the wealth tax, the setting provides a natural treatment group of investors affected by the reform and a control group of investors who were exempt from the wealth tax at the time of the reform. Using a difference-in-differences approach, we find that investors do respond to changes in the equity premium, but their response is gradual. Short-run responses are modest, in line with previous evidence, but portfolio adjustment continues for several years. In the long run, the response is consistent with standard portfolio theory and moderate levels of risk aversion.

A tax-induced change in the equity premium

Identifying how investors respond to changes in the equity premium is difficult. Expected returns are not directly observable, and changes in expected returns typically affect all investors at the same time. The Norwegian wealth tax provides an unusual natural experiment that helps overcome both problems.

Norway taxes households on their net wealth above a threshold. From 1992, publicly listed equities received preferential treatment: for wealth-tax purposes, stocks were valued at only 75% of their market value, while safe financial assets such as bank deposits were valued at their full market value. For households subject to the wealth tax, this raised the after-tax return on equities relative to safe assets. The size of this tax-induced equity premium depended on the household’s marginal wealth-tax rate, while households below the wealth-tax threshold were unaffected.

Figure 1 Wealth taxation of stocks and the tax-induced equity premium

Note: See Fagereng et al. (2023) for details on the wealth tax reforms and the construction of the tax-induced equity premium.

The preferential treatment was introduced in 1992 and abolished in 1998. The reforms therefore provide changes in the equity premium in opposite directions: an increase in 1992 and a decrease in 1998. Importantly, the taxation of capital income was unchanged, so the reforms altered the relative after-tax return on stocks and safe assets without changing their pre-tax returns.

We combine these reforms with administrative data covering the financial portfolios of the Norwegian population. The data allow us to compare households exposed to different changes in the tax-induced equity premium and, crucially, to follow their portfolio adjustments over several years.

Investors respond strongly – but slowly

Figure 2 shows how investors adjusted their portfolios following the changes in the tax-induced equity premium. The response is gradual. In the years after the introduction of preferential tax treatment in 1992, households affected by the reform steadily increased the share of their financial wealth invested in stocks compared to control households. Between 1993 and 1997, the difference grew by about two percentage points. When the preferential treatment was removed in 1998, the process went into reverse: affected households gradually reduced their stock-market exposure.

Figure 2 The effect of changes in the equity premium on portfolio allocation

Note: See Fagereng et al. (2023) for details on the empirical strategy, estimation, and additional results.

The 1998 reform reduced the after-tax equity premium of affected households by about 30 basis points. Within two years, affected households reduced the share of their financial wealth invested in stocks by a modest 0.5 percentage points relative to control households. This is only about 24% of the eventual response. The rest builds slowly. After five to six years, the share has dropped by about two percentage points, bringing it back to the level prevailing in 1993. An analysis based only on the first year or two would therefore conclude that investors respond weakly to changes in expected returns – a wrong conclusion.

Over a longer horizon, the conclusion is very different. The cumulative response is about four times larger than the short-run response. When interpreted through a standard portfolio model, the long-run responses imply coefficients of relative risk aversion between 1.8 and 2.8. These are moderate values, in line with those conventionally used in portfolio models.

The apparently weak sensitivity of portfolios to expected returns therefore need not imply that investors are extremely risk averse. Instead, investors appear to move slowly towards their desired portfolios because of frictions in adjusting their portfolios. This also helps reconcile our findings with previous evidence: over short horizons, our estimates resemble the modest portfolio sensitivities documented using survey measures of expected returns. It is only by following investors for several years that the much larger response becomes visible.

How large are the frictions needed for portfolios to adjust slowly?

In a frictionless portfolio model, investors should adjust immediately when expected returns change. Our evidence instead points to the presence of adjustment frictions. These need not be large. For the average household in our sample, we estimate that the gain from making the full portfolio adjustment immediately, rather than spreading it over the following five years, is only about $3. Even small costs of paying attention, gathering information, or rebalancing a portfolio may therefore be sufficient to generate substantial inertia.

The speed of adjustment also varies systematically across investors. Households that, based on observable characteristics, are expected to face lower adjustment costs respond considerably faster. Importantly, these faster-adjusting investors also own a disproportionate share of the stock market. Aggregate asset demand therefore depends not only on how quickly the average household adjusts, but also on which investors hold the stocks.

Our results do not identify the precise source of these frictions. Investors may pay attention to their portfolios only occasionally, face costs of acquiring and processing information, procrastinate, or prefer to rebalance gradually through new savings rather than actively trading existing assets.

Implications for asset prices and policy

A growing asset-pricing literature argues that asset demand is far less elastic than conventional models assume, so that relatively small shifts in demand can move prices substantially (Gabaix and Koijen 2021). Slow portfolio adjustment is a household-level mechanism for exactly this: if investors are slow to move capital towards assets offering higher expected returns, their demand is inelastic in the short run.

Our evidence provides a microeconomic perspective on this mechanism. In the short run, investors respond little to a change in the equity premium, making their demand for equities relatively inelastic. Over longer horizons, however, capital gradually moves and demand becomes considerably more responsive. The elasticity of asset demand may therefore depend crucially on the horizon over which it is measured.

There are also implications for policies that alter relative returns across assets. If policymakers want to encourage households to hold particular types of assets, a small initial response need not mean that the policy has little behavioural effect. Portfolio responses may take several years to materialise, so estimates based on short post-reform periods can substantially understate long-run responses.

Finally, changes in the equity premium affect not only how much stock existing investors hold, but also whether households participate in the stock market at all. We find that the increase in the equity premium induced substantial entry into the stock market. Yet when the incentive was removed, these new investors largely remained. This asymmetry points to an important role for one-time entry costs: once households have overcome the cost of entering the stock market, a reversal of the initial incentive need not induce them to leave. Our estimates imply a one-time entry cost of about $800 and a recurring participation cost of about $89, in line with existing estimates of participation costs.

Taken together, our results suggest that the apparently weak response of investors to expected returns can be misleading. Investors do respond substantially to changes in the equity premium – they just take time to do so. Accounting for this slow adjustment can reconcile household portfolio behaviour with standard levels of risk aversion and may help explain why asset demand appears so inelastic in the short run.

Source : VOXeu