skip to content

Cambridge Endowment for Research in Finance (CERF)


Voting and Trading: The Shareholder’s Dilemma

Adam Meirowitz (University of Utah) and Shaoting Pi, (Research Associate, Cambridge Centre for Finance (CCFin) and Cambridge Endowment for Research in Finance (CERF))

December 2021


In the study of enterprise ownership comparisons between ownership by investors of capital, suppliers of labor, or customers is widespread (e.g., Hansmann 1988; Jensen and Meckling 1979; Fama and Jensen 1983). Little attention, however, has been allocated to the connections between governance by owners and the ease with which ownership status can be changed. In our paper “Voting and Trading: The Shareholder’s Dilemma,” we explore this point in the context of publicly traded firms by studying how market liquidity can impact the ability of shareholders to effectively govern. Several scholars have studied the direct effects of selling on governance (e.g., Admati and Pfleiderer, 2009; Edmans, 2009; Edmans and Manso, 2011). Our focus is on how opportunities to trade shares impact the incentives for voting and governance by shareholders; thus, we focus on indirect or equilibrium connections.

The informal rationale for governance by shareholder voting is seemingly compelling. Shareholders are united by the concern for return on their investment in the firm they govern, and thus they will have strong incentives to enact policies that enhance the firms’ value. Reflexively, we might then expect that this leads to a problem of common values and although economic theory does provide reasons to be cautious, problem of common values tends to admit institutions that can aggregate information effectively. We contend, however, that this assessment is premature.

This optimistic conception ignores the fact that shareholders may also trade shares in a market, and prices may depend on information revealed from governance decisions. The presence of a market turns out to be particularly relevant. Shareholders may have perverse incentives when voting over firm policy in the presence of liquidity. The key feature in our account is that current shareholders can decrease or increase their holdings in the firm. The desirability of either of these actions depends on the price at which they will transact. Thus, shareholders will not only care about influencing firm decisions, but they will also care about influencing market share prices. If there is any information conveyed by voting, then market prices must also react to votes. This then creates the possibility that strategic voting will allow for the creation of informational advantages over the market, which can be translated into informational rents by strategic voting and trading.

To flesh out the incentives faced by shareholders in the presence of liquidity, we develop a simple model of voting and trading. We find that voting for the policy option that a shareholder believes to be optimal for the firm is only optimal for the shareholder if she turns out to be the pivotal voter (that is the vote is nearly a tie). In all other realizations of the votes by others, her payoffs are maximized by casting a vote for the option she believes to be worse for the firm and capitalizing on the informational advantage she has over the market. Equilibrium then requires that voting must be sufficiently noisy so as to balance (i) how the market reacts to votes with (ii) the incentive to select the policy that is seemingly best for the firm. Thus, equilibrium will be consistent with the idea that there is not too much information contained in voting even if shareholders in aggregate possess a lot of relevant information. In the limit, as the number of shareholders gets large, the probability of making the decision that is better for the firm is bounded away from 1; information aggregation fails. In fact, we find that the limiting probability of making the correct decision is lower than the probability that a single agent with one signal would make the correct decision.

One way to think about these equilibria is through a counter factual. If voting were more correlated with information, then incentives to take advantage of informational rents would be too strong, and there would be dominant incentives to vote against one’s information. Rather, in equilibrium, these incentives are either not present or just balanced with incentives to correctly influence policy, and voting is seemingly random not particularly informative. It is instructive to draw an analogy with the absence of arbitrage opportunities in equilibrium to a canonical trading model. Here, incentives to influence market prices through voting can be driving equilibrium behavior, even though shareholders don’t see desirable opportunities to manipulate prices by voting in equilibrium. If these features were not balanced in equilibrium, a shareholder would see opportunities to deflate (inflate) prices prior to purchasing (selling) shares through casting votes that are seen as strong votes of negative (positive) signal. Interestingly, the model also has equilibria in which voters are very lopsided and completely uninformative. These equilibria may match what we sometimes see in practice when shareholders are seen as rubber-stamps even if they possess strong information.

We contribute to how we conceive of corporate governance. The takeaway is that there is room for scholars to rethink how institutional features impact the motives of shareholders when making firm decisions. Opportunities to behave strategically in one domain may have spill-over effects and distort behavior in another domain. Thus, there are analytical insights that can be obtained by rethinking a range of questions about governance with a broader eye on available levels of participation. Our primary results are negative in that they point to failures in information aggregation and therefore suggest that shareholders may not accurately steer firm policy or serve as effective controls on management misbehavior. But more broadly, thinking about the connections between trading and voting makes it possible to better understand how a broad range of institutional features might impact voting and governance in ways that previous work has missed.

We also contribute to a deeper understanding of empirical patterns relating the presence of votes to stock volatility and price. Here there are two points to make. The first point relates to a tradition of modeling how public signals or events impact trading. Empirical work documents a pattern of high volume without high price effects has led to some to fit richer models of information processing to marker behavior after important events (e.g., Kim and Verrecchia, 1991, 1994; Kandel and Pearson, 1995; Bollerslev, Li and Xue, 2018). In the case of shareholder votes our model provides an alternative and possibly simpler explanation. The public event of learning how shareholders vote is not very informative, but shareholders all possess private information after the vote and have incentives to trade on this information. In fact, the prediction of our model is that a vote will provide almost no information and so price effects will be minimal, but every shareholder will buy or sell shares after voting and in expectation the number buying will be almost the same as the number selling; thus, even after the spike in trading volume following a vote, we may see minimal price effects.





Admati, Anat R., and Paul Pfleiderer. "The “Wall Street Walk” and shareholder activism: Exit as a form of voice." The Review of Financial Studies 22.7 (2009): 2645-2685.


Bollerslev, Tim, Jia Li, and Yuan Xue. "Volume, volatility, and public news announcements." The Review of Economic Studies 85.4 (2018): 2005-2041.


Edmans, Alex. "Blockholder trading, market efficiency, and managerial myopia." The Journal of Finance 64.6 (2009): 2481-2513.


Edmans, Alex, and Gustavo Manso. "Governance through trading and intervention: A theory of multiple blockholders." The Review of Financial Studies 24.7 (2011): 2395-2428.


Fama, Eugene F., and Michael C. Jensen. "Separation of ownership and control." The journal of law and Economics 26.2 (1983): 301-325.


Hansmann, Henry. "Ownership of the Firm." Journal of Law, Economics, & Organization 4.2 (1988): 267-304.


Jensen, Michael C., and William H. Meckling. "Rights and production functions: An application to labor-managed firms and codetermination." Journal of business (1979): 469-506.


Meirowitz, Adam, and Shaoting Pi. "Voting and Trading: The Shareholder's Dilemma." Available at SSRN 3445233 (2021).


Kandel, Eugene, and Neil D. Pearson. "Differential interpretation of public signals and trade in speculative markets." Journal of Political Economy 103.4 (1995): 831-872.


Kim, Oliver, and Robert E. Verrecchia. "Trading volume and price reactions to public announcements." Journal of accounting research 29.2 (1991): 302-321.


Kim, Oliver, and Robert E. Verrecchia. "Market liquidity and volume around earnings announcements." Journal of accounting and economics 17.1-2 (1994): 41-67.




Climate change, discounting, and the tragedy of horizons

Andreas Charisiadis, CERF Research Assistant

November 2021

This year’s UN Climate Change Conference, COP26, has been in the centre of a controversial debate over the extent to which current climate change mitigation measures are (in)sufficient to limit the rise in average global temperatures to 2°C – let alone 1.5°C – above pre-industrial levels. Inherent in the discussion about the appropriate intensity of climate action is a fundamental social dilemma. While most of the costs associated with climate change mitigation measures are incurred today, their ensuing benefits are to a large extent enjoyed by future generations. Conversely, inaction today – while unlikely to generate material repercussions in the short run – is bound to have dramatic consequences in the more distant future. Climate change therefore poses an intergenerational challenge. Specifically, insufficient climate action today represents a manifestation of the ‘tragedy of horizons’, whereby the current generation avoids costly abatement policies to the detriment of future generations – a phenomenon Chichilnisky (1996) referred to as “dictatorship of the present”.

Exponential discounting and the welfare of future generations

Intertemporal choice problems are commonly analysed under the assumption of exponential discounting, whereby any future benefit or cost is discounted at a constant rate. Alas, relying on exponential discounting when conducting cost-benefit analyses for environmental action produces optimal policies which place almost negligible value on environmental damage materialising in the distant future, say several hundred years from now. At virtually any chosen discount rate, the future costs arising from climate change become unimportant in present-value terms so long as the horizon under consideration is sufficiently long. At constant discount rates, justifying large public expenditures on climate change mitigation therefore becomes an intricate endeavour. This gives rise to the question of how climate action – which is of sufficient scale to avoid disastrous consequences for future generations – can be motivated in an intertemporal optimisation framework.

Declining discount rates – a way forward?

One possible approach involves departing from exponential discounting at a constant rate and instead allowing for time-variation in the discount rate. For instance, hyperbolic (or quasi-hyperbolic) discount functions yield discount rates which decline with the length of the time horizon, i.e. impatience is larger in the short run than in the long run. In other words, benefits and costs arising in the distant future are discounted at lower rates than if they were to materialise sooner. The welfare of future generations thus receives a relatively larger weight under such discounting regimes. The use of declining discount rates might therefore be helpful in rationalising the need for more pronounced climate policy action today. Importantly, employing declining discount rates in environmental cost-benefit analyses need not rely on the assumption that all individuals have intertemporal preferences with declining impatience. In fact, even if all members of society are using a constant personal discount rate, the representative agent – who duplicates the optimal behaviour of the group of individuals – has a discount rate which decreases with the time horizon, so long as individuals exhibit decreasing absolute risk aversion (Gollier and Zeckhauser 2005). A similar conclusion is reached by Heal and Millner (2014), who analytically aggregate a set of heterogeneous discount rates into a single ‘consensus’ rate, wherein each preference is assigned an equal weight. This time-dependent aggregate rate is shown to decrease monotonically to the smallest discount rate in the population.

Achieving intergenerational equity

Equal treatment of current and future generations is at the heart of the debate around the appropriate choice of discounting regime when evaluating welfare across generations. Yet, the notion of equal treatment is far from unambiguous. For instance, Ramsey (1928) argued that discounting the utility of future generations would be “ethically indefensible” – yet, so long as there is positive economic growth and utility functions are concave (i.e. agents prefer to smooth their intertemporal consumption path), using a positive rate to discount future consumption streams (as opposed to utility) can be consistent with equal treatment of current and future generations.

However, not discounting the welfare of future generations at all – i.e. setting the social utility discount rate to zero as Ramsey (1928) suggested – is not uncontroversial. In fact, Arrow (1999) argued that failing to discount would produce logical fallacies. For illustration, he considered a stylised economy in which output takes the form of a continuous stream of perishable goods. In this economy, the current generation has access to an investment opportunity, whereby for each unit of consumption foregone, the investment yields a continuous consumption stream (say, α) in perpetuity. In the absence of discounting, any present value calculation would advocate foregoing consumption in order to invest a fraction arbitrarily close to 100% of the current generation’s income in the project, regardless of how small the value of α is. Arrow (1999) regarded such high sacrifices by the current (or, in fact, any) generation as morally unacceptable, and therefore concluded that the welfare of future generations should be discounted at a positive rate.

Reflecting the possibility of extinction

In a recent contribution, Chichilnisky et al. (2018) argue that – contrary to the findings of Arrow (1999) – refraining from discounting the welfare of future generations need not lead to logical inconsistencies. They derive that maximising the unweighted expected utility across all generations can, in fact, be logically coherent as long as (a) the expected total number of individuals who will ever live before extinction is finite and (b) each individual’s utility function is bounded above and below. In this framework, the only source of value decay stems from the possibility of extinction occurring at an uncertain future date. Consequently, the value placed on future consumption should reflect the fact that potential future generations may not come into existence. Chichilnisky et al. (2018) therefore argue that – when modelling the benefits and costs of climate policy – any discounting of future generations’ utility should purely reflect the risk of extinction. This results in an ‘extinction discounting rule’, whereby the weight placed on future generations declines in line with their decreasing probability of existence.

Overall, mitigating climate change is without doubt an intergenerational challenge. Standard (exponential) discounting regimes have been shown to be of limited usefulness for environmental cost-benefit analysis, given their negligible weight on the distant future. Different alternatives have been proposed in response to this dilemma, such as declining discount rates which place a relatively larger weight on future generations, or the assignment of equal weights to all potential future generations conditional on their existence. How such discounting regimes may be implemented in current and future policy making remains an open question.



Arrow, K.J., 1999. Discounting, Morality, and Gaming. In: Portney, P.R. and Weyant, J.P. (eds.), Discounting and Intergenerational Effects. Resources for the Future Press: Washington D.C., pp.13-21.

Chichilnisky, G., 1996. An axiomatic approach to sustainable development. Social Choice and Welfare, 13(2), pp.231-257.

Chichilnisky, G., Hammond, P.J. and Stern, N.H., 2018. Should we discount the welfare of future generations? Ramsey and Suppes versus Koopmans and Arrow. Warwick Economics Research Papers series, 1174.

Gollier, C. and Zeckhauser, R., 2005. Aggregation of heterogeneous time preferences. Journal of Political Economy, 113(4), pp.878-896.

Heal, G.M. and Millner, A., 2014. Agreeing to disagree on climate policy. Proceedings of the National Academy of Sciences, 111(10), pp.3695-3698.

Ramsey, F.P., 1928. A mathematical theory of saving. Economic Journal, 38(152), pp.543-559.


Shaoting Pi, CERF Research Associate

October 2021

Shareholder Collaboration on Corporate Governance


Activist shareholders, who exert a large influence on firms’ policies, typically only own small stakes and frequently interact with fellow shareholders of target firms. Recent empirical evidence (e.g., Crane, Koch, and Michenaud (2019), McCahery, Sautner, and Starks (2016), and Wong (2019)) also suggests that shareholder collaboration has been substituting for concentrated holdings on corporate governance and is orchestrated rather than spontaneously arising. However, it is largely unexplored how a lead activist shareholder organizes shareholder collaboration in the first place. In particular, how does the lead activist shareholder convince fellow active shareholders to collaborate on activism? How does the lead activist win over passive shareholders? Is a larger coalition including more fellow active shareholders more convincing to passive shareholders? What is the optimal size of the lead activist’s coalition?


To answer these questions, I present a model in which an activist with small stakes launches a shareholder campaign to force her proposal on the target firm. The model captures the heterogeneity of the activist’s fellow shareholders; there are two types of fellow shareholders, active and passive. On the one hand, active shareholders can actively engage in the activist’s campaign, such as by communicating or collaborating with the activist. On the other hand, passive shareholders do not actively join the campaign but can vote with their shares if a proxy fight occurs. Passive shareholders will vote against the activist unless convinced that the activist’s proposal has sufficiently high value.


My analysis underscores two problems that the small activist must simultaneously overcome. The first problem is how to afford the shareholder campaign. The second problem is how to win support from passive shareholders. I first analyze how the lead activist shareholder overcomes these two problems, respectively. Then, I analyze how the interplay between these two problems can endogenize the optimal size of the activist’s coalition with fellow active shareholders.


To mitigate the first problem, the lead activist shareholder seeks to communicate and collaborate with fellow active shareholders. The activist needs to figure out a persuasive communication strategy and design the optimal collaboration contract. I find that the activist, to persuade coalition members to collaborate, must reveal her information to them. As a result of the communication and collaboration with coalition members, the activist’s coalition is akin to a fictional large shareholder who collectively owns the shares of the activist and coalition members. Thus, building the coalition with fellow active shareholders enables the lead activist to afford shareholder activism costs more easily, mitigating the first problem.


We might expect that the activist would therefore enlarge the coalition as much as possible. However, perhaps surprisingly, I show that the activist must limit her collaboration with fellow active shareholders to win over passive shareholders. I first show that the activist’s messages are not persuasive to passive shareholders in equilibrium. The intuition is that passive shareholders do not take the activist’s messages at face value, while the lead activist has incentives to exaggerate the value of her proposal. So, when the activist launches the costly proxy fight, passive shareholders know that the activist’s payoff from implementing her proposal is higher than her expenditures on the proxy fight. In other words, passive shareholders’ expectation of the proposal’s value is conditional on the activist’s expenditures on the shareholder campaign. The higher expenditures the activist has to undertake, the higher expectation of the proposal’s value passive shareholders have. From passive shareholders’ perspective, if the activist is willing to undertake more costs from having a smaller coalition, then this can convincingly signal that the activist’s proposal has a high value. Consequently, to convince passive shareholders that the value of the proposal is sufficiently high, the activist needs to limit the size of her coalition.


Recall that the lead activist must solve the two problems at the same time. The analysis above reveals that building a coalition with fellow active shareholders mitigates the first problem (how to afford activism costs) but exacerbates the second problem (how to win over passive shareholders). Accordingly, the activist must strike a balance between enlarging the coalition to have more fellow active shareholders share more costs and shrinking the coalition to convince passive shareholders of the value of her proposal. Taking this quest as a starting point, I derive the optimal size of the activist’s coalition.


In equilibrium, because the activist can optimize her coalition-building to simultaneously overcome the two problems (how to afford the campaign and how to win the campaign), the activist’s coalition makes the proxy-fight threat more credible to the board in equilibrium. As a result, considering the higher risk of losing the proxy fight due to potential shareholder collaboration, boards are more likely to reach agreements with activists before proxy fight voting.


Overall, this paper finds that collaborating with fellow active shareholders spreads activism costs but jeopardizes passive shareholders’ support, which endogenizes the optimal boundary of shareholder collaboration. It also suggests that the prevalence of shareholder collaboration may contribute to the new trend that activists and boards reach agreements before proxy-fight voting much more often than in the past.





Crane, A. D., A. Koch, and S. Michenaud (2019). Institutional investor cliques and governance. Journal of Financial Economics 133 (1), 175–197.


McCahery, J. A., Z. Sautner, and L. T. Starks (2016). Behind the scenes: The corporate governance preferences of institutional investors. The Journal of Finance 71(6), 2905– 2932.


Pi, Shaoting (2020), Speaking with One Voice: Shareholder Collaboration on Activism. Available at SSRN 3599237.


Wong, Y. T. F. (2019). Wolves at the door: A closer look at hedge fund activism. Management Science. 2020 Jun;66(6):2347-71.



                                                   Liquidation, Bailout, and Bail-in: Insolvency Resolution Mechanisms and Bank Lending

                                    By Prof. Bart Lambrecht, Cambridge Centre for Finance and Cambridge Endowment for Research in Finance

                                                                                                        27 September 2021


The 2008-2010 financial crisis and the COVID-19 pandemic have highlighted the importance of an orderly insolvency resolution mechanism (hereafter IRM). Liquidation and bailout were the main IRMs prior to the crisis. In response to the criticisms on the bank bailouts during the financial crisis, regulators on both sides of the Atlantic devised new regulatory frameworks that attempt to minimize the use of public money to recapitalize failing banks. In this context the bail-in tool is probably the most important regulatory innovation. Bail-in is a statutory power in the hands of resolution authorities that permits them to write down part of the bank's liabilities or to convert the bank's liabilities into equity in order to preserve the bank as a going concern.

Existing papers on IRMs typically focus on one single type of IRM (usually liquidation or bailout). We lack papers that compare how firms behave and fare differently under the three IRMs. Important questions remain therefore unanswered. For example, conventional wisdom might predict that the guarantees and safety net provided by bailouts encourage excessive lending and risk-taking compared to liquidations and bail-ins. Is this indeed the case? How do different IRMs  affect payout, leverage, the loss in default, insolvency rates and the net value created by a bank? A comparative analysis of this type is missing. Prior studies have also ignored the role of IRMs for payout despite the fact that regulatory restrictions on bank payout during the financial and COVID-19 crises highlight the importance of payout policy for stress testing.

Lambrecht and Tse (2021) develop a unifying, dynamic model for IRMs that addresses the above questions from a microprudential perspective, i.e. they study the effects of insolvency regulation on an individual bank. The paper explores how the three IRMs affect the payout rate, as well as the quantity and quality of loans when these three decision variables are set by risk averse inside equityholders. Insiders can invest in risky assets (loans) of which the return follows a jump diffusion process. The diffusion component reflects continuous shocks to loan returns, whereas the jumps correspond to rare, negative shocks (hereafter referred to as ``crashes"). Crashes arrive according to an exogenous Poisson process, but the fraction of the assets that is destroyed by the crash (i.e. the crash risk exposure, a proxy for ``loan quality") is a decision variable under insiders' control (e.g. through collateral requirements). Assets with a higher exposure to crash risk carry a higher expected return.

The paper’s model shows that, from a micro-prudential perspective, banks adopt the lowest payout rate and create the most value net of any recapitalization costs under the bailout regime. On the downside, insiders have a stronger incentive to put banks at risk of insolvency, increasing the insolvency rate within the banking sector. The exposure of bank assets to crashes can, however, be kept low by giving insiders skin in the game in the event of a bailout. Excessive risk taking can be curbed by penalizing (rather than rewarding) managers for failure. To avoid bailouts with public money, a fraction of total bank payouts during good times can be put in a bailout fund to cover expected bailout costs. This can be implemented through a proportional tax on payouts, without distorting insiders' incentives. Such a bailout fund is viable if banking (net of recapitalization costs) is a positive NPV activity before and after the bailout. Furthermore, a bailout fund could resolve bank insolvency in a relatively speedy fashion.


Under the liquidation regime banks are least prone to insolvency but they incur the largest loss given default. Since managers receive nothing in liquidation and enjoy limited liability, they do not care about the size of the loss in default. A regulatory regime shift from bailout to liquidation therefore not only reduces the insolvency rate within the banking sector, but also increases the loss given default. This highlights a regulatory tradeoff that hitherto has not been recognized.

If the aim is to keep the amount of lending as well as the banks' exposure to crashes low then bail-ins with debt-to-equity conversion can be a superior alternative to liquidation or bailouts. The price to pay is that banks grow more slowly and generate less value under this bail-in regime. Lambrecht and Tse (2021) show that bail-ins with straight debt write-down create incentives for higher leverage, higher crash risk exposure, higher cost of debt, higher payout, and lower insider claim values than their debt-to-equity conversion counterpart. The increasing usage of principal debt write-down (PWD) bonds in recent years might arguably be a cause for concern with regulators.

The model also highlights a number of caveats associated with bail-ins more generally. First, banks need a sufficient amount of unsecured creditors that can be bailed in to avoid a bank run. The bail-in unravels if depositors are at risk. Second, whether bail-ins mitigate managers' incentives to issue low quality loans, depends on managers' payoff in a bail-in. As with bailouts, it is important that managers' fortunes remain closely linked at all times to the state of the bank; managers may have to be punished in the event of heavy losses. The BRRD stipulates that management should in principle be replaced in a bail-in. The model shows that merely replacing managers exacerbates moral hazard problems if managers have no liability and walk away scot-free. Finally, while a bail-in turns an insolvent bank into a solvent one, it does not inject any new capital (unlike bailouts). Bail-ins may therefore not resolve a bank's liquidity problems.



Lambrecht, B.M, and A. Tse, (2021), Liquidation, bailout, and bail-in: Insolvency resolution mechanisms and bank lending, Journal of Financial and Quantitative Analysis, forthcoming


What do capital structure models tell us about the payout policy?

Shiqi Chen

August 2021

At the beginning of the COVID crisis, a massive drop in global dividends occurred as firms struggled to preserve cash in this unprecedented world health crisis. While Royal Dutch Shell ad British Petroleum cut their dividend for the first time since World War II and the 2010 Deepwater Horizon Disaster, respectively, Exxon continued its thirty-seventh increase in the dividend. The CEO of Chevron, Michael Wirth, said that "the dividend is our number one priority and it is very secure. We're taking actions to preserve cash. It will have some impact on production in the near term, but we've stayed with our financial priorities, which include protecting the dividend." Eighteen months since the outbreak of the pandemic in March 2020, firms now are on track to recover dividends. Shell's dividend now stands at 24 cents, which has increased by 50\% compared to 16 cents since the first cut in 2020. The global dividend is forecast to bounce back to the pre-covid level as early as this year.


This event highlights the already well-documented fact of firms' payout policy: dividend smoothing, that is, once dividend is in place, firms are reluctant to cut it. Brav et al. (2001) conduct a survey on 384 financial executives and in-depth interviews with another 23, and conclude that "maintaining the dividend level is a priority on par with investment decisions. Managers express a strong desire to avoid dividend cuts, except in extraordinary circumstances". Skinner (2008) includes repurchases and shows that the aggregate payout is also smooth. However, the payout policy is not determined in isolation. Indeed, a firm's investment, financing and payout decisions are interdependent through the sources and uses of funds constraints. The constraint reduces the degree of freedom in the decision-making process. This implies that managing payout is going to have a ripple effect on the other two policies. In turn, if a firm adheres to a leverage target, shocks are absorbed by payout and investment, making payout smoothing hard to achieve. Thus, interesting questions arise -- whether the existing capital structure models can be reconciled with payout smoothing? What are the implications for financing and investment decisions when firms stick with a particular payout policy and vice versa?


Chen and Lambrecht (2021) develop a framework to examines the dynamic interactions of the three corporate policies. More specifically, rather than deriving the firm's optimal policies, this paper takes the selected payout or financing policy as exogenously given and examines the resultant implications on the remaining policy through the sources and uses of funds constraint, and the balance sheet identity. With this simple approach, the paper discovers interesting results and generates empirical predictions that can be brought to the data directly.


The paper shows that a positive (negative) NDR target can amplify (dampens) the influence of income shocks. First, if payout is given, positive (negative) gearing requires the firm to rebalance by investing (disinvesting) after positive income shocks, implying a procyclical (countercyclical) investment strategy. Second, suppose investment is fixed, a firm with a positive (negative) NDR target displaces a procyclical (countercyclical) payout policy. Both results suggest that a very negative NDR is less empirically plausible. While payout smoothing is hard to achieve under a very negative leverage target, the paper shows that it can be achieved if the investment policy is positively correlated with net income. Furthermore, the paper shows that a Lintner-style payout smoothing can be reconciled with a leverage target if the firm only partially adjusts toward the target. The maximum level of payout smoothing is achieved if the speed of adjustment is between zero and one-half. On the contrary, if the firm adopts a pecking-order style financing strategy, payout smoothing is easy to accomplish as long as the debt capacity is not yet reached. Under such circumstances, the incomes shocks are completely absorbed by changes in net debt, indicating that the greater degree of payout smoothing, the stronger negative relation between net incomes and variations in net debt.


Based on the analysis, the paper categorizes firms into different types.

  1. It predicts that firms with a significant positive net debt ratio target often have a stable, high net income and easy access to the capital market as a frequent rebalancing of the capital structure is required. In addition, these firms' risky assets should be either very liquid to enable asset sales or tangible to serve as collateral.
  2. Mature, stable and lack of growth opportunities firms are more likely to adopt a net debt ratio target around zero. These firms maintain payout levels by adopting a procyclical investment policy. Investment and payout are financed via retained income.
  3. The paper conjectures that firms with a significant negative net debt ratio target are rare because a very negative NDR implies empirically implausible features.
  4. Young and private firms that are not yet at their stationary states are candidates for a pecking-order style capital structure. With a pecking-order preference, the firm's NDR can be changed freely in response to heavy investment needs and income shocks. Firms that have just reached maturity are also candidates for pecking-order financing. They sit on a pile of cumulated debt and start paying down the debt, thereby reducing NDR to a long-run optimal NDR target.


All these results suggest that firms can switch between different capital structure regimes during their life-cycle.




Brav, Alon, John R. Graham, Campbell R. Harvey, and Roni Michaely. "Payout policy in the 21st century." Journal of financial economics 77, no. 3 (2005): 483-527.


Chen, S., & Lambrecht, B. M. (2021). Do capital structure models square with the dynamics of payout?. Available at SSRN 3854109.


Skinner, Douglas J. "The evolving relation between earnings, dividends, and stock repurchases." Journal of financial economics 87, no. 3 (2008): 582-609.


Oliver, Joshua. “UK Dividends RECOVERY Lags behind Other Markets.” Financial Times. Financial Times, July 30, 2021.


Sheppard, David. “Shell Raises Dividend and LAUNCHES Share Buybacks after Oil Prices Jump.” Subscribe to read | Financial Times. Financial Times, July 29, 2021.


Skinner, Douglas J. "The evolving relation between earnings, dividends, and stock repurchases." Journal of financial economics 87, no. 3 (2008): 582-609.


Stevens, Pippa. “Chevron CEO Says the Dividend Is the Company's No. 1 Priority and Is 'Very Secure'.” CNBC. CNBC, March 24, 2020.






















Why do green stocks outperform?

Mehrshad Motahari
15 July 2021

Green, or environmentally friendly, companies have generated significantly higher stock returns compared to their brown peers in recent years. This gap in the average returns of green and brown stocks was more pronounced during the first COVID wave last year, when the market crashed (Albuquerque et al., 2020). Theoretically, however, this ‘green premium’ seems counter-intuitive. Instead, brown companies should generate greater returns to compensate for their higher levels of climate risk. This is because brown companies are not only exposed to physical damages caused by their environmentally unfriendly practices but are also facing immense social and regulatory pressure to transition to more sustainable entities. This blog post looks at two recent studies by Pástor et al. (2021a; 2021b) in order to explain why green stocks have outperformed and how this is consistent with the theory.

Pástor et al. (2021b) elegantly posit how environmental friendliness should be linked to stock returns using an equilibrium model. The unique aspect of this model is that it accounts for investors’ preferences for sustainability, in addition to climate risk exposure, as forces that shape the returns of green stocks. Specifically, investors in this setting derive utility not just from financial wealth but also from the social impact of holding green companies. The implication of this preference for green holdings is that investors are willing to pay more for greener stocks, resulting in lower expected returns of these stocks in the cross-section. In equilibrium, stock returns vary by exposure to a systematic green factor that captures investors’ demand and appreciation for green investments at each point in time.

There is recent empirical evidence in support of Pástor et al. (2020)’s model. For example, Bolton and Kacperczyk (2021) use carbon emissions to classify firms into green and brown and show that firms with higher emissions generate higher stock returns. More importantly, they find that this result appears only in recent years, particularly after the Paris Agreement, which attracted investors’ attention to climate change. However, a large number of studies that use different environmental-friendliness measures, such as environmental, social and governance (ESG) ratings, report contradictory findings. Atz et al. (2021) conduct a meta-analysis of more than 1,000 research papers and find that the vast majority of these studies report a positive relationship between ESG and measures of financial performance, including stock returns.

Pástor et al. (2021a) reconcile the puzzling outperformance of green stocks by arguing that their higher realised returns in recent years are due to people’s growing appetite for green practices. This unanticipated increase in investors’ demand for the stock and consumers’ demand for the products of green companies has increased their market value, as captured by higher returns. However, this does not mean that green companies will generate higher expected returns going forward. Pástor et al. (2021a) establish this by showing that green companies only outperform when there is a shock related to bad climate news. After accounting for climate-concern shocks and sustainable fund flows in response to them, green stocks underperform brown ones.

Although Pástor et al. (2021a) present an intuitive picture regarding the recent performance of green stocks, further research is needed to understand how this will develop in the future and affect other areas of the market. Particularly, we do not know how much longer it takes for investors’ preferences to adjust so that we observe a consistent premium associated with climate risks. In fact, green stocks may still outperform in years to come as investors experience new climate-concern shocks. This can have consequences in other areas. As one example, Pástor et al. (2021a) show that part of the reason value stocks performed poorly in recent years is that most of them happen to be brown stocks. There may be other more specific implications that remain to be explored.


Rising Temperatures, Melting Ratings
Patrycja Klusakab, Matthew Agarwalabc, Matt Burkeab,
Moritz Kraemerde, and Kamiar Mohaddesf

a University of East Anglia, UK
b Bennett Institute for Public Policy, University of Cambridge, UK
c Centre for Social and Economic Research on the Global Environment, UEA, UK
Centre for Sustainable Finance, SOAS, UK & e Goethe-University, Frankfurt, Germany
Judge Business School & King’s College, University of Cambridge, UK

 This column was first published on VoxEU

[Standfirst] Enthusiasm for ‘greening the financial system’ is welcome, but does the explosion of ‘green’ finance indicators reflect the science? Using artificial intelligence, we construct the world’s first ‘climate smart’ sovereign credit rating and warn of climate-driven downgrades as early as 2030.

Climate change is “the biggest market failure the world has seen” (Stern 2008), with wide-ranging implications for stability – financial, economic, political, social, and environmental. As estimates of the economic consequences of climate change continue to grow, financial markets and business leaders face increasing pressure to factor climate risks into decision making. Climate change will hit markets from all directions. In boardrooms and at AGMs, what were once token whispers of eco-marketing have become serious discussions of extreme weather events, reputational risks, activist movements (from shareholders and consumers), regulatory and transition risks, asset stranding, and environmental litigation. In response, investors and regulators are calling for climate risk disclosures and a clear demonstration that portfolios and business models are consistent with the Paris Climate Agreement. Central bankers, finance ministers, the International Monetary Fund and United Nations are in on the action (see, for instance, Brunnermeier and Landau, 2020).

All this enthusiasm for ‘greening the financial system’ is welcome, but a fundamental challenge remains: financial decision makers lack the necessary information (see, for instance, Edmans, 2021). It is not enough to know that climate change is bad. Markets need credible, digestible information on how climate change translates into material risks. Instead, an explosion of ESG (environmental, societal, and governance) ratings and voluntary, ad hoc, unregulated corporate climate disclosures has created a confusing world of unfamiliar, incomparable, and conflicting metrics.

A chief concern is the lack of scientific foundations in risk disclosures (see, for instance, Fiedler et al., 2021). It is easy to see why. Climate models operate at global scales and project impacts over decades and centuries. Financial models do not. How should a high-frequency trading algorithm (think nanosecond resolution) adjust to the possibility that climate may reduce global output in 2100 by 10%? How should corporate climate disclosure address issues largely beyond their control, such as the carbon intensity of the national electricity grid, or the direction of government flood strategies? Most disclosures present companies as if they are independent of their physical (geographical) and macroeconomic surroundings. But this ignores crucial context. Climate change does not just affect firms individually, it affects countries and economies systemically. No corporate climate risk assessment is complete without also considering the effect of climate on sovereigns. Without scientific credibility, economic evidence, and decision-ready metrics, the field of green finance is open to charges of greenwash. Getting it wrong will be costly.

This is what motivated us to bridge the gap between climate science and real-world financial indicators (Klusak et al., 2021). Rather than constructing a new metric, we focused on one that is eminently familiar to financial decision makers: the sovereign credit rating. By linking climate science with economic models and real-world best practice in sovereign ratings, we simulate the effect of climate change on sovereign credit ratings for 108 countries under three different warming scenarios (see Figure 1).

We were guided by a single overarching principle: to remain as close as possible to climate science, economics, and real-world practice in the field of sovereign credit ratings. To the best of our knowledge, we are the first to simulate the effect of future climate change on sovereign credit ratings. Our approach means we can evaluate the effect of climate on ratings under various climate-economic scenarios and can provide initial estimates of the effects of climate-induced sovereign downgrades on the cost of public and corporate debt around the world.

Figure 1. Bridging the gap between climate science and financial indicators

Notes: Blue boxes (top row) represent the status quo in climate science, climate-economics, and sovereign credit ratings. Economic models translate scientific projections of temperature and precipitation changes into macroeconomic impacts (white box). Green boxes and arrows describe our novel approach to closing the remaining gap between climate economics and ratings.

To this end, we develop a random forest machine learning model to predict sovereign credit ratings and training it on ratings issued by S&P (one of the largest credit ratings agencies) from 2015-2020. Next, we combine climate economic models and S&P’s own natural disaster risk assessments to develop a set of climate-adjusted data describing various warming scenarios. We fed these climate-adjusted macroeconomic data to our ratings prediction model to simulate the effect of climate change on sovereign ratings. Finally, we calculate the additional cost of corporate and sovereign capital due to climate-induced sovereign downgrades (Figure 1, purple).

We focus on sovereign ratings because they are decision-ready. This is distinct from ESG ratings which, even if backed by credible science, still require investors to determine how they relate to material risk. In contrast, sovereign ratings are already used in a range of financial decision-making contexts (e.g. under Basel II rules, ratings directly affect the capital requirements of banks and insurance companies). They cover over US$ 66 trillion in sovereign debt, acting as ‘gatekeepers’ to global financial markets. Sovereign downgrades increase the cost of both public and corporate debt, influencing overall economic performance and significantly affecting fiscal sustainability.

We document three key empirical findings. In contrast to much of the climate-economics literature, we find material impacts of climate change as early as 2030. Under RCP 8.5 (a high emissions scenario that closely traces recent historical emissions), we find that 63 sovereigns suffer climate-induced sovereign downgrades of approximately 1.02 notches by 2030, rising to 80 sovereigns facing an average downgrade of 2.48 notches by 2100 (on a 20-notch scale). Figure 2 depicts the magnitude and geographical distribution of sovereign ratings changes predicted by our model by 2100 under RCP 8.5, showing that the most affected nations include Chile, China, Slovakia, Malaysia, Mexico, India, Peru and Canada all exceeding 5 notches downgrades. More importantly, our results show that virtually all countries, whether rich or poor, hot or cold, will suffer downgrades if the current trajectory of carbon emissions is maintained.

Figure 2. Global climate-induced sovereign ratings changes (2100, RCP 8.5)



Second, our data strongly suggests that stringent climate policy consistent with the Paris Climate Agreement will result in minimal impacts of climate on ratings – with an average downgrade of just 0.65 notches by 2100. Figure 3 shows that the most affected countries are Chile and China with climate-induced sovereign downgrades of 2.56 and 2.05 notches, respectively.


Third, we calculate the additional costs to sovereign debt – best interpreted as increases in annual interest payments due to climate-induced sovereign downgrades – in our sample to be between US$ 22–33 billion under RCP 2.6 (low emissions), rising to US$ 137–205 billion under RCP 8.5. These translate to additional annual costs of servicing corporate debt ranging from US$ 7.2–12.6 billion under RCP 2.6, and US$ 35.8–62.6 billion under RCP 8.5.


Figure 3. Global climate-induced sovereign ratings changes (2100, RCP 2.6)



There are caveats. Due to a lack of scientifically credible quantitative estimates of how climate change will impact social and political factors, these are excluded from our model (Oswald and Stern, 2019). Thus, our findings should be considered as conservative. Moreover, our results should be understood as scenario-based simulations rather than predictions. High emissions scenarios (e.g. RCP8.5) closely track recent observed trajectories and remain useful over near- to midterm timescales (Schwalm et al 2020). But the pace of renewables deployment and climate policy (e.g. regulations banning sales of new petrol and diesel vehicles) offer hope that future trajectories may fall closer to low emissions scenarios such as RCP 2.6 (Hausfather and Peters 2020).  We do not comment on the relative probabilities of any given warming scenario playing out in practice. Despite these caveats, our results are qualitatively similar when changing the time series of ratings used to train the prediction model, restricting the sample to investment grade sovereigns, and varying assumptions about the degree of temperature volatility within the baseline climate-economic model.


The key take-home messages are that it is possible to ‘do climate finance’ without compromising on scientific credibility, economic rigour, or decision-readiness. Existing climate science and economics can support credible, decision-ready green finance indicators. This research is of interest to investors, sovereigns and credit ratings agencies alike. Governments issue ever-longer dated bonds, of which life insurance companies and pension funds are eager buyers, thus enabling them to match their own long-term liabilities. Therefore, investors should consider the long-term creditworthiness of sovereign issuers. Currently there is no reliable yardstick for assessing sovereign creditworthiness beyond the current decade and this research fills this gap. Data on adaptation and resilience will be increasingly important. National statistical offices could play a decisive role, using the recently adopted UN System of Environmental Economic Accounts – Ecosystem Accounts (SEEA-EA) as a framework for tracing environmental investments and expenditure. Our study offers a first methodological approach to extend the long-term rating to an ultra-long-term reality. Based on the methodology applied here future research could focus on developing ultra-long ratings not only for sovereigns but also for other issuers including corporates.



Brunnermeier, M.K., and Landau J-P., (2020). Central Banks and Climate Change. VoxEU column 15 January 2020.


Edmans, A., (2021). The Dangers of Sustainability Metrics. VoxEU column 11 February 2021.


Fiedler, T., Pitman, A.J., Mackenzie, K., Wood, N., Jakob, C., and Perkins-Kirkpatrick, S.E., (2021). Business Risk and the Emergence of Climate Analytics. Nature Climate Change 11, 87–94.


Hausfather, Z., & Peters, G. P. (2020). Emissions - the “Business as Usual” Story is Misleading. Nature, 577, 618–620.


Kahn, M.E., Mohaddes, K., Ng, R.N., Pesaran, M.H., Raissi, M., and Yang, J.C., (2019). Long-Term Macroeconomic Effects of Climate Change: A Cross-Country Analysis. NBER Working Paper 26167.


Klusak, P, Agarwala, M., Burke, M., Kraemer, M., and Mohaddes, K., (2021). Rising Temperatures, Falling Ratings: The Effect of Climate Change on Sovereign Creditworthiness. Bennett Institute Working Paper.


Oswald, A. and Stern, N., (2019). Why are Economists Letting Down the World on Climate Change? VoxEU column 17 September 2019.


Schwalm, C. R., Glendon, S., & Duffy, P. B. (2020). RCP8.5 tracks cumulative CO2 emissions. Proceedings of the National Academy of Sciences of the United States of America, 117(33), 19656–19657.


Stern, N. (2008). The Economics of Climate ChangeAmerican Economic Review. 98:1-37.


This column was first published on VoxEU.



Demand for information

Shiqi Chen

May 2021


We are entering into an era of Big Data, in which the demand and supply of data and information are growing exponentially.  Such unprecedented demand and supply have triggered reforms in many sectors, where the financial industry is at the forefront of the trend. According to the reports by IBISWorld, the revenue of the U.S. financial data providers industry will reach $14.6 billion in 2021, and the expected annual growth rate is 8.2%. Financial data providers refer to those who collect and pack data from various sources such as brokers, regulatory filings, stock exchange feeds, and supply to financial institutions, investors, analysts, corporate executives etc.. 


The increasing importance of data information rests on the common questions faced by every participant in the financial market: How to make financial decisions under uncertainty. The growing complexity and surging uncertainties in the market induce participants to engage in a broader scale of information search, especially through different information intermediaries, in order to make better decisions and improve financial return. Examples are abundant: companies perform due diligence before investment, mergers and acquisition decisions; producers purchase consumer data for production and marketing purpose; fund managers rely on analysts' report to adjust portfolios; retail investors invest in mutual fund hoping to draw on their information advantage.


The substantial demand gives rise to many interesting questions.  Do agents bounded by incomplete information behave differently compared with those with perfect information? What is the influence of information acquisition on financial behaviour? In particular, given the dynamic nature of information flow and learning, how does the influence of information vary across states? What is the relation between public information and private information? Is more public information going to crowd out private information production, or the opposite?


Existing literature has shed light on some of the questions raised above. For example, earlier papers by Gennotte (1986), Brennan (1998), and others show that investors with incomplete information have different portfolio allocations compared with those with perfect information, and introduce the concept of “estimation risks”. Barberies (2000) and Xia (2001) study if stock returns are predictable from a signal such as dividend yield, does the traditional argument -- investors with long investment horizon (e.g. young investors) should invest more into equity than those with relatively short investment horizon (e.g. retired age investors)-- still hold. Even though using a very different modelling approach, both papers show that this is not necessarily the case.


More recently, Gao and Huang (2018), and Goldstein, Yang and Zuo (2020) use the implementation of the EDGAR system to study how modern information dissemination technology affects the relation between public and private information production. Gao and Huang (2018) find that after the full implementation of the EDGAR system, trades by investors with internet access become more informative, and the quality of sell-side analyst report increases significantly. Their findings suggest that public information does encourage private information production. Goldstein, Yang and Zuo (2020) focus on the staggered implementation process and discover a decrease in the investment-to-price sensitivity. They suggest that the propagation of public information can crowd out private information production incentives.


These intriguing results demonstrate that information does affect agents' behaviour along various dimensions. However, some interesting questions are still underexplored, especially from a theory perspective. For example, how does the demand for information varies over time? Do investors conduct more information acquisition in good or bad times? How do different factors such as risk preferences, beliefs, volatility affect the demand? Answer to these questions can help to explain agents' financial behaviour, for example, the trend-chasing phenomena, and passive and active investment choices as active investment can be considered as one way of information acquisition. It can also improve our understanding of the value created by information intermediaries. Bhattacharya et al. (2009) show that although there is a positive relation between news coverage and internet IPOs in the 1990s, the news frenzy fails to explain the interest bubble. The news information factor can only explain 2.9% of the astonishing 1646% return difference between internet and non-internet firms. This, to some extent, implies that preference or demand for information is state-dependent, so is the value of information.


Therefore, exploring how investors demand for information vary over different states can help to understand many market phenomena that otherwise remain puzzling. Let me stop here by referring to the interview of Robert Shillers in the New York Times: “ The fundamental problem is that the information obtained by any individual -- even one as well-placed as the chairman of the federal reserve -- is bound to be incomplete. "





Barberis, N., 2000. Investing for the long run when returns are predictable. The Journal of Finance55(1), pp.225-264.


Bhattacharya, U., Galpin, N., Ray, R. and Yu, X., 2009. The role of the media in the internet IPO bubble. Journal of Financial and Quantitative Analysis, pp.657-682.


Brennan, M.J., 1998. The role of learning in dynamic portfolio decisions. Review of Finance1(3), pp.295-306.


Gao, M. and Huang, J., 2020. Informing the market: The effect of modern information technologies on information production. The Review of Financial Studies33(4), pp.1367-1411.


Gennotte, G., 1986. Optimal portfolio choice under incomplete information. The Journal of Finance41(3), pp.733-746.


Goldstein, I., Yang, S. and Zuo, L., 2020. The real effects of modern information technologies (No. w27529). National Bureau of Economic Research.


Rodriguez, A. (2021). Financial Data Service Providers. [online] IBISWorld. Available at:


Xia, Y., 2001. Learning about predictability: The effects of parameter uncertainty on dynamic asset allocation. The Journal of Finance56(1), pp.205-246.




Climate finance

Mehrshad Motahari, CERF Research Associate
15 April 2021


The risk of climate change and its potentially drastic consequences has reshaped many industries in recent years. The financial industry has been no exception to this trend. Laurence D. Fink, the founder and chief executive of BlackRock, announced last year that his firm will make environmental sustainability a core decision-making objective and that he believes environmental awareness will soon lead to a ‘fundamental reshaping of finance’ (Times2020). This global concern about climate change has also brought about a new strand of literature in finance looking at the effects of climate change in financial markets (Hong et al., 2020). This blog post provides a brief summary of some of the key findings of this literature, together with their implications regarding the financial industry.

Barnett et al. (2020) provide one of the major recent studies, giving theoretical insights into the role of climate change in finance. They draw on continuous time decision theory to estimate the social cost of carbon. Under this framework, asset prices reflect the environmental damage of carbon emissions due to the uncertainty regarding the future effects of climate change on human welfare. The empirical evidence supports this and establishes that there is a risk premium associated with long-run climate change risk (Engle et al., 2020). A firm’s exposure to this risk factor is determined by their greenhouse gas emissions, among other measures of environmental friendliness.

In a recent study, Bolton & Kacperczyk (2021) explore the climate risk premium globally by looking at the distribution of corporate carbon emissions across 77 countries. Carbon emission in this setting captures the exposure to the risk associated with transitioning from fossil fuels to renewable energy. They find that companies with higher carbon emissions generate higher stock returns in most areas of the world, except for Africa, Australia, and South America. This positive premium is higher in countries with lower GDP per capita, less democratic systems, less developed healthcare systems, and those whose economic output relies more on their manufacturing sector. Also, they show that this carbon premium has risen significantly after the Paris agreement, which led to a rise in investor awareness.

However, it is not clear whether climate risks are priced correctly. Several studies show that investors do not pay attention to climate change risks and underreact to long-term climate trends (Hong et al., 2019; Krueger et al., 2020; Painter, 2020). Consequently, salient climate events (such as abnormally hot days) attract investors’ attentions to the problem, leading to firms with high carbon emissions underperforming compared to those with low emissions (Choi et al., 2020).

The characteristics of investors who hold the stock can also determine how climate change risks are priced. For example, Alok et al. (2020) show that fund managers based in regions with frequent climate disasters overreact to negative climate events and underweight stocks affected by climate disasters more heavily. Therefore, stocks facing high climate disaster risks are more likely to be underpriced when they are held by such fund managers.

The climate finance literature also highlights several important implications for the financial industry and, in particular, asset managers. Focardi & Fabozzi (2020) argue that the process of transitioning low carbon emissions will have costs as well as opportunities for asset managers. On the one hand, climate change will introduce various new sources of risk, including climate regulatory and compliance risk, physical damage to assets and companies, and adverse social and economic impacts. On the other hand, Focardi & Fabozzi (2020) argue that portfolios and indices can be constructed in a way that would have low carbon footprints without penalising returns.

Overall, the recent findings show that the financial industry is headed in the right direction when it comes to climate awareness and action. The new costs and hurdles in the way of investors also have not led to their disengagement in the markets. Krueger et al. (2020) conducted a survey of investors and found that they generally consider risk management and engagement to be the better approach for addressing climate risks than divestment. Probably the most important remaining question is whether the financial sector can do even more to tackle climate change and, if so, how.



Alok, S., Kumar, N., & Wermers, R. (2020), ‘Do fund managers misestimate climatic disaster risk?’, The Review of Financial Studies 33(3), 1146–1183.

Barnett, M., Brock, W., & Hansen, L. P. (2020), ‘Pricing uncertainty induced by climate change’, The Review of Financial Studies 33(3), 1024–1066.

Bolton, P. & Kacperczyk, M. (2021), Global Pricing of Carbon-Transition Risk, Technical report, National Bureau of Economic Research.

Choi, D., Gao, Z., & Jiang, W. (2020), ‘Attention to global warming’, The Review of Financial Studies 33(3), 1112–1145.

Engle, R. F., Giglio, S., Kelly, B., Lee, H., & Stroebel, J. (2020), ‘Hedging climate change news’, The Review of Financial Studies 33(3), 1184–1216.

Focardi, S. M. & Fabozzi, F. J. (2020), ‘Climate change and asset management’, The Journal of Portfolio Management 46(3), 95–107.

Hong, H., Li, F. W., & Xu, J. (2019), ‘Climate risks and market efficiency’, Journal of Econometrics 208(1), 265–281. Special Issue on Financial Engineering and Risk Management.

Hong, H., Karolyi, G. A., & Scheinkman, J. A. (2020), ‘Climate finance’, The Review of Financial Studies 33(3), 1011–1023.

Krueger, P., Sautner, Z., & Starks, L. T. (2020), ‘The importance of climate risks for institutional investors’, The Review of Financial Studies 33(3), 1067–1111.

Painter, M. (2020), ‘An inconvenient cost: The effects of climate change on municipal bonds’, Journal of Financial Economics 135(2), 468–482.

Times, N. (2020), ‘BlackRock CEO Larry Fink: Climate crisis will reshape finance’, New York Times.

Defining, Detecting and Measuring Asset Price Bubbles

Charlie Woodman, CERF Pre-doctorate

March 2021

One of the most frequently asked questions in the financial news media this year has been whether the stock market is in a ‘bubble’? Since the onset of the pandemic in early 2020, the S&P500 experienced a -35% return between the middle of February and the middle March before bouncing back to record levels by September. One year since the low in February, the index is now sitting around $3900, representing an annual return of more than 75%. This has led many to question, and in some cases proclaim, that US and international stock markets are in a bubble. But what exactly is a bubble, and how can we discern one?


A bubble is defined as the difference between price and fundamental value, where fundamental value is defined as the discounted sum of conditional expected future payoffs (Jarrow et al., 2010). Broadly speaking, there are two types of bubble: rational and irrational. During an irrational bubble, price deviates from fundamental value as a consequence of behavioural biases and limits to arbitrage, which prevents new information from being fully incorporated into prices. Irrational bubbles represent an arbitrage opportunity and a failure of market efficiency. Testing for an irrational bubble requires a model for fundamental value, which is unobservable, or on the model implied dynamics of a bubble component that has irrational origins. Thus, tests for irrational bubbles are plagued by the joint hypothesis problem. Moreover, theories of irrational bubble formation are most often designed to fit a pre-conceived notion that a bubble is a period of dramatic price ascent followed by a spectacular collapse. In other words, the starting point becomes one of an idealised price trajectory, rather than a deviation of price from fundamental value. This is unsatisfactory because there are many possible explanations for a rapid increase in prices that do not require behavioural explanations, chief among which is a decrease in the discount rate, which may be driven by changes in risk preferences or investors rationally updating their expectations about future returns. Bubbles can also occur during periods of sustained price decline, so long as fundamental value is falling at a greater rate. This leads to significant sample bias when designing and testing models using a handful of examples that are chosen on the basis of their convenient price characteristics.


The rational bubble literature can be decomposed into that which models prices in discrete-time and that which models prices in continuous-time. In the discrete-time setting, a rational bubble is born from a failure of the transversality condition in the rational asset pricing equation for stocks. Put simply, the bubble is expected to grow at a rate equal to the discount rate. The implied price behaviour is explosive, that is the expected price tomorrow is equal to the price today multiplied by a factor greater than 1. Phillips, Wu and Yu (2011) and Phillips, Shi and Yu (2015) tested this implication by applying recursive econometric tests for explosiveness to the NASDAQ and the S&P500. The results of these tests offer support for the dotcom bubble thesis and also identify several other shorter periods of explosiveness before and after. However, the discrete-time rational bubble framework does not permit the reformation of bubbles after they have burst and by implication, a bubble can only exist if it existed since the start of trading. Furthermore, discrete-time bubbles can only exist for infinitely lived assets. To that extent, the empirical literature is removed from the theory that motivates it, such that it is difficult to say whether the results are attributable to bubbles or other possible sources of explosive behaviour. For example, explosiveness could also be a consequence of a convex decrease in discount rates, which is not synonymous with the definition of a rational bubble.


In a continuous-time framework, bubbles can arise on finitely lived assets, exhibit finite duration and reform after they have burst. Moreover, continuous-time bubbles can be detected without need of a model for fundamental value and where a model is needed, it can be independently verified without incurring the joint hypothesis problem. Continuous-time bubbles are characterised as periods in which the discounted price process is a strict local martingale, which we understand as a local martingale that is not a martingale. An assets price process is a strict local martingale if volatility is a convex function of price. Therefore, continuous-time bubbles differ from discrete-time bubbles insofar as the emphasis shifts from explosiveness in price to explosiveness in volatility. The price paths produced by continuous-time bubbles are often similar to the typical J-shaped runups associated with historically recognised bubble periods, but the model also embeds a much more diverse set of potential price trajectories that are not captured by discrete-time alternatives. This is because strict local martingales are strictly continuous-time phenomena, reflecting the additional flexibility of continuous-time models in finance. Note also that both discrete and continuous-time rational bubbles are consistent with the notion of no-arbitrage, but neither can exist unless the market is incomplete (Jarrow et al., 2010).


Existing testing procedures for continuous-time bubbles can be classified into two groups: parametric and non-parametric. Parametric procedures postulate a model for price, estimate the parameters that govern the process and check if the estimates fall within the interval that renders price a strict local martingale. Non-parametric testing procedures estimate the distribution of volatility over the observable price interval and then extrapolate into the tail to check that volatility is a convex function of price. However, existing parametric approaches are slow to respond to changing market conditions and non-parametric methods often produce false positive signals in real-time applications.


Research on the theoretical modelling and empirical detection of asset price bubbles remains a fertile area of research. For example, recent research by Fusari et al. (2020) uses option price data to detect bubbles. The premise is that call option prices can contain bubbles but put option prices cannot, such that one can estimate the magnitude of the bubble in the underlying asset by calculating the difference between observed and model implied prices of call options using a model that accurately prices puts. Another recent approach from Bashchenko and Marchal (2020) applies machine learning techniques, specifically Recurrent Neural Networks (RNN’s) with Long-Short Term Memory (LSTM) cell architecture to estimate a parametric model that classifies assets as true martingales or strict local martingales at each time interval. The model is trained on simulated price data and achieves an out-of-sample accuracy in excess of 83%. Finally, recent advances in the theoretical literature continue to provide new insights and new testable hypotheses that will contribute to an improved understanding of bubbles, and the implications of these phenomena for investors, policy makers and the wider economy.





  • Bashchenko, O. and Marchal, A., 2020. Deep Learning for Asset Bubbles Detection. Available at SSRN 3531154.
  • Fusari, N., Jarrow, R. and Lamichhane, S., 2020. Testing for Asset Price Bubbles using Options Data. Available at SSRN 3670999.
  • Jarrow, R.A., Protter, P. and Shimbo, K., 2010. Asset price bubbles in incomplete markets. Mathematical Finance: An International Journal of Mathematics, Statistics and Financial Economics20(2), pp.145-185.
  • Phillips, P.C., Wu, Y. and Yu, J., 2011. Explosive behavior in the 1990s Nasdaq: When did exuberance escalate asset values? International economic review52(1), pp.201-226.
  • Phillips, P.C., Shi, S. and Yu, J., 2015. Testing for multiple bubbles: Historical episodes of exuberance and collapse in the S&P 500. International economic review56(4), pp.1043-1078.














Firms' Capital Structure Dynamics, Market Competition, and Industry Dynamics


Shiqi Chen, CERF Research Associate

February 2021


Debt-equity conflict is undoubtfully one of the core paradigms of corporate finance research. It is well-known that, once debt is in place, the misalignment of interest between debt holders and equity holders can lead to asset substitution and inefficient underinvestment problems, which are initially identified by Jensen and Meckling (1976) and Myers (1977), respectively.  More recently, studies by Admati et al. (2018) and DeMarzo and He (2020) highlight another consequence of debt-equity conflicts: the leverage ratchet effect, that is, when firms cannot commit to future debt levels, once debt is in place, equity holders not only are reluctant to reduce leverage voluntarily, but also have an incentive to increases the firm’s leverage to the detriment of debt holders. They show that such debt-equity conflicts due to non-commitment give rise to leverage dynamics that differ substantially from what is implied by the static trade-off model, and may help to explain many empirical phenomena that otherwise remain puzzling—for example, the reluctance of distressed firms to recapitalize. 


However, it is important to note that an individual firm is not a solo player in the market. Firms' entry, exit, production and financing decisions need to be considered in a broader context. Indeed, firms' capital structure dynamics and product market decisions are closely intertwined. If debt-equity conflicts persist among firms, the corresponding influence can aggregate rapidly at the industry level and have a profound impact on product market competition, output prices and other dimensions of industry dynamics. In turn, the degree of product market competition and equilibrium output prices are key determinants of industry players' profitability and survivorship. Therefore, product market behaviour affects the evolution of debt-equity conflicts and funding choices at the firm level. Such interaction (shown in the figure below) is highlighted in Zingales (1998), "in the absence of a structural model, we cannot determine whether it is the product market competition that affects capital structure choices or a firm's capital structure that affects its competitive position and its survival". 



What has happened in the oil industry in 2020 mirrors such interactions. In 2020, the price war between Russia and Saudi Arabia, coupled with prolonged pandemic and subsequent collapse in global demand had tumbled many heavily indebted oil and gas producers. The West Taxes Intermediate was even trading in the negative territory for the first time in April. Since 2008, the surges in crude export have led to a shale boom in North America, as well as an all-time high aggregate debt level in the oil and gas sector. According to the report by Haynes and Boone, 46 North American oil and gas producers have filed Chapter 11 bankruptcy in 2020, among which 14 are billion-dollar bankruptcies. The imminent consolidation and reshuffle within the industry are going to create further fluctuations in oil prices and variations in the survivors' capital structure. These emphasize the intriguing interdependencies between firms' financial decisions, market competition and industry dynamics.  


In the article " Industry Dynamics and Capital Structure (Non)Commitment", CERF Research Associate Shiqi Chen and collaborator Hui Xu (University of Lancaster) attempt to address these interactions. It develops a competitive equilibrium model to understand how the debt-equity conflicts arising from the absence of equity holders' commitment to future debt levels affect industry dynamics and the corresponding feedback effect on firms' financial decisions. 


The article shows that shareholders' resistance to leverage reduction and incentives to increase leverage make debt financing more expensive. As a result, entry into the industry becomes harder, which reduces the degree of market competition and raises the equilibrium output price. The increase in the output price, in turn, improves the profitability of industry incumbents and makes the shareholders willing to wait longer before shutting down firms, thereby alleviating inefficient liquidation and the agency costs generated by non-commitment. By looking at the stationary industry distribution of firm in terms of debt-scaled cashflow, they find that, compared with the commitment case, non-commitment and the resultant higher output price increase the number of firms in the high leverage region and the overall average industry leverage. More firms now stand close to the exit boundary. Such distributional effect gives rise to a higher frequency of entry and exit, and consequently, a higher market turnover rate at the equilibrium. The results suggest that debt-equity conflicts at the firm level can ramp up and have profound implications on industry dynamics. 




References mentioned:

Admati, A. R., P. M. DeMarzo, M. F. Hellwig, and P. Pfleiderer (2018): “The leverage ratchet effect,” Journal of Finance, 73(1), 145–198.

DeMarzo, P., and Z. He (2020): “Leverage dynamics without commitment,” Journal of Finance, Forthcoming.

Jensen, M. C., and W. H. Meckling (1976): “Theory of the firm: Managerial behavior, agency costs and ownership structure,” Journal of Financial Economics, 3(4), 305–360.

Haynes, and Boone (2020): “Haynes and Boone, LLP Oil Patch Bankruptcy Monitor (31 December),” Available at:  (Accessed 11 February 2021).

Myers, S. C. (1977): “Determinants of corporate borrowing,” Journal of Financial Eco- nomics, 5(2), 147–175.

Zingales, L. (1998): “Survival of the fittest or the fattest? Exit and financing in the trucking industry,” Journal of Finance, 53(3), 905–938.



When Sheer Predictive Power is not Good Enough:
Towards Accountability in Machine Learning Applications

by Thies Lindenthal (CERF Fellow) and Wayne Xinwei Wan

The law is clear: Housing-related decisions must be free of discrimination, at least in terms of gender, age, race, ethnicity, or disabilities. Easier said than done for the plethora of machine learning (ML) empowered systems for mortgage evaluation, tenant screening, i-buying schemes or other ‘disruptions’. A rapidly expanding literature explores the potential of ML algorithms, introducing novel measurements of the physical environments or using these estimates to
improve the traditional real estate valuation and urban planning processes (Glaeser et al., 2018; Johnson et al., 2020; Karimi et al., 2019; Lindenthal & Johnson, 2019; Liu et al., 2017; Rossetti et al., 2019; Schmidt & Lindenthal, 2020). These studies, again and again, demonstrated the undisputed power of ML systems as prediction machines. Still, it remains difficult to establish causality o to understand the internal mechanism of the models. An “accountability gap”
(Adadi & Berrada, 2018) remains: How do the models arrive at their prediction results? Can we trust them not to bend rules or cut corners?
This accountability gap holds back the deployment of ML-enabled systems in real-life situations (Ibrahim et al., 2020; Krause, 2019). If system engineers cannot observe the inner workings of the models, how can they guarantee reliable outcomes? Further, the accountability gap also leads to obvious dangers: Flaws in prediction machines are not easily discernible by classic crossvalidation
approaches (Ribeiro et al., 2016). Traditional ML model validation metrics such as
the magnitude of predictions errors or F 1 -scores can evaluate the models’ predictive performance, but they provide limited insights for addressing the accountability gap. Training ML models is a software development process at heart. We believe that ML developers therefore should follow best practices and industry-standards in software testing. Particularly, the system testing stage of software test regimes is essential: It verifies whether an integrated system performs the exact function as required in the initial
design (Ammann & Offutt, 2016). For ML applications, this system testing stage can help to close the accountability gap and to improve the trustworthiness of the resulting models. After all, thorough system testing has verified that system is not veering off into dangerous terrain but stays on the pre-defined path.
System testing should be conducted before evaluating the model’s prediction accuracy, which can be considered as the acceptance testing stage in the software testing framework. In recent years, several up-to-date model interpretation algorithms have been developed, which attempt
to reduce the complexity by providing an individual explanation that solely justifies the prediction result for one specific instance (Lei et al., 2018; Lundberg & Lee, 2017; Selvaraju et al., 2017; Ribeiro et al., 2016). However, most of the current local interpretation tools are qualitative and require human inspection for each individual sample. Thus, these tools for model verification do not easily scale up with large sample size. One example – to demonstrate the general approach In this paper, we develop an explicit system-testing stage for an ML-powered classifier for images of residential real estate. In formalizing a novel model verification test, we first define categories of relevant and irrelevant information in the training images that we are interested in testing.
Then we identify the elements of the input images that are found to be most relevant for classification by the ML model (i.e., which pixels matter most?), using a local model interpretation algorithm. Finally, we calculate what proportion the interpretable information originates from our defined categories of relevant/irrelevant information, and we use this proportion as the model verification test score. High scores imply that the model bases its
predictions on meaningful attributes and not on irrelevant information, e.g. in the background of the images.
Specifically, we augment an off-the-shelf image classifier that has been re-trained to detect architectural styles of residential buildings in the UK (see my previous blog post for CERF) . This type of computer-vision based classifier is selected as an illustration due to its popularity in real estate and urban studies (Naik et al., 2016), although our approach also extends to other ML classifiers, e.g. in text-mining (Fan et al., 2019; Shen, 2018).
Following architects’ advice, we define facades of houses, windows and doors as the most relevant attributes for classifying building styles, and we consider trees and cars as the irrelevant information. These objects are detected in the input images using the object detection algorithms. Further, we implement the local interpretable model-agnostic explanation algorithm (LIME) – one of the popular local model interpretation tools – to find the areas in the input images that best explains the predictions.
Finally, by comparing these interpretable areas and the areas of the objects, we calculate the verification test score/ratio for this exemplar model. Our results reveal that the classifier indeed selects information from house, windows, and doors for predicting building vintages, and it also excludes the irrelevant information from the trees as we hope. More importantly, these findings
improve the trustworthiness of the prediction results, as well as the associated implications between building vintages and real estate values (Johnson et al., 2020; Lindenthal & Johnson, 2019; Schmidt & Lindenthal, 2020). However, we find that the model also considers information from the cars for its predictions.
Our study contributes to the growing literature that applies ML in real estate and urban studies from two aspects. Firstly, we propose a ML application framework with an additional system testing stage, which aims to address the accountability gap and improve the trustworthiness of the results. Using a commonly applied computer vision model in the literature as an example, we demonstrate the capability of our approach to check whether the model is under the threat
of capturing undesirable information for predictions. Secondly, we extend the existing qualitative model-interpretation techniques to a formal
quantitative test. Methodology-wise, this helps to scale up the model interpretation analyses for a large sample size, which is essential for most of the applications in real estate and urban studies. In summary, our proposed method extends for other ML models and, due to the essence of closing the accountability gap, this study has important implications for ML applications in real estate and urban studies, as well as in other subjects beyond.
Fig 1: First, find areas that are relevant when e.g. describing a home’s vintage

Fig 2: Second, compare to the image areas that actually lead to a specific classification: How good
is the overlap?


Do Firm Locations Affect Stock Prices?

Mehrshad Motahari, CERF Research Associate
15 January 2021


A large body of literature documents how firms' geographical locations can affect their stock returns. For example, Pirinsky and Wang (2006) show that the stock returns of firms headquartered in the same geographical area comove with each other. They argue that this comovement is not related to economic fundamentals but to the trading patterns of local investors. Various papers attribute this excess comovement to local bias that induces local investors to take larger positions in local stocks. Bernile et al. (2015) suggest that even institutional investors overweigh firms whose 10-Ks frequently mention the investors' state. In contrast, Kumar et al. (2013) show that retail trades cause comovement in local stocks, whereas institutional trades mitigate the issue.

Local bias also leads to the incorporation of the behaviours and preferences of local investors in the prices of local stocks. Korniotis and Kumar (2013) highlight that local risk tolerance affects the returns of local stocks. Specifically, they argue that US state-level heterogeneity in economic conditions leads to variations in investor risk tolerance across states, and heterogeneous risk tolerance results in variations in the cross section of stock returns. In other words, the economic conditions of the region in which a firm is based can affect its stock price, irrespective of the firm’s fundamentals.

In the article ‘Geographic Heterogeneity, Local Sentiment, and Market Anomalies’, CERF Research Associate Mehrshad Motahari, shows that market anomalies (i.e. strategies that beat the market, such as momentum) have different performances for stocks headquartered in different US states. In other words, if we break the US cross section down into states in which anomalies have recently worked well and those in which anomalies have worked poorly, we observe that the first group will continue to have a better performance in the future. Using a famous anomaly variable such as momentum, the study shows that we can predict how well momentum predicts future returns by taking a firm’s headquarters into account. To illustrate, if we construct the momentum strategy (i.e. going long on high momentum stocks and low on low momentum ones) for stocks headquartered in either California or Texas in 2020 and find that this strategy works better for Californian stocks, it will likely continue to generate higher alphas for stocks in California in 2021.

This pattern can be explained by arguing that investors in different regions have different levels of sentiment. Local investors in states experiencing a relatively higher level of sentiment are more likely to buy excessively or overpay for local stocks. In the presence of local bias, short-selling impediments and information uncertainty, this behaviour exacerbates stock overpricing. The resulting mispricing is more severe in states experiencing higher sentiment and will persist due to limits to arbitrage.

The study also looks at analyst forecasting errors as a proxy for information uncertainty surrounding stocks. The idea is that investor biases and sentiment levels are more likely to be reflected in prices when the stock is hard to be valued. In line with this, the findings show that geography predicts the performance of anomalies only for stocks experiencing higher levels of analyst forecasting errors.

Overall, the findings of this research and other papers in this area imply that it is preferable to tilt a portfolio towards stocks in specific geographic regions when devising systematic trading strategies to exploit mispricing. More importantly, studies on this subject establish that firms’ locations have more extensive effects on stock prices than previously documented. That is, the location of a stock can determine how the fundamentals of the stock will be priced in the cross section and relative to other local and non-local firms.



Bernile, G., Kumar, A., and Sulaeman, J. (2015) ‘Home away from home: Geography of information and local investors’, Review of Financial Studies, vol. 28, pp. 2009–2049.

Korniotis, G. M. and Kumar, A. (2013) ‘State-level business cycles and local return predictability’, Journal of Finance, vol. 68, pp. 1037–1096.

Kumar, A., Page, J. K., and Spalt, O. G. (2013) ‘Investor sentiment and return comovements: Evidence from stock splits and headquarters changes, Review of Finance, vol. 17, pp. 921–953.

Pirinsky, C. and Wang, C. (2006) ‘Does corporate headquarters location matter for stock returns?’, Journal of Finance, vol.61, pp. 1991–2015.