Do political systems have a lasting effect on climate change concern? Evidence from Germany after reunification

I use data from the German Socioeconomic Panel (SOEP) (2019), a nationally representative longitudinal survey of private households, collecting information on demographic, attitudinal, social and economic variables from nearly 15 000 households and over 25 000 individuals annually since 1984. Until 1989 data were collected exclusively among residents of the Federal Republic of Germany (FRG). The survey was extended to include residents of the former GDR in 1990, allowing comparisons between individuals who had lived under two opposing political systems. Since 2009, SOEP elicits participants' concern for the consequences of climate change, asking: 'How worried are you about the consequences of climate change?'. Available responses are on a three-point scale: '1. Very concerned', '2. Somewhat concerned', and '3. Not at all concerned'. The outcome of interest is a binary variable, equal to 1 for participants that respond 'Very concerned' to the climate worry question. In robustness tests I extend the definition of climate concern to include individuals responding 'Somewhat concerned'. Figure B1 in the appendix shows the distribution of climate concern as elicited by SOEP.

To construct the main independent variable, I use information on participants' country of residence before reunification. SOEP participants are asked: 'Where did you live in 1989?'. Available responses are '1. East Germany including East Berlin' , '2. West Germany including West Berlin' and '3. Abroad'. The treatment variable East, indicates individuals living in East Germany in 1989 (Alesina and Fuchs-Schundeln 2007, Heineck and Sussmuth 2013, Fuchs-Schundeln and Haliassos 2020). It is intended to capture the effect of assignment to the group exposed to the totalitarian regime of the GDR because of their location at the time of the division or their location of birth, on contemporary climate concern, relative to the group that did not experience life under the GDR. For the remainder of this letter, I use the term 'East (West) German' to refer to individuals that lived in the GDR (FRG) in 1989. Individuals living abroad in 1989 are removed from the analysis. Figure B2 shows the evolution of East and West German's climate concern over time.

I account for demographic and economic variables that can affect climate change attitudes (Franzen and Meyer 2010, Franzen and Vogl 2013, Poortinga et al 2019). Models control for employment status, sex, the presence of children under 16 in the household, whether the respondent lived in a large, medium or small city, or the countryside at age 16, and a quadratic polynomial in age to capture variation in climate concern that is due to age differences. Finally, I include education years, and a quadratic polynomial of the natural logarithm of household income in constant 2016 Euros.

I restrict the sample to SOEP participants who were at least 12 years old at the time of the reunification of Germany in 1990. I remove individuals who were interviewed in SOEP's expansion subsamples that were exclusively targeted at immigrant and refugee households. Table A1 in the appendix shows descriptive statistics of the variables used in the analysis.

Let the relationship between climate concern of individual i living in country c, Y_ic , and the country's political system PS_c , be described by:

$$\begin{equation} Y_{ic} = a_{c}+b\times PS_{c}+X_{ic}+\epsilon_{ic}. \end{equation} \tag{ 1 }$$

Using individual level data and variation in political systems prevailing across countries to assess this relationship will lead to biased estimates of b. Individuals sort across countries according to unobserved characteristics that can simultaneously determine their climate concern. To address biases raised by the endogeneity of the political system, I use the natural experiment created by the division of Germany. I test whether differences in individual experience of political systems, determined by the division of Germany in 1949 have a lasting effect on contemporary climate change attitudes. As East and West Germany faced the same political and economic system before the division and after reunification, comparing East and West Germans after reunification offers insights into the influence of life in the GDR on climate change attitudes (Alesina and Fuchs-Schundeln 2007). The division and reunification of Germany has been used to estimate the impact of political systems in general, and communism in particular, on individual preferences (Alesina and Fuchs-Schundeln 2007, Friehe and Pannenberg 2020), attitudes (Rainer and Siedler 2009a, van Hoorn and Maseland 2010, Brosig-Koch et al 2011, Bauernschuster and Rainer 2012, Heineck and Sussmuth 2013, Carl 2018, Lippmann et al 2020) and economic outcomes (Fuchs-Schundeln 2008, Rainer and Siedler 2009b, Friehe and Mechtel 2014, Fuchs-Schundeln and Masella 2016, Ong and Theseira 2016, Beblo and Gorges 2018, Campa and Serafinelli 2018, Lippmann and Senik 2018).

The post-WWII partition of Germany was agreed in September 1945 between the Western Allies and the Soviet Union following a long negotiation (Fuchs-Schundeln and Schundeln 2005, Redding and Sturm 2008, Becker et al 2020). The part of Germany that was not annexed by the Soviet Union or Poland was divided into American, British and Soviet occupation zones, while a French zone was created later from areas under American and British control (Redding and Sturm 2008). Berlin was jointly occupied by Western and Soviet forces. Following growing tensions between the Western Allies and the Soviet Union, the GDR was founded in 1949 comprising territories under Soviet control, adopting a communist model of governance. The regions under allied control formed the FRG, adopting liberal democracy and market economy institutions. In the early years of the partition, transit and trade between the two States were possible. By 1952 the border was sealed, breaking the links between the West and East German States, and movement across the border became nearly impossible after 1961 (Burchardi and Hassan 2013). After four decades of division that appeared to be permanent, discontent in East Germany in 1989 led to widespread protests that resulted in the unexpected fall of the Berlin Wall and the reunification of Germany less than one year later (Alesina and Fuchs-Schundeln 2007, Fuchs-Schundeln 2008, Prantl and Spitz-Oener 2019).

There are reasons to expect that features of the GDR's political system, including State ideology, the centrally planned economy, limits to information dissemination, and lack of electoral accountability, affected environmental conditions and the public's future climate attitudes ¹ . State ideology did not acknowledge that environmental problems could arise under communism, as these were perceived to be a symptom of capitalist production (Jones 1993). While the GDR introduced ambitious environmental policies, these were not monitored or enforced as industrial growth objectives systematically outweighed environmental protection in decision making (Jones 1993, Dominick 1998). Planners and politicians faced no electoral competition in the single-party GDR and had no incentive to adopt environmentally friendly policies (Dominick 1998). At the same time, the centrally planned economy did not provide sufficient incentives for developing less environmentally damaging production methods and was slow to respond to the increasing environmental degradation (Jones 1993, Dominick 1998, Rink 2002). Finally, the GDR's information policy restricted the dissemination of environmental facts and opinions, shielding decision making from public scrutiny (Rink 2002). Independent environmental groups began forming in the early 1980s in response to worsening environmental problems. These groups were tolerated by the State as their influence was limited, but were nevertheless monitored and faced interference (Jones 1993, Rink 2002).

To assess whether the GDR's political system has a lasting effect on individual climate concern, I estimate panel random effects models:

$$\begin{equation} Y_{it} = \alpha_{s}+T_{t}+\beta_{1}East_{i}+\beta_{2}X_{it}+\epsilon_{it} \end{equation} \tag{ 2 }$$

where Y_it is the concern indicator for respondent i in year t. The variable of interest East_i is binary, indicating individuals who lived in the GDR prior to reunification. Coefficient β₁ captures the difference in the probability of declaring climate concern between those who were exposed to the totalitarian system of the GDR against to those who were not. α_s are German State (Länder) indicators controlling for state-specific time-invariant characteristics including local culture, natural resource endowments, climate and topography that may affect climate concern. T_t is a set of year-specific effects, absorbing variation common across all SOEP participants within a given year. Matrix X contains the set of individual level demographic variables described earlier. Finally, _it is an error term. I report standard errors clustered at the individual SOEP participant level. Marginal effects from probit and pooled OLS models are shown in the appendix. In all cases estimates are identical in terms of magnitude and statistical significance.

The empirical strategy requires that the partition of Germany exogenously divided a homogeneous population. The GDR-FRG border was drawn by the Western Allies and the Soviet Union primarily according to military considerations and it is unlikely that the local population, environmental conditions or future climate concerns had much influence in the delineation (Alesina and Fuchs-Schundeln 2007, Redding and Sturm 2008, Becker et al 2020). A threat to identification comes from pre-existing differences in economic and cultural characteristics as the border followed existing administrative limits (Becker et al 2020). There is mixed evidence regarding the pre-division homogeneity of economic characteristics between regions that formed the two States. On the one hand, evidence suggests that the German economy was well integrated by the end of the Weimar Republic (Wolf 2009), and there were no systematic differences in per capita income across German regions (Alesina and Fuchs-Schundeln 2007, Redding and Sturm 2008, Friehe and Mechtel 2014, Lippmann et al 2020). On the other hand, Becker et al (2020) find differences in self-employment, female employment and the share of blue-collar employment between East and West regions. Results reported here could reflect these differences, to the extent that these economic characteristics determine climate concerns and are not captured by individual level controls or the State dummies. There is also evidence that the East was significantly more left wing in the years before the division (Becker et al 2020). Nevertheless, as left wing views tend to be positively correlated with climate concern this likely attenuates estimates reported in this paper ² . Since climate change concerns did not exist in the early half of the 20th century, there is no risk that predivision differences in climate change attitudes are driving the results.

A further threat to identification arises from the differential treatment of East and West Germany from the occupying forces (Becker et al 2020). War reparations to the Soviet Union depleted East Germany's industrial capital. This may have led to over-reliance on the country's natural resource endowment and delayed technological development, eventually affecting climate attitudes. Nevertheless, Dominick (1998) suggests that West and East German economies were similar in their willingness to tradeoff environmental quality for economic growth and in their effort to maximize industrial output growth in the first decades following the division, while both were reliant on coal resources (Renn and Marshall 2016). As a result many environmental problems were as severe and in few cases worse in the West (Dominick 1998). Differences in the treatment of the environment started to appear in the early 1970s as the centrally planned economy did not provide sufficient incentives for innovation, and as planners and politicians did not face electoral scrutiny over their environmental policy (Jones 1993, Dominick 1998, Rink 2002). With regards to pre-division differences in environmental attitudes, it appears that by the early 20th century environmental conservation efforts were developed across all German States, later consolidated under the 'Reich Nature Protection Act' (Ditt and Rafferty 1996). Furthermore, Niedertscheider et al (2014) show similar forest cover between East and West Germany in 1951 just after the division.

To further isolate the influence of the political system from regional characteristics, I focus on the climate concern of individuals who crossed the inner-German border after reunification (Brosig-Koch et al 2011, Becker et al 2020). Specifically, I estimate models comparing between (i) East (West) and West (East) Germans currently in the West (East), and (ii) East (West) Germans currently in the West (East) against East (West) Germans currently in the East (West). To assess the main result's stability and examine whether the relationship between GDR experience and climate concern is driven by region-specific characteristics, I also estimate equation (2) replacing East with an exhaustive set of indicators for each of the States that comprised the GDR. If GDR-wide institutional features instead of regional characteristics drive differences in climate attitudes between West and East Germans, estimates of the GDR State indicators should be similar. To account for possible biases introduced by the self-selection of Berlin's residents (Bursztyn and Cantoni 2015, Becker et al 2020), I test the results' robustness when removing them from the analysis.

Data on climate concern were collected between 2009 and 2018, raising the possibility that part of the estimated effect may be due to divergence between East and West German regions after 1990 that is not captured by the State dummies. To assess this, I examine the relationship between GDR experience and environmental concern. SOEP collected environmental concern data since 1984 in West Germany and since 1990 in East Germany, allowing me to track environmental concern since immediately after reunification. Although not identical, environmental and climate concern may be driven by similar mechanisms (Helm et al 2018), permitting the use of general environmental concern as a proxy for climate concern. I also show estimates from equation (2) when introducing year-State effects, and when controlling for the regional unemployment rate to account for the influence of time-varying regional characteristics on climate concern.

I examine whether the influence of the GDR's political institutions varies with the duration of the experience, replacing the East indicator with a continuous Years in East variable, measuring the years a respondent has lived in the GDR (Carl 2018, Becker et al 2020). To account for non-linearity in the influence of GDR experience, I also show estimates when Years in East enters models in a quadratic polynomial and flexibly with dummies.

To assess the evolution of climate concern over time for East and West Germans, I estimate:

$$\begin{align} Y_{it}& = \alpha_{s}+I_{it}+\beta_{1}East_{i}+\sum_{r = 2}^{5}\gamma_{r}R_{r}+\sum_{r = 2}^{5}\delta_{r}East_{i}\nonumber\\&\quad\times R_{r}+\beta_{2}X_{it}+\epsilon_{it} \end{align} \tag{ 3 }$$

where R_r are biannual indicators, with R₂ = 1 for years 2011–2012, and R₅ = 1 for years 2017–2018. R₁ = 1 for years 2009–2010 is the omitted category. Coefficient β₁ shows the difference between East and West German's climate concern in years 2009 and 2010. Coefficients γ_r show the change in West Germans' climate concern in period r, relative to 2009–2010. Coefficients on the $East_{i}\times R_{r}$ interactions are interpreted as measures of convergence between East and West Germans' attitudes.

To account for heterogeneity in the effect of East across individuals with different migration opportunities and experiences with the GDR's environmental policy, I introduce a set of birth cohort dummies and their interactions with the treatment indicator (Fuchs-Schundeln 2008, Becker et al 2020) and estimate:

$$\begin{align} Y_{it}& = \alpha_{s}+T_{t}+\beta_{1}East_{i}+\sum_{r = 2}^{3}\zeta_{r}C_{ir}+\sum_{r = 2}^{3}\eta_{r}East_{i}\nonumber\\&\quad\times C_{ir}+\beta_{2}X_{it}+\epsilon_{ist} \end{align} \tag{ 4 }$$

where the cohort indicators C_ir capture differences in climate concerns across birth cohorts of former West Germans. C_i2 = 1 for cohorts 1949 to 1960 and C_i3 = 1 for individuals born on or after 1961. Birth-year cohorts before 1948 are the omitted category. Coefficient β₁ captures the difference in climate concern between East and West Germans born before 1949. Coefficients ζ describe differences in concern across West Germans' birth cohorts, while coefficients η show the differences across East German birth-year cohorts.

Table 1 shows estimates from equation (2), assessing the effect of exposure to the GDR's political system on individual climate concern, using SOEP data from 2009 to 2018. The estimate in column 1 comes from regressing climate concern only on the treatment indicator East, while models in the remaining columns progressively introduce year dummies, German State dummies and individual-level demographic controls. The GDR's influence on climate concern is negative and statistically significant, while its magnitude is stable irrespective of the set of covariates. The specification adjusting for the full set of controls (column 3) suggests that experience with the GDR's political system lowers the probability of climate concern by 4.5 percentage points. Table A2 in the appendix shows the full set of estimates. Table A3 shows results from random effects probit and pooled OLS models. In all cases results are identical in magnitude and statistical significance.

Table 1. Climate concern and experience of life in the GDR.

	(1)	(2)	(3)
East	$-0.067^{\ast\ast\ast}$	$-0.043^{\ast\ast\ast}$	$-0.045^{\ast\ast\ast}$
	(0.005)	(0.009)	(0.009)
State	No	Yes	Yes
Year	No	Yes	Yes
Controls	No	No	Yes
Observations	142 274	142 274	142 274

Note: Data from the German Socioeconomic Panel Survey from 2009 to 2018. Each column presents estimates from a different panel random effects model. The dependent variable is binary, equalling 1 for SOEP participants declaring 'Very worried about the consequences of climate change'. The main independent variable 'East' is binary, indicating participants who lived in the GDR before the reunification of Germany. Sociodemographic controls are: a set of employment status indicators; a binary variable indicating women; a set of indicators controlling for urbanization of respondents'place of residence until age 16; education years; age and age²; ln(HouseholdIncome) and ln(HouseholdIncome)². Standard errors, clustered at the individual respondent level in parentheses. $^{\ast}~p\lt0.1$, $^{\ast\ast}~p\lt0.05$, $^{\ast\ast\ast}~p\lt0.01.$

Table 2 shows estimates from models focusing on the climate attitudes of individuals that moved across the former inner-German border after reunification. In columns 1 and 2 the main independent variable East equals 1 for East Germans that have since reunification moved West. In column 1, I restrict the sample to individuals living in West German States, and compare East Germans who moved West, against West Germans (I exclude Berlin residents from the analysis). The estimate suggests that the difference in East and West Germans' climate concern persists among individuals living in the West: GDR experience lowers the probability of climate concern by 4.7 percentage points. Self-selection may attenuate the estimated effect of GDR experience, as the group of East Germans in the West presumably chose to move because they were closer to attitudes prevailing among West Germans. The model in column 2 compares East Germans in the West, against East Germans in the East. The estimated effect on climate concern is positive but statistically insignificant. The positive sign can be attributed to the selection of East Germans who moved West or signal convergence between East and West attitudes. Columns 3 and 4 repeat the analysis, this time defining a West binary variable, equaling 1 for West Germans in the East. West Germans in the East are 6.2 percentage points more likely to declare climate concern, relative to East Germans in the East (column 3). Finally, West Germans in the East are more likely to be concerned about the consequences of climate change relative to West Germans in the West, but the estimate is not statistically significant. However, results referring to West Germans in the East are based on a small number of cases.

Table 2. Climate concern and experience of life in the GDR.

	East Germans in West	East Germans in West	West Germans in East	West Germans in East
	vs	vs	vs	vs
	West Germans in West	East Germans in East	East Germans in East	West Germans in West
	(1)	(2)	(3)	(4)
East	$-0.047^{\ast\ast\ast}$	0.016
	(0.012)	(0.03)
West			$0.062^{\ast\ast\ast}$	0.089
			(0.024)	(0.071)
Observations	102 372	37 513	34 314	99 113

Note: Data from the German Socioeconomic Panel Survey from 2009 to 2018. Each result comes from a different panel random effects model. The dependent is binary, equalling 1 for SOEP participants declaring Very concerned about the consequences of climate change. The independent variable 'East' (West) is binary indicating East (West) Germans who on the year of the survey were found in the West (East). The first column compares East against West Germans in West Germany. The second column compares East Germans in the West against East Germans in the East. All models include State and Year dummies and sociodemographic controls. The latter are: a set of employment status indicators; a binary variable indicating women; a set of indicators controlling for urbanization of respondents' place of residence until age 16; education years; age and age squared; ln(HouseholdIncome) and ln(HouseholdIncome)². Standard errors, clustered at the individual respondent level are reported in parentheses. $^{\ast}~p\lt0.1$, $^{\ast\ast}~p\lt0.05$, $^{\ast\ast\ast}~p\lt0.01$.

Table 3 presents a series of estimates to evaluate the main result's stability. The model in column 1 removes individuals currently living in Berlin from the sample. The effect is similar to the baseline result in magnitude and statistical significance. Column 2 expands the definition of the dependent variable to include SOEP participants that are 'Somewhat Concerned' about the consequences of climate change. The result is statistically significant and of comparable magnitude to the baseline. Column 3 compares current residents of the former GDR and former FRG States, replacing East with an exhaustive set of GDR-State indicators. All coefficients on the GDR State dummies are negative, large and statistically significant. Residing in one of the former GDR regions lowers probability of climate concern from 4.4 percentage points in Thuringia, to 9.8 percentage points in Saxony. The null of equality across all coefficients is rejected (p = 0.000). However, testing does not reject equality across the effects of Brandenburg, Mecklenburg-West Pomerania, Saxony-Anhalt and Thuringia (p = 0.103). It appears then that the impact of living in Saxony stands out from the rest of the estimates. This can be plausibly attributed to the remoteness of the region to the FRG, reducing residents access to information from western media in the years of the GDR (Bursztyn and Cantoni 2015, Hennighausen 2015, Friehe et al 2020). Models in columns 4 and 5 account for the influence of time-varying regional economic conditions on individual climate concern, respectively adding State-Year effects and the rate of the regional unemployment in the set of controls. The estimated effect at 4.5 percentage points is identical to the main result. Finally, column 6 shows estimates from equation (2) when the outcome is a binary 'Environmental Concern' variable using SOEP data from 1990 to 2018: experience with the GDR lowers environmental concern by 1.8 percentage points.

Table 3. Robustness tests.

	(1)	(2)	(3)	(4)	(5)	(6)
East	$-0.052^{\ast\ast\ast}$	$-0.027^{\ast\ast\ast}$		$-0.045^{\ast\ast\ast}$	$-0.045^{\ast\ast\ast}$	$-0.018^{\ast\ast}$
	(0.010)	(0.008)		(0.009)	(0.009)	(0.008)
Brandenburg			$-0.065^{\ast\ast\ast}$
			(0.009)
Mecklenburg-Vorpommern			$-0.074^{\ast\ast\ast}$
			(0.012)
Saxony			$-0.098^{\ast\ast\ast}$
			(0.008)
Saxony-Anhalt			$-0.074^{\ast\ast\ast}$
			(0.010)
Thuringia			$-0.044^{\ast\ast\ast}$
			(0.010)
Observations	136 686	142 274	136 686	142 274	142 274	365 770

Note: Models in columns 1 to 5, use data from the German Socioeconomic Panel Survey for years 2009 to 2018. Each result comes from a different panel random effects model. The dependent is binary, equalling 1 for SOEP participants declaring 'Very concerned about the consequences of climate change'. Column 1 removes Berlin residents from the sample. Column 2 extends the definition of concern to include those that are somewhat concerned. Column 3 replaces the East dummy with a series of Region dummies indicating residence in one of the former GDR States- the omitted category is States of the Former FRG. Column 4 adds State-Year controls. Column 5 adds an unemployment control variable. For the model in column 6, data come from the German Socioeconomic Panel Survey for years 1990 to 2018. The dependent variable is binary, equalling 1 for SOEP participants declaring 'Very concerned about environmental protection'. All models include State and Year dummies and sociodemographic controls. The latter are: a set of employment status indicators; a binary variable indicating women; a set of indicators controlling for urbanization of respondents' place of residence until age 16; education years; age and age squared; ln(HouseholdIncome) and ln(HouseholdIncome)². Standard errors, clustered at the individual respondent level are reported in parentheses. $^{\ast}~p\lt0.1$, $^{\ast\ast}~p\lt0.05$, $^{\ast\ast\ast}~p\lt0.01$.

Table 4 presents estimates from equation (2) when replacing the treatment indicator East with a continuous Years in East variable measuring the years (in decades) respondents lived under the GDR's regime. Imposing a linear relationship between climate concern and duration of GDR experience suggests that a decade of life in the GDR decreases the probability of climate concern by 0.6 percentage points (column 1). Assuming a quadratic relationship implies a negative effect that decreases as the length of the experience increases (column 2). Allowing the duration of GDR experience to enter flexibly in column 3, suggests that the probability of climate concern decreases for every interval, relative to no GDR experience, while the null of equality across all coefficients is rejected (p = 0.000). The impact of having lived 40+ years in the GDR appears lower relative to the other experience intervals, plausibly reflecting the evolution of the GDR's environmental policy over time. Recent research suggests individual climate attitudes are related to country-level environmental protection (Levi 2021). As described earlier, environmental protection was a nominal policy objective during the early years of the GDR, but the interest in conservation waned over time as the State continued to focus on expanding industrial output (Jones 1993, Dominick 1998, Rink 2002). The group of individuals with 40+ years of life in the GDR includes those who may retain some memory of early GDR environmental aspirations and policy targets, and their individual climate attitudes may reflect them.

Table 4. Climate concern and the duration of GDR experience.

	(1)	(2)	(3)
Years East (×10)	$-0.006^{\ast\ast}$	$-0.051^{\ast\ast\ast}$
	(0.003)	(0.009)
Years East2 (×10)		$0.011^{\ast\ast\ast}$
		(0.002)
12–19 Years East			$-0.061^{\ast\ast\ast}$
			(0.012)
20–29 Years East			$-0.052^{\ast\ast\ast}$
			(0.011)
30–39 Years East			$-0.040^{\ast\ast\ast}$
			(0.012)
40+ Years East			$-0.025^{\ast\ast}$
			(0.012)
Observations	142 274	142 274	142 274

Note: Data from the German Socioeconomic Panel Survey from years 2009 to 2018. Each column presents estimates from a panel random effects model. The dependent variable is binary, equalling 1 for SOEP participants declaring 'Very worried about the consequences of climate change'. In columns 1 and 2 the main independent variable is the number of years (in decades) the respondent spent in the GDR. In column 3 the variables of interest are dummies indicating respondents that spent 12–19 to 40–41 years in the GDR. The omitted category is zero years in the GDR. All models include State and Year dummies and sociodemographic controls. The latter are: a set of employment status indicators; a binary variable indicating women; a set of indicators controlling for urbanization of respondents'place of residence until age 16; education years; age and age squared; ln(HouseholdIncome) and ln(HouseholdIncome)². Standard errors, clustered at the individual respondent level are reported in parentheses. $^{\ast}~p\lt0.1$, $^{\ast\ast}~p\lt0.05$, $^{\ast\ast\ast}~p\lt0.01.$

The first column of table 5 reports estimates from equation (3), assessing the evolution of climate concern over time. The coefficient on East is negative and statistically significant, suggesting that the probability of climate concern was lower by 4.8% points for East Germans in 2009–2010 compared to West Germans for the same period. The coefficients on the biannual dummies indicate that West Germans' probability of climate concern was fluctuating for the examined period. The probability of climate concern for West Germans was lower by 2.8% points in 2013–2014, but increased by 6.4% points in 2017–2018 relative to the baseline years. Turning to the interaction terms, the positive coefficient on East × Years 2013–2014 implies some convergence in attitudes took place towards the mid-2010s. The estimated influence of the remaining interaction terms is not statistically significant. Table A4 in the appendix repeats the analysis with general environmental concern as the outcome variable.

Table 5. Climate concern over time, across birth cohorts and age.

	(1)	(2)	(3)
East	$-0.048^{\ast\ast\ast}$	−0.018	$-0.109^{\ast\ast\ast}$
	(0.010)	(0.012)	(0.019)
Years 2011–2012	−0.004
	(0.004)
Years 2013–2014	$-0.028^{\ast\ast\ast}$
	(0.004)
Years 2015–2016	0.003
	(0.004)
Years 2017–2018	$0.064^{\ast\ast\ast}$
	(0.005)
East × Years 2011–2012	0.005
	(0.007)
East × Years 2013–2014	$0.016^{\ast\ast}$
	(0.007)
East × Years 2015–2016	−0.001
	(0.007)
East × Years 2017–2018	−0.004
	(0.008)
Born 1949–1960		0.012
		(0.010)
Born ≥ 1961		−0.013
		(0.014)
East × Born 1949–1960		$-0.037^{\ast\ast\ast}$
		(0.012)
East × Born ≥ 1961		$-0.034^{\ast\ast\ast}$
		(0.011)
Age			$-0.007^{\ast\ast\ast}$
			(0.002)
East × Age			$0.012^{\ast\ast\ast}$
			(0.003)
Observations	142 274	142 274	142 274

Note: Data from the German Socioeconomic Panel Survey from years 2009 to 2018. Each column presents estimates from a different panel random effects model. The dependent variable is binary, equalling 1 for SOEP participants declaring 'Very worried about the consequences of climate change'. The main independent variable 'East' is binary, indicating participants who lived in the GDR before the reunification of Germany. All models include State and Year dummies and sociodemographic controls. The latter are: a set of employment status indicators; a binary variable indicating women; a set of indicators controlling for urbanization of respondents' place of residence until age 16; education years; age and age²; ln(HouseholdIncome) and ln(HouseholdIncome)². Standard errors, clustered at the individual respondent level are reported in parentheses. $^{\ast}~p\lt0.1$, $^{\ast\ast}~p\lt0.05$, $^{\ast\ast\ast}~p\lt0.01.$

The second column of table 5 reports estimates from equation (4), testing for heterogeneous effects of the political system across birth cohorts. The estimated coefficient on East is statistically insignificant suggesting that climate concerns of pre-1949 born East Germans are indistinguishable to those of West Germans in the same birth-year cohorts. This lends some support to the identifying assumption that environmental attitudes were homogeneous across the German population in the years prior to the division. Birth cohort indicators capture the difference in climate attitudes for West Germans belonging in the 1949–1960 and post-1961 birth-year cohorts relative to those born before 1949. Coefficients on the cohort dummies are statistically insignificant, suggesting that climate concern did not systematically vary across West German birth cohorts. Finally, interaction terms capture the influence of exposure to the GDR's regime across birth cohorts. Estimates suggest that the probability of climate concern decreases for younger East German birth cohorts that experienced a significant part of their formative or productive years under the GDR regime.

To further examine how the political system influences the relationship between age and climate attitudes, I regress the outcome on the treatment indicator East, the age variable and their interaction. Estimates in the third column of table 5 show a negative relationship between age and concern for West Germans: a ten-year increase in age decreases the probability of climate concern by 0.7 percentage points. The pattern reverses when turning to former GDR residents: a ten-year increase in age increases their probability of declaring concern by 0.5. Evidence from international surveys suggests that the direction, magnitude and statistical significance of the age-climate concern relationship vary across countries with similar contemporaneous institutions and at comparable levels of development (Poortinga et al 2019). The result reported here suggests the possible influence of a country's institutional heritage in determining the direction and size of this relationship. Younger cohorts in countries with recent experience of totalitarianism may be less concerned about environmental degradation relative to older cohorts of their compatriots, and relative to their peers in countries without similar experience.