2. Firms’ exposure to political uncertainty and equity pricing

In this section, we outline the conceptual framework that will guide our empirical strategy. We draw on a strand of research in empirical finance about the impact of uncertainty on the time series and cross section of equity returns (Brogaard and Detzel, Reference Brogaard and Detzel2015; Bali et al., Reference Bali, Brown and Tang2017). Our starting proposition is that firms are exposed to different sources of political risk. They include unstable or polarised political systems (for instance, high executive turnover, lack of democratic accountability or a high level of partisan conflict), poor institutions (corruption or judiciary inefficiency), high levels of socio-political conflicts (both internal and external, including wars) and unfavourable policy actions by governments (from changes in regulation, taxation and spending to outright expropriations and asset seizures) (Julio and Yook, Reference Julio and Yook2012; Bekaert et al., Reference Bekaert, Harvey, Lundblad and Siegel2014; Lehkonen and Heimonen, Reference Lehkonen and Heimonen2015; Azzimonti, Reference Azzimonti2018).

Equity investors can better assess firms’ exposure to political risk and its effect on future returns in a stable political environment—that is, when they have adequate information about the predictable range of possible political outcomes and their impact on the distribution of future returns around their expected meanFootnote ¹. On the contrary, when this information is inadequate—for instance, because the political environment is highly volatile and unpredictable—investors become “unsure about the correct probability laws” governing future mean returns and their distribution (Anderson et al., Reference Anderson, Ghysels and Juergens2009, p. 234). This type of political uncertaintyFootnote ² increases investors’ exposure to reinvestment risk—that is, to “a deterioration in investment opportunities that affects the prospects of long-term investors when they reinvest their wealth in financial markets” (Gonçalves, Reference Gonçalves2021, p. 2154). As a consequence, it should “move” equity prices as investors use the continuous flow of political news to revise their prior beliefs (Pastor and Veronesi, Reference Pastor and Veronesi2013, pp. 521–522). The testable prediction is that political uncertainty should predict future negative aggregate returns.

In turn, how political uncertainty affects the cross section of returns depends on its diversifiability. If it is perceived as a source of systematic risk, theory predicts that innovations in uncertainty should increase investors’ discount rate, leading to a generalised fall in equity prices and an increase in expected returns, as risk-averse investors demand a higher future compensation for bearing more systematic risk (Pastor and Veronesi, Reference Pastor and Veronesi2013). However, there is substantial evidence that sensitivity to uncertainty is higher for firms in industries more dependent on government spending or regulation or with a high level of asset specificity (Belo et al., Reference Belo, Gala and Li2013; Baker et al., Reference Baker, Bloom and Davis2016; Gulen and Ion, Reference Gulen and Ion2016; Kim and Kung, Reference Kim and Kung2017). In this case, theory predicts that risk-averse investors will increase their demand for stocks of firms they perceive as less exposed to political risk, pushing up their prices. On the contrary, stock of firms more exposed to political risk will move in the opposite direction. Hence, some stocks will covariate positively and others negatively with political uncertainty—that is, they will have positive and negative uncertainty betas, respectively. We will test these alternative hypotheses empirically.

Positive beta stocks provide a hedge against future unfavourable states because they increase investors’ wealth exactly when expectations about future investment opportunities deteriorate. However, there is a trade-off between this contemporaneous upside wealth effect and a future downside effect. When uncertainty falls and expectations improve, underpriced negative beta stocks will bounce back, generating higher returns that compensate investors for holding riskier assets (Maio and Santa-Clara, Reference Maio and Santa-Clara2012; Bali et al., Reference Bali, Brown and Tang2017, pp. 472–473)Footnote ³. Hence, a third testable prediction is that hedge stocks should earn in the future lower excess returns (over a safe asset, such as government bonds) than stocks with negative beta—that is, they should earn a negative risk premiumFootnote ⁴.

An important insight from the literature is that the impact of uncertainty on equities may vary across states of the economy and financial markets. For instance, some studies suggest that the demand for stocks with positive uncertainty beta is stronger in “bear” stock markets with low investor sentiment (Nartea et al., Reference Nartea, Bai and Wu2020). This is consistent with the idea that, in bad states, policy changes are more likely and uncertainty about future government actions is higher (Julio and Yook, Reference Julio and Yook2012; Pastor and Veronesi, Reference Pastor and Veronesi2013; Brogaard and Detzel, Reference Brogaard and Detzel2015). However, the impact of this type of uncertainty can be positive if investors expect public intervention to provide a “put” protection that contributes to stabilise asset prices (Pastor and Veronesi, Reference Pastor and Veronesi2013, p. 521). As we discuss later, this insight is especially relevant for our historical study, as a “bear” market dominated during most of the 1930–1936 period and interventions of public authorities on the stock market were systematic and extensive.

3. Political and economic uncertainty in the nascent Spanish democracy

Political scientists observe that “nascent” democracies, born out of the crisis of autocratic regimes, tend to generate much higher political uncertainty than consolidated democracies, as they lack stable rules for the resolution of conflicts, the structure of political interaction can evolve unpredictably and the possibility of an authoritarian reversal or a democratic break-down is not trivial (Lupu and Riedl, Reference Lupu and Riedl2012, pp. 1344–1345). The political environment of interwar Spain reflects these features very well. After the big political surprise of April 1931 (the unanticipated success of republican parties in municipal elections, the voluntary exile of the monarch and the peaceful proclamation of the Republic), the disbandment of dynastic forces and the emergence of new parties and leaders led to an entirely new structure of political competition. Political representation was highly fragmented. Unstable coalitions led to twenty cabinets in 63 months. The snap elections of 1933 and 1936 produced major swings in electoral support. The longevity of the new democratic institutions was highly uncertain. Civil–military relations remained unsettled, with frequent rumours of, and occasional attempts at military coups. The commitment to constitutional rules was weak among parties at the extremes of the political spectrum. The electoral success of regional nationalisms in Catalonia and the Basque Country, and the explicit secessionist goals of Catalan elites, questioned the survival of Spain as a unitary state. Linz (Reference Linz, Linz and Stepan1978) described the political situation of the period as “extremely fluid and unstructured”.

The new regime was also torn between opposed economic preferences of social forces and political parties. As in other countries in interwar Continental Europe, the rise of mass politics led to a violent clash of movements opposed to liberal democracy on the left and right ends of the political spectrum. In spite of their profound differences, they embraced the “primacy of politics”—that is, the willingness to use political power to create “a new socioeconomic order” in which the state “could and should control markets without destroying them” (Berman, Reference Berman2006, pp. 16–17). On the one hand, left parties and workers’ unions pursued extensive social reforms in land and labour markets and were open to possible alterations of private property rights through selective expropriations and nationalisations. On the other end, the Catholic right advocated an authoritarian corporatist regime that would bring employers and employees under the control of the state. With rising political polarisation, aggravated by international factors (such as the rise to power of the Nazi party in Germany), the public discourse became increasingly imbued of revolutionary and counter-revolutionary rhetoric, as the “ghost” of a radical political and socio-economic change was invoked or exorcised as a tool of political mobilisation (González-Calleja et al., Reference González Calleja, Cobo Romero, Martínez Rus and Sánchez Pérez2015, pp. 357–387, 769–844). At the same time, a state of permanent economic crisis reinforced the mobilisation of the working class. Electoral participation rates ranged between 65 and 70 per cent, trade union density escalated to 30 per cent of active workers and strike rates were the highest in Europe (Domènech and Miley, Reference Domènech and Miley2013). Massive waves of collective actions, occasional insurrectional attempts and frequent outbursts of politically- or religion-motivated violence elicited the state’s violent reaction, raising fears of a possible radical change of the political and socio-economic regime (Linz, Reference Linz, Linz and Stepan1978, pp. 187–194)Footnote ⁵.

Uncertainty about economic policies was an important additional dimension. The economy suffered from the international propagation of the Great Depression. In the two years after the fall of the dictatorship of General Primo de Rivera, GDP contracted by 19 per cent, also due to a twin (banking and currency) crisis triggered by the regime change of April 1931 (Betrán and Pons, Reference Betrán and Pons2019). The level of economic activity stagnated between 1932 and 1934 and contracted by another 12 per cent until the spring of 1936 (Albers, Reference Albers2018). The 1931 crisis revived tensions with the Bank of Spain, which in fact had been wrestling with different executives over the stabilisation of the exchange rate since 1928 (Betrán et al., Reference Betrán, Martín-Aceña and Pons2012; Martín-Aceña et al., Reference Martín-Aceña, Pons and Betrán2014; Martínez Ruiz and Nogués-Marco, Reference Martínez Ruiz and Nogués-Marco2014; Jorge-Sotelo, Reference Jorge-Sotelo2020). The following years were characterised by constant uncertainty about governments’ ability to pursue fiscal consolidation and the stabilisation of the exchange rate (facilitated by Sterling’s exit from the Gold Standard and finally achieved at the end of 1933 with a peg to the French Franc and the introduction of binding capital controls). At the same time, governments, private banks and the Bank of Spain engaged in a prolonged dispute on the relaxation of monetary policy to revive a depressed economy (Martín-Aceña, Reference Martín-Aceña1984, pp. 261–271). The disruption of trade relationships added to the uncertainty in the same period (Betrán and Huberman, Reference Betrán and Huberman2022, Reference Betrán and Huberman2024). While the political environment did little to enhance the management of the macroeconomy, economic policy was a significant and separate source of uncertainty for Spanish firms. For this reason, in our empirical analysis we will control for its possible impact.

4. Equity returns and political uncertainty: the dataset

The validity of our theoretical framework rests on the assumption that political uncertainty is a state variable correlated with future investment opportunities. We test this hypothesis empirically. Our proxy for the set of investment opportunities is an equity portfolio of common stocks of twenty-seven Spanish firms regularly traded on the Madrid Stock Exchange. They represent 60 per cent of nominal paid-out capital and provide a balanced representation of the sector-wise composition of the marketFootnote ⁶. In order to identify different phases in the market, Figure 1 plots the 36-month cumulative total return of our capitalisation-weighted market portfolioFootnote ⁷. The regime transition and the early years of the Republic coincided with a long “bear” market that lasted more than three years and reached its bottom level in October 1931, with a cumulative loss of 50 per cent of its 1928 value. The market remained depressed (with occasional and short-lived recoveries) for the following two years. A sustained recovery became visible only at the end of 1933, in coincidence with the electoral victory of the centre-right coalition. The “bull” market peaked at the end of 1935 and turned into a new crash after the victory of the Popular Front in the general elections of February 1936. The correspondence between political and stock market cycles was emphasised by all coeval observers.

Figure 1.

“Bears” and “bulls” in the Madrid stock market, January 1930–July 1936.

This figure shows the 36-month rolling cumulative total return of the capitalisation-weighted equity market portfolio. The zero line signals the transition from “bear” to “bull” markets.

A fundamental characteristic of this period was the continuous intervention of stock exchange authorities (Junta Sindical), in coordination with the Ministry of Finance, during periods of acute market stress. As a rule, interventions on equities took the form of selective temporary price caps (topes) on firms especially affected by selling. Less frequently, trading of specific stocks was limited to transactions among brokers or temporarily suspended. Such selective restrictions, motivated by the need to “prevent financial panics or overly abrupt changes” (Torrente Fortuño, Reference Torrente Fortuño1934, p. 37), were often a response to sharp movements driven by adverse political newsFootnote ⁸. This back-stop mechanism on prices operated as a protective “put” option that limited “bad” volatility, thus partially insuring investors against downside risk in periods of heightened uncertainty. It also enhanced the supply of liquidity by the Bank of Spain through collateralised loans against the pledge of corporate securities (Bartolomé Rodríguez, Reference Bartolomé Rodríguez2017). Restrictions were sufficiently widespread and systematic to be perceived as a permanent management tool and affect investors’ expectationsFootnote ⁹. In fact, they were criticised by groups of traders for interfering with the pricing mechanism and limiting their scope for profits from short-sellingFootnote ¹⁰.

To measure political uncertainty, we use news from the coeval press, in line with recent studies (Mathy and Ziebarth, Reference Mathy and Ziebarth2017; Lennard, Reference Lennard2020; Verdickt, Reference Verdickt2020). Our source is ABC, a widely read conservative newspaper published in Madrid. ABC reported daily stock market data, published professional stock exchange chronicles and represented a fundamental source of information for investorsFootnote ¹¹. Our index is based on count data of keywords systematically associated with news about socio-political conflicts that could make firms’ future profitability more unpredictableFootnote ¹². The underlying assumption, common to an extensive literature (Alesina and Perotti, Reference Alesina and Perotti1996; Voth, Reference Voth2002; Funke et al., Reference Funke, Schularick and Trebesch2016; Cortes and Weidenmier, Reference Cortes and Weidenmier2019), is that time variations in the arrival of “bad” news identify periods during which investors perceived a heightened level of political uncertainty. Our index (expressed in standard deviations) is shown in Figure 2. A structural break is clearly visible at the end of 1930, in coincidence with the first revolutionary mobilisation of republican parties. We can also observe that the level of uncertainty remained high and persistent throughout the periodFootnote ¹³. Peaks of the index were often clustered and coincided with important socio-political processes, such as electoral campaigns and waves of collective action.

Figure 2.

Political uncertainty index.

This figure shows our index of political uncertainty based on ABC news. Values for May 1931 and August-November 1932 are missing due to the suspension of publication of ABC and other conservative newspapers by the government. The index is based on the first principal component of standardised individual time series of political “bad” news. Grey areas indicate months around the municipal elections of April 1931, the constituent elections of June 1931 and the general elections of November 1933 and February 1936.

Importantly, coeval chronicles clearly identified socio-political factors as key determinants of investors’ behaviour. For instance, in January 1932 (index value 3.5), ABC noted that “political directives and social manifestations are the two factors that exercised the most direct influence. … In hard and turbulent periods, [securities markets] are not in control of their own resources, but are slaves of external events” (10 January 1932). One year later, in March 1933 (index value 3.5), the sentiment of ABC’s commentators was that “Spanish stock exchanges […] have become intoxicated by politics” (19 March 1933). In the same vein, during the electoral campaign of 1933 (index value 2.9) they wrote: “The stock exchange has an overly great concern for the electoral process […] and […] in spite of the occurrence of other events, they are neglected and the outlook of the elections remains preponderant […] This attitude is not surprising, because our generation never experienced such an intense political and social struggle” (29 October, 1933). While their views were not necessarily unbiased (ABC was a monarchic newspaper very hostile to left republicans and socialists, and its publication was from time to time suspended by the government), García-Uribe et al. (Reference García-Uribe, Mueller and Sanz2024) find that indices of socio-economic conflict based on news from ABC and La Vanguardia (a moderately pro-Republic newspaper published in Barcelona) are virtually identical and explain most of the economic policy uncertainty of the period.

5. Did political uncertainty predict future returns?

In this section, we test whether political uncertainty was correlated with future investment opportunities, proxied by aggregate equity market returns. We use in-sample predictive regressions for different time horizons in the following multivariate empirical specification:

(1a)$$r_{mkt, m + q} = \alpha _q + \beta _{mkt, q}^{POL} POL_m + \gamma _qX_m + \varepsilon _{mkt, m + q}$$

(1b)$$r_{mkt, m, m + q} = \alpha _q + \beta _{mkt, q}^{POL} POL_m + \gamma _qX_m + \varepsilon _{mkt, m, m + q}$$

where r _mkt,m+q is the market portfolio total return q months ahead, $$ r_{mkt, m, m + q} = r_{mkt, m+1} + \cdots + r_{mkt, m + q} $$ is the continuously compounded cumulative total return over q months, POL is our index of political uncertainty, X is a set of control state variables, q is the forecasting horizon (from 1 to q months) and $\varepsilon _{mkt, m + q}, \;$ $\varepsilon _{mkt, m, m + q}$ are forecasting errors with zero conditional mean. The coefficient of interest is $\beta _{mkt, q}^{POL} \;$ which indicates whether innovations in uncertainty predict positive or negative future returns.

In the benchmark regressions, the first set of control state variables includes three risk factors—size, value and momentum—in line with Fama–French models recently used also in the historical literature (Annaert and Mensah, Reference Annaert and Mensah2014; Ye and Turner, Reference Ye and Turner2014). The size factor is the 1-month ahead excess return of a portfolio of small over large stocks. The value factor is the excess return of a portfolio of high over low dividend yield stocks. The momentum factor is the excess return of a portfolio of winning over losing stocks in the last 6 monthsFootnote ¹⁴. To ensure that POL does not capture the impact of poor economic prospects or other sources of economic uncertainty not directly related to socio-political processes, we also control for three additional state variables. The first one is the aggregate dividend-to-price ratio, a broad indicator of business conditions which has been found to predict returns over various time horizons in different historical contexts robustly (Maio and Santa-Clara, Reference Maio and Santa-Clara2012; Golez and Koudijs, Reference Golez and Koudijs2018). We also include the government bond yield as a key indicator of the nominal interest rate in the money market, which is also found to have predictive ability over short-term time horizons (Rapach et al., Reference Rapach, Wohar and Rangvid2005; Ang and Bekaert, Reference Ang and Bekaert2007). For this purpose, we use the yield of internal perpetual bonds (deuda perpetua interior 4%), a safe financial instrument widely used as a source of liquidity at the discount window of the Bank of Spain. Finally, we control for the possible impact of market liquidity on future returns. Our measure of liquidity is the aggregate market turnover constructed on the base of daily data as the monthly total number of stocks traded divided by the number of stocks admitted to trade on the stock exchangeFootnote ¹⁵.

Table 1 summarises the benchmark estimates of the $\beta _{mkt, q}^{POL}$ coefficient both for cumulative and non-cumulative aggregate returns. As it is standard in the empirical finance literature, we use Newey and West standard errors with pre-whitened series up to q lags, which are robust to heteroscedasticity, auto and serial correlation in the residuals. We find that political uncertainty was a strong predictor of future market returns at short time horizons. One standard deviation increase in POL predicts a statistically significant and negative (non-cumulative) monthly return of −0.6/−0.7 per cent in the following three monthsFootnote ¹⁶, with a cumulative effect up to −2.1 per cent. This implies that peaks of political uncertainty (max value 4.36) predicted a cumulative loss up to −9.2 per cent 3 months ahead. At longer time horizons, the effect gradually wore out and was eventually reversed, with a predicted monthly return of 0.4 per cent and a cumulative return of 1.3 per cent 12 months aheadFootnote ¹⁷.

Table 1.

Predictive power of political uncertainty: benchmark results

The table reports $\beta _{mkt, q}^{POL}$ coefficients from a generalised linear model in which we regress monthly returns, first non-cumulative (top quadrant) then cumulative (bottom quadrant), of the capitalisation-weighted market portfolio up to 12 months ahead on our index of political uncertainty, for the period from January 1930 to July 1936. Controls include the size, value and momentum factors, the market dividend yield, the bond yield and an aggregate measure of market liquidity (turnover ratio). Newey–West standard errors are reported in parenthesis.

***^, ** and * denote statistical significance at 1, 5 and 10% levels, respectively.

As a second step in our analysis, we run various robustness checks to exclude that our measure of political uncertainty captures other confounding state variables. First, we include dummies for pre-election, election and post-election months to control for the possible impact of uncertainty generated by the municipal elections of April 1931 and the general elections of June 1931, November 1933 and February 1936. In the same vein, we include dummies for months that preceded, or coincided with cabinet crises. We also use dummies to capture months of political emergency, such as states of war and alarm declared in coincidence with internal crises such as outbursts of violence after the proclamation of the Republic (April–July 1931), the attempted military coup of August 1932 and the revolutionary insurrection of October 1934. In a second specification, we include three additional monthly indicators based on news from ABC: an index of news related to key macroeconomic policy issues, such as exchange rate, fiscal and monetary policy; an index of trade policy news and an index of news related to the possible break-up of Spain as a unitary state, such as the proclamation of an independent Catalan Republic (April 1931 and October 1934) and the parliamentary debate on the controversial statute of autonomy for Catalonia (finally approved in September 1932)Footnote ¹⁸. We also control for the possible impact of international political instability. We use the International Crises Behaviour (ICB) dataset to construct monthly measures of political disaster risk (Berkman et al., Reference Berkman, Jacobsen and Lee2011) based on the number of international political crises, both at a global level and in Europe, adjusted for characteristics such as the intensity of armed violence and the involvement of Great PowersFootnote ¹⁹. Finally, we add monthly variations in available macroeconomic indicators, such as GDP, wholesale prices, imports, exports and the exchange rate (the gold value of the peseta)Footnote ²⁰.

Table 2 summarises the results for cumulative returns at a time horizon of 3 months. They confirm the previous findings. A one-unit innovation in political uncertainty predicts a cumulative negative return between −1.5 and −2.3 per cent. Dummies for electoral campaigns also predict cumulative negative returns 3 months ahead up to −8.9 per cent. On the contrary, we do not find any significant impact of other economic policy or political news, internal states of exception or international political crisesFootnote ²¹. Overall, this evidence strongly corroborates the predictive power of the dimensions of domestic political uncertainty captured by our news-based indexFootnote ²².

Table 2.

Predictive power of political uncertainty: controls

The table reports $\beta _{mkt, q}^{POL}$ and other coefficients from a generalised linear model. We regress monthly cumulative returns of the capitalisation-weighted market portfolio 3 months ahead on our index of political uncertainty, dummies for election, pre-election, post-election, government crisis and pre-government crisis months, months of internal emergency (state of war and state of alarm), other news-based indices and international political crises for the period from January 1930 to July 1936. In all regressions controls include the size, value and momentum factors, the market dividend yield, the bond yield and a measure of market liquidity. In columns (2), (4) and (5), we also control for the monthly change of (log) GDP, wholesale prices, imports, exports and the peseta/gold exchange rate. Newey–West standard errors are reported in parenthesis.

***^, ** and * denote statistical significance at 1, 5 and 10% levels, respectively.

As a last step, we test the heterogeneity of the predicted impact of political uncertainty across firms. As suggested by an extensive literature, firms use political connections to extract private information, influence policy-making or demand protection (Faccio, Reference Faccio2006). These resources are especially valuable in periods of uncertainty, as they diffuse political risk and mitigate its negative effect—for instance, by allowing firms to access privileged information about policy changes (Liu et al., Reference Liu, Hu and Cheng2021). At the same time, political transitions could damage firms connected to the overturned regime (Dang and So, Reference Dang and So2018). In this perspective, firms connected to ruling parties of the Second Republic should have been less affected by shifts in political uncertainty, while firms connected to monarchic parties, or to political forces of the previous regimes, could be especially exposed to it.

To test those hypotheses formally, we reconstructed the biographies of the 467 directors sitting on the boards of our twenty-seven firms between January 1930 and July 1936. We identified fifty-five “politicians” (11.8 per cent of total), broadly defined as directors with political offices as MPs (elected and candidate), ministers or members of party elites during the Republic. We also found 110 “former politicians” connected to previous regimes (23.5 per cent of total), that is, MPs, senators, ministers or leaders of pre-1923 parties; and ministers, political officers and members of the unelected corporatist assembly (Asamblea Nacional Consultiva) during the dictatorship. Table 3 provides details about their political profilesFootnote ²³.

Table 3.

“Political” directors

The table summarises the profile of directors with political positions during the Second Republic (after April 1931) (top quadrant) and previous political regimes (Monarchy before September 1923, Dictatorship between September 1923 and January 1930, transition period January 1930–April 1931) (bottom quadrant). The narrow definition of “political directors” includes only MPs and ministers affiliated to political parties of the republican regime; the broad definition includes also local and national party leaders with no seats in Parliament or ministerial responsibilities. “New” connote directors with no political positions before April 1931; “old” those with pre-existing political experiences. “Directors appointed by the government” are official representatives in firms in which the state was a shareholder or had the legal power to appoint some directors; those without political connections were usually members of public administrations, including the army.

We measure at monthly frequency the intensity of each firm’s political connections as the share of seats occupied by “politicians” and “former politicians” and interact this measure with the political uncertainty index in a panel regression to test its impact on the cross section of future returnsFootnote ²⁴. Table 4 summarises the results for cumulative returns at a 3-month horizon. Firms with political connections to republican parties were significantly less affected; for a firm with a 10 per cent share of seats occupied by republican directors, the predicted fall from one-unit increase in political uncertainty was up to 41 per cent lower (−1.4 instead of −2.3 per cent) than a firm with no political connections. We do not find comparable systematic effects for other board characteristics, such as connections to other parties, foreign shareholders or to the state through representatives of public administrations. In turn, there is no evidence of a more negative impact from connections to monarchic parties or past regimes. The results hold if we control for time-varying firm characteristics (size, dividend yield, volatility) and other types of political or business connectionsFootnote ²⁵.

Table 4.

Predictive power of political uncertainty: political connections

The table reports results from predictive panel regressions for the cross section of firms. We regress monthly cumulative returns of individual firms 3 months ahead on our index of political uncertainty, and the interaction between uncertainty and a measure of political and business connections based on the share of board seats occupied by “political” directors, foreign directors and representatives of state administrations. Benchmark controls include the market, size, value and momentum factors, the market dividend yield, the bond yield, market liquidity and dummies for election, pre-election, post-election, government crisis and pre-government crisis months, for the whole period from January 1930 to July 1936. Stock time-varying characteristics include size, dividend yield, momentum, volatility and liquidity. Other political and business connections include personal connections to influential political figures (Francesc Cambó, Álvaro Figueroa Conde de Romanones) and financial groups (Banco Urquijo). Newey–West standard errors are reported in parenthesis.

***, ** and * denote statistical significance at 1, 5 and 10% levels, respectively.

It is important to add here that only a small minority of firms benefited from the stabilising properties of connections to republican parties. In fact, the presence of “republican” directors was heavily concentrated in three firms in which ruling parties had the legal power to appoint government representatives: the Bank of Spain, the oil company CAMPSA and the telephone company Telefónica—all firms with long-term special contracts with the government and monopolistic privileges in their respective sectorsFootnote ²⁶. We will further explore their role in the following sections.

6. Did sensitivity to political uncertainty vary across firms?

We now move on to study whether investors perceived political uncertainty as a source of non-diversifiable systematic risk. For each stock we regress its monthly return on the contemporaneous and lagged value of the political uncertainty index, POL, and a set of control factorsFootnote ²⁷. The sensitivity of each stock i to a one-unit increase in political uncertainty at month m is $( {\beta_{i, m}^{POL} + \beta_{i, m-1}^{POL} } )$, that is, the sum of its contemporaneous and lagged β ^POL from the following multivariate regression:

(2)$$R_{i, m} = \alpha _{i, m} + [ {\beta_{i, m}^{POL} \cdot POL_{i, m} + \beta_{i, m-1}^{POL} \cdot \;POL_{i, m-1}} ] + [ {\beta_{i, m}^X \cdot X_{i, m} + \beta_{i, m-1}^X \cdot \;X_{i, m-1}} ] + \varepsilon _{i, m}\;$$

where X is a vector of additional risk factors that could affect returns and β ^X measures the corresponding stock exposure. They include the market, size, value and momentum factors, and the aggregate market liquidity. Regressions for the entire period from January 1930 to July 1936 reveal a significant dispersion in the average β ^POL, ranging from +0.86 to −0.59%. However, the width of confidence intervals also suggests that estimates are not very precise in most casesFootnote ²⁸. This can be due to the fact that firms’ sensitivity changed over time—a plausible hypothesis given the volatility of the political environment.

To explore this issue further, we estimate time-varying exposures of our twenty-seven stocks using 24-month rolling windowsFootnote ²⁹. Figure 3 plots the time series of the median value of the estimated rolling sensitivities, and the upper and lower bounds of different quantiles of its cross-sectional distribution. The data show the presence of positive and negative β ^POL, with significant variations in their dispersion over time. This suggests that investors did not respond to heightened political uncertainty by exiting the market indiscriminately, but adjusted the composition of their equity portfolio. The range of extreme quantiles was significantly larger in the “bear” market of 1932–1934, confirming the hypothesis that sensitivity to uncertainty was especially strong in “bad” states of the financial market.

Figure 3.

Cross-sectional distribution of firm-specific sensitivities to political uncertainty, 1930–1936.

This graph shows for each month the median value of estimated time-varying sensitivities to political uncertainty of the twenty-seven firms included in the equity market portfolio, and the values corresponding to different quantiles of its cross-sectional distribution.

To assess which firms were perceived by investors as better and worse hedges against political uncertainty, Figure 4 shows for each stock the correlation between its median time-varying sensitivity and its percentage of positive β ^POL over the total number of rolling estimatesFootnote ³⁰. The upper right quadrant includes stocks with a positive average β ^POL and more than 60 per cent of positive estimates; the bottom left quadrant identifies stocks with a negative average β ^POL and more than 60 per cent of negative estimates. The median β ^POL measures the typical contemporaneous response of monthly returns to one unit increase in POL; for instance, in the two extreme cases of the insurance company Unión y Fénix and the chemical company Explosivos, it implies a 1.9 per cent increase and a 1.8 per cent decrease, respectively.

Figure 4.

Hedges vs. non-hedges, 1930–1936.

This graph shows for each firm included in the equity market portfolio the correlation between average time-varying sensitivity to political uncertainty (based on median value of rolling estimates) and the unconditional probability of positive estimates. Values for Banco Central are excluded since the stock was suspended from trading for long periods between 1931 and 1936.

Data suggest that, during months of heightened political uncertainty, investors bid up the prices of few selected firms. On the one hand, they were firms traditionally perceived as safe long-term investments, such as the Bank of Spain, a privately owned and highly profitable bank of issue with a sizable commercial business (Robledo, Reference Robledo1988) and Unión y Fénix, a French-Spanish company with a leading position in the cartelised and highly profitable insurance industry (Pons Pons, Reference Pons Pons2003). On the other hand, other “hedge” stocks were young firms created in the 1920s in emerging sectors, with strong market power and international connections. They included CHADE, a multinational electric public utility with a dominant position in Argentina, controlled by the Brussels-based holding SOFINA and with close political connections to Catalan nationalism (Cabana, Reference Cabana2015; Dalla-Corte Caballero, Reference Dalla-Corte Caballero2017); Telefónica, controlled by the U.S. multinational ITT, participated by large Spanish banks and with a monopolistic contract with the state granted by the dictatorship in 1924 (Álvaro Moya, Reference Álvaro Moya2007) and CEPSA, the only Spanish oil refining firm, with oil fields in Venezuela, modern refineries in the Canary Islands, access to supply from foreign oil multinationals and a privileged contract with CAMPSA, controlled by a consortium of Spanish banks, with a minority participation by the state and a monopoly contract for the commercialisation of oil products granted by the dictatorship (Tortella, Reference Tortella, Tortella, Ballestero and Díaz-Fernández2003; Tortella and Quiroga, Reference Tortella, Quiroga Valle and Quiroga Valle2022).

On the other hand, political uncertainty penalised firms in mature industries that had benefited from the extensive pro-business interventionism of the dictatorship, based on tariff protection, subsidies, public procurement contracts and ambitious plans of public works (Palafox, Reference Palafox1991, pp. 79–121; Carreras and Tafunell, Reference Carreras and Tafunell2004, pp. 242–251). These include, for instance, the cartelised iron and steel industry (Altos Hornos and Duro Felguera), the chemical giant Explosivos and the cartelised declining lead industry (Guindos). Similar cases are MZA and Norte, two railway giants heavily dependent on state subsidies under the regulatory framework (Estatuto Ferroviario) established by the dictatorship in the 1920s, which had prepared the ground for a more direct participation of the state in the management of the concessionary companies in view of a future nationalisation (Ortúñez Goicolea, Reference Ortúñez Goicolea2008, Reference Ortúñez Goicolea2016). On the contrary, we do not find a clear impact of political uncertainty on the publicly regulated electric sectorFootnote ³¹.

Importantly, Figure 4 also reveals that many firms returned almost equal numbers of positive and negative estimates. This suggests that investors’ perception of their exposure to political risk varied over time, possibly in response to changes in firm-specific political circumstancesFootnote ³². In the next section, we discuss this hypothesis for a group of firms which became the target of hostile political action by Republican governments.

7. Firms under political attack: how did investors respond?

In the initial phase of the Republic, some firms became especially exposed to attacks by the new political elites either for their conflicts with the executive (Banco de España), their privileged contracts with the state (Telefónica, CAMPSA, CEPSA) or their huge burden on the public finances (MZA, Norte). Their cases give us the opportunity to observe in more detail how shifts in firm-specific political risk affected investors’ perceptions of their hedging properties.

The politically more sensitive cases were the oil and telephone monopolies, stigmatised as the “brainchildren” of the dictatorship. In the case of Telefónica, the terms of the 1924 contract, initially accepted by the provisional government in April 1931, were pronounced illegal by a parliamentary commission and referred to the Parliament a few months later. In the middle of acrimonious political debates, secret negotiations and gestures of appeasement, uncertainties were gradually resolved over the course of 1934, when new amicable terms were agreed with the centre-right government of Alejandro Lerroux, leader of the Radical Party and a long-standing “friend” of the company (Álvaro Moya, Reference Alesina and Perotti2007, p. 80). In a similar fashion, the commercial oil monopoly granted to CAMPSA in 1927 had been questioned by republicans and socialists. Uncertainties were magnified by a legacy of unsettled legal controversies (domestic and international) on the expropriation of pre-existing private companies, an international boycott that made the company heavily dependent on Soviet oilFootnote ³³ and the competitive challenge of CEPSA, controlled by a group close to the old regime and interested in gaining direct access to the Spanish market. As a consequence, with the advent of the Republic, many expected a radical reform, or possibly the abolition of the monopoly. On the contrary, the centre-left government gave up initial plans to challenge the CAMPSA contract (which provided substantial fiscal revenues to the state) and turned into a strong supporter of the company, which then became a target of political attacks from the rightFootnote ³⁴. After the elections of 1933, rumours about a possible revision of the monopoly contract were revived, political interferences in the management of CAMPSA increased, and hostilities with CEPSA and its political supporters intensified, locking the two companies into a permanent dispute among political factions (Velarde, Reference Velarde Fuertes2010; Contreras-Pérez and Peña Díaz, Reference Contreras-Pérez, Peña Díaz and Quiroga Valle2022; Tortella and Quiroga, Reference Tortella, Quiroga Valle and Quiroga Valle2022).

In the case of the Bank of Spain, the 1921 reform of the banking law, which aimed at converting it into a modern central bank (Pons Brias, Reference Pons Brias2022), created scope for policy conflicts on the management of the exchange rate, which was formally entrusted to the Ministry of Finance. The final period of the dictatorship and the 1930–1931 transition were dominated by a permanent dispute on the use of gold reserves (which the Bank and its private shareholders regarded as its main asset) for exchange market interventions. In the crisis of 1931, the Bank intervened reluctantly in support of the banking system and most of its directors dissented publicly with the government’s stabilisation plans. The conflict precipitated a reform of the banking law that increased the representation of the executive on the board of directors and the share of Bank’s profits accruing to the state. The project, loudly resisted by the Bank and its political supporters, was passed in November 1931 in spite of harsh public controversies in the press and the Constituent Assembly. However, the manifest hostility of banking elites cost Indalecio Prieto, a socialist leader, his position at the head of the Ministry, replaced by moderate republican Jaime Carner in December 1931 (Martín-Aceña, Reference Martín-Aceña1984, pp. 244–251; Velarde, Reference Velarde Fuertes2010; Martínez Ruiz and Nogués-Marco, Reference Martínez Ruiz and Nogués-Marco2014; Jorge-Sotelo, Reference Jorge-Sotelo2020; Martínez Ruiz, Reference Martínez Ruiz2022).

Railways were another prominent example of the shift in policy risk brought home by the regime change. The regulatory regime approved in 1924 by the dictatorship (Estatuto Ferroviario) had led to a semi-nationalisation with high costs for public finances and large benefits for private shareholders of the agonising giants MZA and Norte. Its formal suspension by an interim government in July 1930 pushed the sector into a legal, regulatory and financial limbo and left to republican governments and political forces a burdensome legacy of permanent (and inconclusive) negotiations on the objectives and costs of a new reform. Until 1933, while companies struggled to guarantee shareholders’ compensation in case of nationalisation, the centre-left coalition reinforced government’s influence on firms’ management, both indirectly (through the Consejo Superior de Ferrocarriles) and directly (through state commissioners with the power to veto any deliberation of the directors), in view of a future nationalisation at no cost for the state budget. In 1934–1935, on the contrary, centre-right governments, focused on alleviating the fiscal burden of the existing regime, approved an increase in tariffs and elaborated reform projects more in tune with the requests of the companies, pointing to a favourable redemption of concessions, a rationalisation of the sector, a reorganisation of private companies and a new lease regime. However, the hypothesis of a nationalisation at zero cost resurged in 1936 with the seizure of the defaulting third largest railway company (Andaluces) and the proposal of a new co-administration regime, fiercely opposed by companies (Ortúñez Goicolea, Reference Ortúñez Goicolea2000, Reference Ortúñez Goicolea2008).

Did investors respond to shifts in firm-specific political risk? We find this is actually the case. In the bottom part of Figures 5–8 we show the rolling estimates of sensitivity to political uncertainty for the selected firms. The top part of the graphs also shows sensitivity shocks, obtained as residuals from a regression of the estimated sensitivity on fixed firm and month effects (to get rid of the impact of unobservable time-invariant firm characteristics and common shocks) and a lag of the dependent variable (to account for the persistence of rolling estimates).

Figure 5.

Time-varying sensitivity: the banking sector.

Time-varying sensitivity is based on rolling estimates of β ^POL on a 24-month moving window. Shocks to sensitivity are the residuals from a pooled regression of the estimated time-varying sensitivities on a constant, firm- and month-fixed effects, and a lag of sensitivity to account for its autoregressive component.

Figure 6.

Time-varying sensitivity: Telefónica.

Time-varying sensitivity is based on rolling estimates of β ^POL on a 24-month moving window. Shocks to sensitivity are the residuals from a pooled regression of the estimated time-varying sensitivities on a constant, firm- and month-fixed effects, and a lag of sensitivity to account for its autoregressive component.

Figure 7.

Time-varying sensitivity: oil companies.

Time-varying sensitivity is based on rolling estimates of β ^POL on a 24-month moving window. Shocks to sensitivity are the residuals from a pooled regression of the estimated time-varying sensitivities on a constant, firm- and month-fixed effects, and a lag of sensitivity to account for its autoregressive component.

Figure 8.

Time-varying sensitivity: railway companies.

Time-varying sensitivity is based on rolling estimates of β ^POL on a 24-month moving window. Shocks to sensitivity are the residuals from a pooled regression of the estimated time-varying sensitivities on a constant, firm- and month-fixed effects, and a lag of sensitivity to account for its autoregressive component.

In the regime transition of 1930–1931, the hedge properties of the Bank of Spain’s stocks were significantly impaired by its conflicts with the Ministry of Finance, which exposed it to political actions that could limit its autonomy and profitability (see Figure 5). On the contrary, its β ^POL became strongly positive after January 1932; this suggests that investors interpreted the political changes of the end of 1931 as a significant reduction in its exposure to political risk. The appeasement with the Bank also had a positive spillover on the rest of the banking system. In fact, as shown in the same figure, we observe in 1932 a similar switch to positive β ^POL also for private banks (Banco Español de Crédito, Banco Hispano Americano)Footnote ³⁵. This coincided with important regulatory changes, negotiated behind closed doors by banks and the Ministry of Finance. By temporarily altering accounting rules, the agreement protected banks’ official balance sheets from the consequences of the generalised slump of stock market prices accumulated since the advent of the Republic. This intervention not only protected banks’ stability in the short run, but also reduced counterparty risk for the Bank of Spain’s loans against collateralised securities (Jorge Sotelo, Reference Jorge-Sotelo2019, pp. 231–248)Footnote ³⁶. Therefore the political appeasement was credibly signalled by a pact on regulatory forbearance that favoured the recovery of public trust in the banking system as a wholeFootnote ³⁷.

Similarly, Telefónica stocks switched from positive to negative β ^POL when the terms of its contract went under attack, but switched back to positive β ^POL as the appeasement with the “friendly” Radical party was negotiated and finally agreed (Figure 6). The β ^POL of CAMPSA and CEPSA moved along opposed and crossing paths until 1933 in response to shifts in their specific (and negatively correlated) political circumstances, but then moved jointly into negative β ^POL when the struggle of political factions exacerbated interferences and conflicts (Figure 7). In the case of the railway sector, on the contrary, the policy deadlock on the Estatuto Ferroviario made investors permanently wary of the hedge properties of MZA and Norte stocks, even when governments seemed to abandon the implicit threat of a nationalisation without compensation and moved towards a reform more favourable to the interests of the companies and their shareholders (Figure 8)Footnote ³⁸.

Overall, the four cases discussed here confirm that our measure of time-varying sensitivity to political uncertainty responded systematically to shifts in firm-specific exposure to political risk. In general, we find that changes in firms’ exposure to political risk were quickly reflected in shifts in investors’ perception of their hedge properties in periods of heightened political uncertainty.

8. Did hedges earn lower future excess returns?

As explained in Section 2, asset pricing theory predicts that safer “hedge” stocks should earn lower future excess returns compared with riskier stocks with negative sensitivity to political uncertainty. We test this prediction formally by using a portfolio-sorting approach. At the end of each month, m, we rank stocks by their time-varying β ^POL estimated in the past 24 months, and sort them into tertile portfolios (each composed of nine stocks). The first (third) tertile includes stocks with low (high) β ^POL. Then, we calculate the return of each portfolio, both equally and value weighted, for holding periods from m + 1 to m + 12Footnote ³⁹. This procedure is repeated each month until the end of the sample. Finally, we estimate the average expected performance of the first and third tertile portfolios and that of the hedge portfolio—that is, a portfolio long on high β ^POL stocks and short on low β ^POL stocks.

Table 5 summarises the average characteristics of stocks included in the tertile portfolios. In comparison with stocks in the first tertile (low sensitivity), stocks included in the third tertile account on average for a slightly larger share of the overall market portfolio and show higher β ^POL (by construction), higher market capitalisation and lower negative momentum. On the contrary, we cannot observe any statistically significant differences in the mean of other stock characteristics, such as dividend yield, liquidity or sensitivity to market risk. The fact that high sensitivity stocks do not differ systematically from low sensitivity stocks in terms of liquidity excludes the possibility that hedging was in fact a “flight to liquidity”. Looking at the mean cumulative returns of both equally weighted and capitalisation weighted tertile portfolios at different time horizons, we also find that high β ^POL stocks performed significantly better than median and low sensitivity stocks at any time horizon until 12 months, and show lower volatility and downside risk for holding periods until 6 monthsFootnote ⁴⁰.

Table 5.

Average characteristics of stocks included in the tertile portfolios

The top section of this table shows the mean values of β ^POL, market beta, size (market capitalisation), dividend yield and momentum of stocks included in the tertile portfolios sorted by their sensitivity to political uncertainty. The bottom section tests the statistical significance of the differences between tertile 3 portfolio (high β ^POL) and the other two. Columns (a) refer to equally weighted tertile portfolios, columns (b) to market-capitalisation weighted tertile portfolios.

***^, ** and * connote significance at the 1.5 and 10% levels, respectively.

More formally, we estimate the risk-adjusted 1-month ahead excess return of tertile portfolios over government bonds using a four-factor model, which includes the contemporaneous and lagged market factor and the size, value and momentum risk factors. Table 6 shows the unadjusted excess return of value weighted high and low β ^POL portfolios and the hedge portfolio, their average α (i.e. the risk-adjusted excess return) and the loadings of each risk factorFootnote ⁴¹. We find that the risk-adjusted returns for the high and low β ^POL tertiles are 0.4 and −0.6% in the CAPM model and 0.3 and −0.8% in the four-factor model, with risk-adjusted spreads of 1.0 and 1.1% respectively (all coefficients are statistically significant at the 5 or 1% levels). These results confirm that the risk-adjusted spreads are driven both by the underperformance of low β ^POL portfolios and the outperformance of high β ^POL portfoliosFootnote ⁴².

Table 6.

Tertile portfolios’ risk-adjusted performance, 1-month holding period

This table shows unadjusted and risk factor-adjusted returns of capitalisation weighted tertile portfolios with high and low sensitivity to political uncertainty, and their hedge spreads, for a 1-month holding period. Newey–West standard errors are reported in parenthesis.

***^, ** and * denote statistical significance at 1.5 and 10% levels, respectively.

The finding of a positive uncertainty premium runs against the theoretical predictions of asset pricing. Our interpretation is that, in a situation of persistently high political uncertainty, investors went systematically long on hedges and short on firms more exposed to political risk. This hypothesis is supported by Figure 9, which shows the correlation between the median β ^POL of individual stocks and their price in December 1935 (the peak of the bull market of 1934–1935) relative to October 1931 (the bottom of the market slump that followed the proclamation of the Republic in April).

Figure 9.

Sensitivity to political uncertainty and long-term price performance.

This figure shows the correlation between the median value of time-varying sensitivity of individual stocks and their price in December 1935 relative to October 1931.

We observe that by the end of 1935 the price of most hedge stocks (Banco de España, Unión y Fénix, CEPSA, Telefónica) had increased between 20 and 100 per cent relative to 1931; on the contrary, the value of most stocks with typical negative β ^POL was still a fraction (in some cases, a very small one) of its 1931 levelFootnote ⁴³. This evidence supports the hypothesis of a structural (rather than transitory) adjustment in the composition of the average investor’s portfolio towards firms perceived as less exposed to political risk and against firms with negative sensitivity to uncertainty, whose prices recorded a long-term decline over the periodFootnote ⁴⁴. This also suggests that the protective “put” option offered by public authorities to holders of negative β ^POL stocks, through price floors and trading restrictions, possibly slowed down but did not prevent the adjustment of risk-averse investors’ preferences.