Political uncertainty and its long-run spatial effects Expected loss of market access along the temporary intra-Austrian border

August 3, 2015

WORKING PAPER WAS PREPAIRED FOR THE 11th EHES CONFERENCE 2015 IN PISA. First draft – Please do not cite without permission

Abstract How does a short period of political uncertainty affect regional economic growth in the long-run? We use the decade of post-war uncertainty in to analyze spatial effects in economic development. We focus on the temporary intra-Austrian border between the Western occupation zones (WOZ) and Soviet occupation zone (SOZ). A hand-collected dataset about municipal residents based on census data, food-vouchers and electorate data allow us to test the role of an expected loss of market access. We find that municipalities that are located in the WOZ but close to the temporary intra-Austrian border relatively shrink by around 30% in comparison to municipalities further away from the SOZ. During the period of occupation, relative shrankage was largest from 1949 to 1953 – excatly these years when tensions between the USA and the Soviet Union increased and a potential threat of division was feared. Relative shrinkage of temporary border regions does not stop before the mid 1980s. We show three sources of the persistent effect: First, young people are more likely to leave the temporary border regions, settled further away and get kids in ther years after. Second, relocation of industries out of the SOZ and Vienna towards the WOZ avoid to settle near by the temporary border. Third, investments of Marshall Aid was less likely in WOZ regions near the temporory border. Our findings suggest that a short period of uncertainty - or instabilty - can shape regional economic perspectives in the long run.

JEL-Code: N14, N94, R11, R12, R23.

Keywords: Uncertainty, Occupation of Austria, Market Access, Spatial Inequality, Relocation

Christian Ochsner1 Felix Roesel Ifo Institute – Leibniz-Institute Ifo Institute – Leibniz-Institut for Economic Research for Economic Research Dresden Branch Dresden Branch Einsteinstr. 3 Einsteinstr. 3 01069 Dresden, Germany 01187 Dresden, Germany Phone: +49(0)351/26476-26 Phone: +49(0)351/26476-28 [email protected] [email protected]

1 1 Introduction

Very little is known how uncertainty - especially political uncertainty - shifts regional growth pattern in the long-run. We use the unique period of uncertainty in post World War Two (WW2) Austria to analyze long-lasting regional effects based on political uncertainty. The randomly drawn temporary intra-Austrian border between the Western occupation zones (WOZ) and the Soviet occupation zone (SOZ) is used as an exogenous shock for the municipalities that are located close to this temporary border. First, we anlyze the municipal growth pattern for the entire post war period. Our overall results indicate a relative shrinkage of municipalities within the WOZ but that are located close to the temporary border of around 29 percent. This result is also robust when we analyze specific subsamples, when we control for market access to European core markets or when we estimate the relative shrinkage on (county) level. In general, smaller municipalities close the the temporary intra-Austrian border shrink relatively more than larger ones. The relative shrinkage does not stop before the mid 1980s. Furthermore, convergence can not be observed. Second, during the period of occupation from 1945 until 1955, the relative shrinkage of border municipalities was largest after 1949 until 1953. This coinsides with the foundation of the German Democratic Republic and with the beginning of the European Reconstruction Programm (ERP), better known as Marshall-Plan and a generall increase in tensions between the two main occupation forces in Austria: The USA and the Soviet Union.

Our study focuses on post-war Austria. We use the randomly drawn occupation zones by the allies to estimate the resulting regional inequality. After World War 2, Austria was divided into four occupation zone. The Western Occupation Zone (WOZ) was occupied by the US, GB and France. The Soviet Occupation Zone (SOZ) was occupied the Soviet Union. As Berlin after World War 2, Vienna was divided into four occupation zone. Figure ?? illustrates the drawing of the occupation zones in Austria. From 1945 onwards until 1955, Austria faced a decade of uncertainty about its future. The threat of a potential division of the country was severe, especially after the foundation of the German Democratic Republic (GDR).

We construct a hand-collected dataset to evaluate population growth patterns on district and munic- ipal level as well as the evolution of industrial work places on the district level. We compare regional population growth separately for the Western and Eastern Occupation Zone. We expect that the threat of a potential division affects municipalities and districts within the WOZ differently: Regions within the WOZ but close to the border will have a relative decline in population and industrial work places whereas regions further away from the SOZ will gain. Based on Redding and Sturm (2008) we assume that the expectation of a market based economy in the WOZ leads to a relocation of work

2 Allied Occupation Zones: Czech Republic United States Lower Austria Soviet Union Slovakia Vienna United Kingdom France

Provincial Capital Upper Austria Germany Burgenland

Hungary Styria Salzburg Vorarlberg Tyrol

Carinthia

Switzerland

Slovenia 0 75km

Figure 1: The occupation zones by the four allies between 1945 and 1955

Czech Republic

Slovakia Vienna

Upper Austria Lower Austria Germany

Burgenland

Styria Hungary Salzburg Vorarlberg Tyrol

Carinthia

Switzerland

Italy 0 75 km Slovenia

France United States United Kingdom Soviet Union Provincial Capital

Note: The map shows the four occupation zones from July 1945 until October 1955 in different colors. The bold lines within Austria show state/province borders, the thin lines shows districts borders. force and production facilities away from the border. Regions within the WOZ that are located close to SOZ will decline since they expect a loss of input and output market access.

Figure 2 shows strong evidence for our hypothesis. The figure uses data for all 1487 municipalities (territorial status of 2012) that have been located in the Western occupation zone. The dotted line show the average municipal population dynamics (indexed to 100 for the year 1939) from 1869 until 2001 for municipalities that have been located within 75 kilometers to the temporary intra-Austrian border. The black line represents municipal population growth dynamics for municipalities within the WOZ that have been located further away 75 kilometers to the temporary border. In the 60 years prior World War Two, population dynamics between these two municipal group differs only slightly whereas post war population dynamics is strongly in favor for municipalities further away from the temporary border. Interestingly, all municipalities within the WOZ grow around 15% percent during the years of WW2 until 1946. This is caused by refugees from the SOZ. Hence, differences in population dynamics have not been caused by the initial migration flow. Instead, the different growth pattern of municipalities started during the period of occupation and seem to be long lasting.

We apply a Fixed Effect estimation to test this findings econometrically. Furthermore, we observe a long lasting difference of the regional population dynamics within the WOZ after Austria was

3 Grafik neu ab 1869:

200

180

160

140

120

100 Population(1939100) =

80

60

FigureDistance 2: Municipal to temporary population intra-Austrian dynamics border: within the < Western 75 km occupation > 75 zones km

200

180

160

140

120

100

Population (1939 (1939 100) = Population 80

60

Distance to temporary intra-Austrian border: < 75 km > 75 km

Note: The graph shows population dynamics (indexed to 100 for the year 1939) for 1487 Austrian municipalities that have been located in the Western occupation zones from 1945 to 1955. The dotted line shows municipal population dynamics (arithmetic average) for municipalities that are located within 75 kilometers to the temporary intra-Austrian border, the solid line shows municipal population dynamics for municipalities further away than 75 kilometers to the temporary intra-Austrian border. Data for the years 1869 to 1939, for 1951 and for 1961 onwards are taken from the Statistical Office of Austria. Data for 1946 and 1948 are taken from municipal population statistics based on food vouchers. Municipal population data for 1949, 1953 and 1956 are estimated based on municipal changes of electorate. The red area indicates the period of World War Two (1939 - 1945). The gray area shows the period of Austrian occupation by the allied troops (United States, United Kingdom, France, and Soviet Union) from 1945 until 1955.

unified in 1955. We want to investigate the three possible channels how these long-lasting effects can be explained. The first channel is based on Redding et al. (2011): Once an industry location is relocated, it will never go back to its originate region. We want to test this idea by analyzing the development of industrial work force in fast growing industries during the „Wirtschaftswunder“. Second, the persistent differences might occur due to the specific age structure of migrants that leave the border zone in the WOZ as well as the SOZ during the period of uncertainty. If younger people migrate to municipalities further away from the temporary intra-Austrian border, this would shape the age structure in the respective regions: municipalities close to the temporary border might face a relatively lower birthrate in the subsequent decades which might cause municipal growth in the long- run. Third, Marshall aid, especially the ERP (European Reconstruction Programm) might influence the observed population dynamics. We have to figure out which regions get more Marshall Aid than others and how this aid causes the different municipal population dynamics.

4 UK Plan, Aug. 1944 SU Plan, Nov. 1944

UK Plan, Jan. 1945 SU Plan, Apr. 1945

Planned Occupation Zones: United States Soviet Union United Kingdom France

Figure 3: Planed and realized occupation zones by the allies

UK Plan, Aug. 1944: SU Plan, Nov. 1944: UK Plan, Jan. 1945:

SU Plan, March 1945: Liberation, May 1945: Occupation, 1945-1955:

France United States United Kingdom Soviet Union

Note: The upper maps and the lower left map show different occupation plans by the allies – especially by United Kingdom and the Soviet Union – during the war. The lower middle map shows the demarcation lines on May 9 1945. The allied troops occupied these regions until July 1945, when they withdraw to their assigned occupation zones. The lower right map shows the assigned occupation zones for the four allies. These occupation zones have been in place from July 23/24 1945 until the signing of the Austrian State Treaty on October 25 1955.

2 Historical background

In the final stage of World War Two allied troops liberated Austria from Nazi occupation. The Austrian borders were reestablished as they existed before the annexation of Nazi-Germany in 1938.

Austria became occupied by the fourGermany allied troops from the end of WW2 until the Austrian state treaty in 1955. Vienna, as Berlin in Germany, was also divided into four occupation zones.2 Based on this treaty, Austria was reestablished as an independent and neutral state. The four occupation forces, the United States, Great Britain, France and the Soviet Union left the country immediately and the restriction of free movement of people and goods, especially between the Western and the Soviet occupation zones were abolished.

The decision how to zone Austria is essential for our empirical identification. Especially, the zoning of Austria should be drawn independent of economic considerations. Hence, it is essential that the borders between the region we use for our empirical estimation and the Soviet occupation was exogenous to economic issues. In fact, we find no evidence that zoning of Austria is based on economic rather than on military considerations (see Williams (2000) or Erickson (1950)). First, in the case of an eventual

2The Nazis formed the so called Greater Vienna Area (Gross-Wien) in 1938. Even though the incorporated territories officially belong to Vienna until 1954, the division of Vienna was based on its boundaries prior the formation of Gross-Wien. The areas outside of the main city borders belong to the Soviet occupation zone.

5 Figure 4: Austrian Newspapers around the failing of State Treaty in 1949/1950

Note: The picture shows front pages of Austrian newspapers from August 1949 until May 1950. The incidence of State Treaty related headers was largest from November 1949 to February 1950. The headers are from the following newspa- pers: „Neues Österreich - Organ der demokratischen Einigung“(jointly published by the three main democratic parties – Austrian People’s party (ÖVP), Social democrats party (SPÖ) and the Austrian communist party (KPÖ) – in occupied Austria) and „Kleine Zeitung“published as an independent newspaper in Graz (Styria). Copies of these newspapers are based on micro forms and were taken in the Austrian National Library (Österreichische Nationalbibliothek) in Vienna. defeat of the Nazis, a remodeling of Austria was discussed. Especially the United Kingdom under prime minister Churchill promoted to establish a new state consisting of Austria and Southern parts of Germany. However, from 1943 onwards the main planning efforts were about a potential division of Austria into occupation zones. Plans for occupation zones where elaborated from the United Kingdom and the Soviet Union. Figure ?? shows different plans of occupation zones and the realized plan of a plan of April 1945. The first plan from August 1944 just consisted of a British and Soviet occupation zone with a divided Vienna. Later stages involved a US (Soviet plan from November 1944) and a French occupation zone (UK plan from January 1945) too. The definitive occupation zones based on a plan from the Soviets in April 1945. The occupation borders were draw according to political and geog-stratigic considerations rather than on economic reasons (see Karner und Stangler (2005)). As an example, the Soviet Union insisted to get the Northern parts of Upper Austria. The main target of this proposal was, the the Soviets did not want that the US occupation zone bordering directly Czechoslovakia. Hence, military consideration were main driving forces of drawing occupation zones (Erickson (1950)). The drawing of the border line between the Western and Soviet occupation zone seems to be exogenous with regard to economic considerations. In addition, a mean comparison of

6 the relative pre-WW2 size of industrial sectors does not show statistically significant differences for 21 out of 25 economic sectors.3 This indicates that the zoning of Austria was not drawn with respect any industrial specialized region.

3 Theoretical Considerations

In this section, we motivate the relative shrinkage of municipalities with respect to an expected loss of market access based on an economic geography model by Helpman (1998). A simplified adoption of the model is given by the following equation:

real  nominal nominal nominal Populationregion w = f wregion , ptrade , pnon−trade (1)

The population in a region, P opulationregion, is endogenously determined by the real wages workers earn in that region. Migration of workers with respect to real wages will equalize equilibrium real wages across regions but might result in an unequal distribution of people between regions. Real wages within a region are function of agglomeration and dispersion forces f(.). The function is determined by

nominal nominal regional specific nominal wages, wregion , the price of tradeable consumption goods, ptrade and the

nominal nominal nominal price of non-tradeable consumption goods, pnon−trade. wregion depends on a firm’s market access in a region. Increasing returns to scales and increasing transport costs lead to a concentration of production facilities near larger markets. This agglomeration forces will increase nominal wages for workers in larger markets. However, firms close to larger markets face also dispersion forces due to increased number of lower-priced competitors. This force leads to a decline in nominal wages. Beside the workers’ nominal income, nominal prices of consumption goods are taken into account for

nominal settlement decision. ptrade is the nominal price for tradeable goods in a region. Nominal prices of tradeables will fall (increase) in a regions with larger (smaller) market access due to the implied transportation costs and regional price-competition. The nominal price of non-tradeable consumption

nominal nominal goods, pnon−trade, increases with larger and falls with smaller market size. pnon−trade is the main dispersion force in the model. Taking all three aspects into account, a worker decision whether to migrate or not depends on the realized real wage which is the difference of the nominal wage and the sum of nominal prices for tradeables and non-tradeables. According to Redding and Sturm (2008), an increase in transport costs, e.g. due to a loss of market access, will reduce benefits from agglomeration

3We employ a t-test to test district wise mean differences between the Western and Soviet occupation zone of work-occupation. The data are obtained from the Volks, Berufs- und Betriebszählung vom 17. Mai 1939 for the Alpen- und Donaureichsgaue.

7 forces more than dispersion forces. 4 Due to the migration of workers, regions that face a loss of market access will shrink whereas other regions will grow.

We aply the idea of Helpman (1998) to evaluate how a potential loss of market access leads to a relative shrinkage of municipalities. We argue, that not only a realized loss market access determines the future size of a region. In addition, expectations during a period of political uncertainty might contemporaneous migration decisions of workers and investment decision of firms. Ceteris paribus, an expectet loss of market access can shape regional growth pattern in the long run. The following equation adopts equation 1 to post-WW2 period in Austria:

real h  nominal nominal nominal i Populationmuni,p>1955 w = Ep=occupation f wmuni , ptrade , pnon−trade (2)

The parameter p stands for the respective period. Municipal population in the periods that are followed by the period of occupation, P opulationmuni,p>1955, is a function of expected nominal wages and nominal prices for tradeables and non-tradeables. We hypothesize that migration and investments decision during the period of occupation is highly influenced by the expectation of a potential loss of market access in municipalities that are located within the WOZ but close to temporary intra-Austrian border. We expects a relative decline in municipal and regional population growth of entities with a geographic proximity to the Soviet occupation zone. However, this relative decline should be larger for municipalities or regions with a smaller internal market. Hence, we expect that the relative decline of larger municipalities close to the temporary border is less pronounced as the relative decline smaller municipalities since larger municipalities benefit from a larger home-market.

4 Data

Our basic dataset is a balanced panel about population of all Austrian municipalities based on the territorial status of 2014. We collect a unique data set for the years 1869 until 2001 based on three different sources: First, the Statistic Office of Austria provides municipal census data for entire period based on 2014 territorial status. Secondly, we add these data with municipal population data based on food-vouchers for the years 1946 and 1948. Third, we estimate municipal population data for the years 1949, 1953 and 1956 based on the municipal change of total electorate. We obtain these data

4Redding and Sturm (2008) calibrate the sizes of regions based on a loss of market access (increase nominal nominal nominal in transport costs) with German data. They find that wregion + ptrade > pnon−trade due to an increase in transport costs, for example due to a loss in market access. Hence, the drawbacks of smaller nominal wages and more expensive nontradeables exceed the benifits from lower nominal prices for non-tradeables.

8 from parliamentary and presidential elections. The detailed population pattern for the years 1946 to 1956 allows us to get a very detail view of municipal population pattern during the occupation period. To ensure the territorial comparability of the data based on food-vouchers and electorate data, we aggregate municipalities which merged (or split up) between 1946 and 2014.5 We end up with a panel with 44,783 observations (2,357 municipalities in 19 time observation).

Table 1 provides a summary of Austrian municipalities together and separately for the four occupa- tion zones. Linz and Vienna are dropped from the sample for all our empirical estimates because of their division into two (Linz) and four (Vienna) occupation zones from 1945 to 1955. Our summary statistics contain data about the number and share of municipalities by occupation zone, pre-WW2 characteristics such as population and population density, pre- and post-WW2 average municipal population growth rates as well as time in-variant municipal classification into alpine or hilly munic- ipalities. All in all, 1,487 (63%) out of the overall sample of 2,357 municipalities are located in the three Western occupation zones. Out of these, 442 (19%) municipalities are located in the US, 704 (30%) in the UK and 342 (15%) in the French occupation zone.

In addition to municipal population data, we employ a bundle of further expationary varibles on the municipal or district level. These are municipal latitude and longitude data to measure the distance to the temporary intra-Austrian border, district data about pre- and post-WW2 sectoral workforce occupation and employment shares, gross and net birthrates, net migration and sectoral and regional data about investments of the European Reconstruction Program (ERP) from 1948 to 1953. We describe these data when introduced in our empirical estimation and in our Appendix, section 9 .

5A detailed description of these data sources and the transformation to the 2014 territorial status is given in Appendix, section 9 .

9 Table 1: Summary statistics

Municipalities Population in 1939 Annualized Pop Growth Municipal Characteristics No. Share Mean Std. Dev. Min Max pre-WW2 post-WW2 Density Hilly Alpine (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

Austria (w/o Vienna) 2,356 1.00 2,072.0 6,190.2 70 208,016 0.44 0.61 170.9 0.51 0.26

Western Occ. Zones (WOZ) 1,487 0.63 1,898.3 6,675.2 70 208,016 0.33 0.88 180.5 0.54 0.42 United States Zone 441 0.19 1,987.0 4,324.1 135 77,170 0.34 1.00 149.2 0.70 0.16 10 United Kingdom Zone 704 0.30 2,070.6 2,070.6 175 208,016 0.30 0.37 169.6 0.63 0.37 France Zone 342 0.15 1,429.3 4,721.7 70 81,710 0.38 1.76 243.0 0.14 0.86

Soviet Occ. Zone (SOZ) 868 0.37 2,224.3 3,048.6 135 48,583 0.63 0.16 152.8 0.45 0.00

Linz (US & SU Occ. Zones) 1 0.00 128,177.0 - - - 1.81 0.70 1663.3 0.00 0.00

Vienna (All Occ. Zones) 1 - 1,770,938.0 - - - 1.38 -0.20 5310.2 0.00 0.00 Note: The summary statistics use only population data that are taken from the Statistic Office of Austria. Column (1) and (2) show the number and the share of municipalities (territorial status in 2014) for Austria and according to the four occupation zones. Linz and Vienna are listed separately due to their division in different occupation zones from 1945 to 1955. Columns (3) - (6) show the mean, standard deviation, minimum and maximum of municipal population in 1939 for Austria and according to the four occupation zones. Columns (7) and (8) show the annualized municipal growth rates from 1869 to 1939 (pre-WW2) and from 1939 to 2001 (post-WW2). Column (9) shows municipal population density in 1939. Density is defined as the total municipal population per square kilometer that is amenable to settle (settlement area). Columns (10) and (11) show the shares of hilly and alpine classified municipalities. The rest are municipalities that are classified as flat-land municipalities. 5 Estimation strategy and results

5.1 Estimation framework

We use a Fixed Effect model to analyze the relative shrinkage of municipalities within the WOZ with respect to geographic proximity to the temporary intra-Austrian border. Our estimation strategy has the following form:

Popgrowthit = α + β [Borderi ∗ postWW2 ] + δi + dt + εi,t (3)

The left-hand side variable, P opgrowthit, is the annualized municipal population growth rate. The dependent variables on the right read as follows: α is a constant. The coefficient of interest – our interaction coefficient – is β. β measures the relative shrinkage of municipalities close to the temporary intra-Austrian border compared with municipalities further away in a given period. In our baseline specification, Borderi is a dummy equals one (zero) if a municipality in the WOZ is located within (outside of) a certain distance strip to the temporary intra-Austrian border. postW W 2 is a dummy which is zero prior WW2 (until 1939) and one afterwards. We use different specifications of Borderi and postW W 2 when we analyze different subsamples or different time periods. δm are municipal

fixed effects and dt are time (in our specification period) fixed effects. εi,t is a heteroscedastic robust error term clusterd on municipal level. Concerning the coefficient of interest β, the FE estimation is equivalent to a Difference-in-Difference estimation approach as in Redding and Sturm (2008).

The specification (3) also allows to control for observed and unobserved fixed effects that might influence municipal population dynamics but are time invariant. For example, a potential influence of municipal characteristics as the geographic landscape, sea level, time invariant spatial externalities and even differences in regional cultural norms that might municipal population growth rates. However, in section 7 we apply a matching approach to compare subsamples of border and control municipalities to address potential growth dynamics based on time-invariant municipal characteristics. The period time dummies, dt, control for common shocks which influence municipal population dynamics in the WOZ. This are, for example, the shocks of the First and Second World Wars and the shock of the occupation period when average municipal population within the WOZ increases by roughly 15% due to the flight from the SOZ. Last but not least, time dummies also capture the declining fertility behavior in the post-WW2 period.

11 5.2 Baseline estimation results

Table 2, table 3 and table 9 (see Appendix) contain our baseline estimation results. We classify the periods after 1939 until 2001 as postWW2 period. We start in 1939 because due to the fact that the first population data are only available for 1946 (food vouchers). However, it is essential to treat the period from 1939 to 1946 as postWW2 period since in the direct aftermath of WW2, a huge number of internal refugees from the Soviet zone and from Vienna migrate to the Western zones. Hence, this migration patterns are a direct effect of the occupation which we address for.

Column (1) in table 2 shows a relative average shrinkage for the overall post-WW2 period of 0.47 percent p.a. WOZ-municipalities with a geographic proximity of 75 kilometers to the temporary intra- Austrian border face a relative decline of total population of more than 29 percent in comparison to municipalities further away. This effect is highly statistical significant. The distance threshold of 75 kilometers is obtained by investigating heterogeneous population dynamics depending on the distance to the temporary border. Column (2) in table 2 introduce distance dummies for municipalities that are located within distance strips of 25 kilometers to the temporary border. We interact four distance strips with the post-WW2 period. The coefficient is highly statistically significant for the first three 25 kilometer strips, but is insignificant afterwards. Hence, the temporary border effect has no negative impact on municipal growth dynamics for municipalities further away than 75 kilometers to the temporary border.

Columns (3) and (4) in table 2 contain heterogeneous time effects by comparing border municipalities with the entire WOZ (column (3)) and within the US and UK zones only (column (4)). We interact every single period after 1939 with the group of temporary border municipalities. Three time effects are worth to be mentioned: First, temporary border effects for the entire WOZ in the treatment period from 1939 until 1946 shows only a slow and a weak statistically significant relative population decline. Hence, internal refugees from the Soviet zone have only a weak settlement bias in favor for municipalities further away from the temporary border. Second, the relative decline of temporary border municipalities starts to be strong and highly statistically significant after 1946 and is visible during the entire post-WW2 period but with a reduction of the relative decline in the last decades. Third, the relative shrinkage of border municipalities within the US and UK zone is only temporary and dies out in the 1980 (column (4)). This leads to three striking results: First, border municipalities within the US and UK zone face only a relative decline of their population for approximately 30 years after the temporary border was abolished. Second, US and UK zones municipalities grow at the same relative rate since the 1980s as they have done prior WW2. This is, as we argue, a strong incidence for steady state municipal population growth. Third, ongoing persistent effects in column (3) are caused

12 Table 2: Basic results of the temporary intra-Austrian border effect

Dependend Variable: Annualized Population Growth Spatial Effects Time Effects WOZ by Strips WOZ US and UK Zone (1) (2) (3) (4) Border x postWW2 -0.47*** (0.04) Border 00-25km x postWW2 -0.48*** (0.05) Border 25-50km x postWW2 -0.46*** (0.06) Border 50-75km x postWW2 -0.54*** (0.06) Border 75-100km x postWW2 -0.10 (0.10) Border x Year 1939-46 -0.26* -0.45** (0.15) (0.18) Border x Year 1946-51 -0.65*** -0.44** (0.14) (0.20) Border x Year 1951-61 -0.63*** -0.42*** (0.07) (0.09) Border x Year 1961-71 -0.77*** -0.28*** (0.07) (0.08) Border x Year 1971-81 -0.47*** -0.13* (0.06) (0.07) Border x Year 1981-91 -0.28*** 0.00 (0.05) (0.07) Border x Year 1991-01 -0.25*** 0.02 (0.05) (0.06) Constant 0.24*** 0.24*** 0.24*** 0.30*** (0.03) (0.03) (0.03) (0.03) Year Effects YES YES YES YES Municipality FE YES YES YES YES No. of Obs. 20,804 20,804 20,804 16,016 R-Squared (adj.) 0.14 0.14 0.14 0.14 Note: Data are a panel of 1487 Austrian Municipalities that have been located in the Western occupation zones after WW2. Data are official census data from the Statistical Office of Austria and a population census based on food vouchers in 1946. The left-hand-side variable is the annualized rate of growth of municipal-population, expressed as a percentage. Population growth rates are for 1869-1880, 1880-1890, 1890-1900, 1900-1910, 1910-1923, 1923-1934, 1934- 1939, 1939-1946, 1946-1951, 1951-1961, 1961-1971, 1971-1981, 1981-1991 and 1991-2001. Border is a dummy that equals one (zero) if a municipality lies within (outside of) 75 kilometers to the temporary intra-Austrian border to the Soviet occupation zone. postWW2 is a dummy which is zero prior World War Two (1939) and one afterwards. Column (1) shows the overall temporary intra-Austrian border effect for the entire post-WW2 period. Column (2) shows the overall temporary intra-Austrian border effect for the entire post WW2 period by distance strips of 25 kilometers. Distance strips are dummy variables that equals 1 if a municipality is located in one of the respective strips. Column (3) and (4) show time effects of the relative shrinkage of the municipalities within 75 kilometers to the temporary intra-Austrian border. The year dummies equals one for the respective sub-periods. Column (3) compares the relative shrinkage of border municipalities with municipalities from the entire WOZ, column (4) does it for US and UK occupation zones only. The Fixed Effect estimation uses time and municipality fixed effects. Standard errors are heteroskedasticity robust. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

13 by a massive and ongoing population and economic push in municipalities in Vorarlberg and Tyrol (French zone).6

Table 9 shows the relative shrinkage of municipalities in the WOZ during the overall post-WW2 period for all four regional subsamples. The overall relative decline in the US and UK zone together is 15% for the entire period (-0.24% p.a.). Border municipalities within the UK zone relatively shrink more than municipalities within the US zone. However, these results are highly (weakly) statistically significant for the UK (US) zone. The significant negative temporary border effect in the US zone is somehow surprising. First of all, large industrial cities such as Linz, Steyr or Wels and their surrounding municipalities belong to the border municipality group where we would predict an increasing population due to the post-WW2 economic boom. Secondly, the river of Danube provides cheap shipping cost which would have favored regions near by the temporary border in the US zone. However, this geographic and industrial preconditions seem to be less important than the shock of a potential division of Austria.

Table 3: Heterogeneous size effects of the temporary intra-Austrian border

Dependent Variable: Annualized Population Growth Municipality size City size Small Medium Large > 5,000 > 10,000 (1) (2) (3) (4) (5) Border x postWW2 -0.71*** -0.37*** -0.24*** -0.40* -1.05** (0.07) (0.07) (0.07) (0.22) (0.40) Constant 0.03 0.16*** 0.54*** 1.45*** 2.12*** (0.04) (0.03) (0.06) (0.17) (0.43) Year Effects YES YES YES YES YES Municipality FE YES YES YES YES YES No. of Obs. 6,944 6,958 6,902 882 266 R-Squared (adj.) 0.17 0.14 0.14 0.22 0.26 Note: Data, the dependent as well as the explanatory variables are the same as in table 2 . Column (1) - (3) show heterogeneous shrinkage effects depending on municipality size in 1939. Column (1) contains municipalities in the lower third, column (2) municipalities in the middle third and column (3) in the upper third of the population range in 1939. Column (4) and (5) show the entire post-WW2 effects of the temporary intra-Austrian border for cities with more than 5,000 inhabitants in 1939 (column (4)) and with more than 10,000 inhabitants in 1939 (column (5)). The Fixed Effect estimation uses time and municipality fixed effects. Standard errors are heteroskedasticity robust. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

Now, we turn to the question whether the persistent effects of the temporary border is driven by the size of municipalities. Up to now, our results show the average shrinkage of border municipali- ties independent of their absolute population. However, regional implications beyond the municipal

6As we show in section 6, especially Tyrol get a huge positive economic shock due to the relocation of businesses and production facilities, especially from Vienna, in the late 1940s. This might facilitate the rise of these regions (see Mähr, 1989 and Hofbauer, 1992)

14 viewpoint might be much lower when the relative shrinkage is mainly driven by small municipalities. Therefore, we divide our municipal sample into three subsamples based on their population in 1939.7 Columns (1) - (3) in table 3 show the relative shrinkage of border municipalities since WW2 for the smallest third (column (1)), for the middle third (column (2)) and the largest third (column (3)) of all municipalities in the WOZ. Columns (4) and (5) show the temporary border effects for cities that have more than 5,000 and 10,000 inhabitants in 1939. We observe declining relative shrinkage patterns by municipal size: The smallest third of municipalities relatively shrink three times as much as the largest third. However, all size subsamples observe a highly statistically significant relative decline of border municipalities. However, the declining shrinkage pattern in municipal size is not uniform: The relative shrinkage of cities that are located within 75 kilometers to the temporary border is higher than the relative shrinkage of the largest third municipalities: Cities with a population with more than 5,000 (10,000) relatively shrink by around 25% (60%) during the entire WW2 period.8

5.3 Effects during occupation until 1955

Our main empirical findings so far show a significant relative shrinkage of municipalities that have been located close to the temporary intra-Austrian border. We now turn to the question of the relative decline of border municipalities during the period of occupation. Especially, we are interested if the relative shrinkage of border municipalities was uniform during the period of occupation or caused by political events in the years 1945 to 1955. Especially, we are interest if the takeover by the Communists in Czechoslovakia and Hungary 1948 or the failing negotiations about the Austrian State Treaty in 1949/1950 – which coincides with the foundation of the GDR in the Soviet occupation zone in Germany – leads to a contemporary reaction in WOZ-internal migration patterns. For this propose, we first enlarge our municipal population data with an additional population census based on food vouchers in 1948. Second, we estimate municipal population (Pd opi,t) for the years 1949, 1953 and 1956 (nearest observable date around the State Treaty in October 1955) based on the change of municipal electorate (Eleci,t) for parliamentary or presidential elections. We use the first presidential election 1951 (Eleci,1951) as reference because for 1951 we get census based municipal population data as well as the number of municipal electorate. Hence, the following municipal population estimates are applied:

7This issue can also be addressed due to an analysis on the district level. We do this in section 7 . 8This result is mainly driven by cities in Vorarlberg and Tyrol, where we observe relocation of industries from the SOZ and Vienna during the period of occupation (see section 6).

15   Eleci,1949 Popd i,1949 = ∗ Popi,1951 ∗ Se1949 (4) Eleci,1951

  Eleci,1953 Popd i,1953 = ∗ Popi,1951 ∗ Se1953 (5) Eleci,1951

  Eleci,1956 Popd i,1956 = ∗ Popi,1951 ∗ Se1956 (6) Eleci,1951

9 Set is a scaling factor for municipal population. Hence, Set normalizes the first term on the right hand side in equations 4 - 6 in that way that the estimated overall population in Austria equals the official yearly Austrian population.10 We end up with municipal population data during the period of occupation for the years 1946, 1948, 1949, 1951, 1953 and 1956.

Table 4 contains our occupation results for entire WOZ (column (1)) and the US and UK zones together (column ((2)) and separately (columns (3) and (4)). The first sub-occupation period holds from 1939 to 1946 and shows the relative shrinkage of border municipalities. Once again, this period is classified as occupation period because the internal emigration wave out of the SOZ and Vienna take place immediately after the end of WW2. Hence, the estimation results for the first occupation period is mainly driven by the initial settlement decisions of refugees. We find heterogeneous attractiveness for settlement: Border municipalities within the US zone were relative attractive for immigrants from the SOZ due to US food provision (Mähr 1989) whereas border municipalities in the UK zone are not. Especially the Eastern parts of Styria were less attractive for internal refugees due to the initial occupation by the Red Army until July 1945 (see Iber et al. (2008)).11. Overall, we find for this period only weak relative lower attractiveness of municipalities close to the temporary border. The years 1946 and 1947 show a statistically significant decline of border municipalities. However, we do not observe any significant relative relocation of people within the US and UK zones. Hence, the relative shrinkage in these years was caused by the relative growth of municipalities in the French zone. In the following year (1948/49) only US border municipalities relatively shrink enormously, whereas by taking the overall WOZ effect into account, no significant shrinkage of border regions is observed. The effect in the US coincides with takeover by the Communists in Czechoslovakia and the

P P op 9 i c i,1949 For example, the scaling factor for 1949 is S1949 = . e P opAustria,1949 10Yearly population data for Austria outside of official census years are available only for the entire country. 11We adress the intervention of the Red Army in Central and Eastern Styria from May until July 1945 in our robustness checks in table 7

16 Table 4: Heterogeneous time effects of the intra-Austrian border during the period of occupation until 1955 (1956)

Dependend Variable: Annualized Population Growth WOZ US and UK UK Occ.Zone US Occ.Zone Occ.Zone (1) (2) (3) (4) Border x Year 1939-46 -0.26* -0.45** -1.16*** 0.70*** (0.15) (0.18) (0.23) (0.27) Border x Year 1946-48 -0.65** -0.27 0.02 -0.73 (0.27) (0.42) (0.41) (0.88) Border x Year 1948-49 -0.81 -0.24 1.45 -2.95*** (0.93) (1.11) (1.71) (0.91) Border x Year 1949-51 -0.40*** -0.55*** -0.52*** -0.61** (0.12) (0.14) (0.16) (0.25) Border x Year 1951-53 -0.63*** -0.26* -0.47*** 0.07 (0.14) (0.15) (0.16) (0.29) Border x Year 1953-56 -0.20 0.13 0.02 0.31 (0.13) (0.16) (0.13) (0.35) Constant 0.34*** -0.02 -0.17 0.23 (0.11) (0.15) (0.13) (0.32 ) Year Effects YES YES YES YES Municipality FE YES YES YES YES No. of Obs. 19,318 14,872 9,139 5,733 R-Squared (adj.) 0.01 0.01 0.01 0.16 Note: Data are the same as in table 2 . In addition we use food vouchers data for 1948 and estimated municipal population data for 1949, 1953 and 1956 based on changes in municipal electorate. Border indicates municipalities within WOZ that are located within 75 kilometers to the temporary intra-Austrian border. The year dummies equals one for the respective sub-periods. Column (1) indicates the border effect during the occupation period for the whole WOZ (from 1939 until 1956). Column (2) includes municipalities in the US and UK occupation zones together. Columns (3) and (4) include municipalities in the UK and US occupation zones separately. The Fixed Effect estimation uses time and municipality fixed effects. Standard errors are heteroskedasticity robust. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%. beginning of the Marshall Aid in Austria.12 The years after 1949 show a highly significant shrinkage of border municipalities for all regions under consideration. This are strong evidences that the internal migration pattern after 1949 is highly driven by a potential threat of division after the failing of the State Treaty. Austrian people, as shown in section 2, have been well aware of the upcoming disputes among the negotiators and the raising threat of an unilateral treaty for the WOZ and the SOZ respectively. The relative shrinkage of border municipalities hold on until 1953. During the last occupation period, the relative shrinkage of border municipalities stops.

Table 10 in Appendix summarizes the relative shrinkage of border municipalities for the entire occu-

12The effects of the Marshall Aid will be analyzed in section 6. However, internal US zone migration might be affected by Marshall Aid since regions further away from the temporary border receive more ERP funds than direct border provinces.

17 pation period. The average relative shrinkage of municipalities within 75 kilometers to the temporary border is 6.5%. Concerning US und UK zones seperately, the overall effect is largest in UK zone, whereas the direct occupation effects in the US zone are unclear.

6 Channels of the long-lasting border effects

6.1 Cohort specific migration pattern and fertility

Here it comes. A litte bit of a hope. Nothing worth to hope for. Destiny is is generosity.

6.2 Relocation of industries during occupation

In his study on the history of large Austrian companies, Mathis (1987) reports several cases of shifting industrial businesses from Eastern to Western Austria after WW II. Remarkably, most places of new places of production are located in the very west of Austria (Tyrol, Vorarlberg). We argue that espe- cially the high level of uncertainty, war destruction and Soviet deconstruction drives relocation. First, uncertainty leads east Austrian corporations to found settlements in western Austria. The electrical engineering Brown, Boverie & Cie. (BBC) concern established a Western Austrian firm in Tyrol’s capital Innsbruck in 1947/1948 due to the uncertain future of its Soviet controlled head quarter in Vienna. The Schmidtstahlwerke (mechanical engineering sector) founded two branches in Innsbruck and Graz. Mathis (1987) reports also of production related impacts of lost market access. The ma- chine building Maschinenfabrik Andritz AG near Graz (within our treatment area) reports of heavy shortages in material which leads to a stop in production for almost quarter a year. Also the Julius Roemer KG in Nettingsdorf (paper mill, 400 workers in 1938) which is located within the WOZ but close to the intra-Austrian border suffers from scarcity of raw materials. Regular production after WW II could not start until 1947 when the firm hired a wood mill in the more western city of Dellach (Carinthia) which is located outside our treatment area. Western Austria was also much more popular in founding new businesses than the Soviet zone. The textile producing J. Kunert & Söhne concern established new production facilities in the western Austrian Rankweil (Vorarlberg) after socialization of its origin place of production in Czechoslovakia in 1949. Another example is the new founding of the Tiroler Röhren- und Metallwerke AG in Hall (Tyrol). Put generally, the demarcation line was seen as a massive problem in âhindering businessâ (WIFO 1945, WIFO 1946) especially in the direct aftermath of WW II. The western parts of Austria has been perceived as an even safer place to start business.

18 Table 5: Migration pattern of the „fertile cohorts“and fertility behavior on district level

SOZ WOZ Distance to temporary intra-Austrian Border Difference < 75 km > 75 km = (3) - (2) (1) (2) (3) (4) „Fertile cohorts“ at the End of WW2 Pre-WW2: „fertile cohorts“as share of total population (1939 Census) Age cohorts 20-41 in 1945 34.51 35.15 35.57 -0.42 Age cohorts 16-46 in 1945 49.55 50.27 51.02 -0.74 Pre-WW2 vs. Post WW2: Growth of „fertile cohorts“ between 1939-1961 (annualized) Age cohorts 20-41 in 1945 -1.65 -1.36 -0.97 -0.39** Age cohorts 16-46 in 1945 -1.43 -1.02 -0.52 -0.50** Fertility Pre-WW2: Gross birth rate per 1,000 capita (yearly data) 1933 17.83 19.35 20.15 -0.80 1934 17.07 17.98 18.77 -0.79 1935 16.49 18.13 18.38 -0.24 1936 16.04 17.88 18.70 -0.81 Post-WW2: Gross birth per 1,000 capit(yearly/annualized data) 1948 18.74 19.80 22.44 -2.64*** 1951 15.85 17.84 20.29 -2.45*** 1952 16.11 18.00 20.25 -2.25*** 1953 16.50 17.93 20.20 -2.27*** 1954 16.58 17.98 20.29 -2.31*** 1955 17.77 18.76 20.98 -2.22*** 1951-1961 15.59 16.83 19.00 -2.17*** 1961-1971 18.10 19.63 22.25 -2.61*** 1971-1981 12.54 13.49 15.25 -1.76*** 1981-1991 11.55 12.26 13.54 -1.28*** 1991-2001 10.99 11.60 12.93 -1.33*** Post-WW2: Net birth per 1,000 capita (annualized) 1951-1961 5.53 7.45 11.34 -3.88*** 1961-1971 4.92 7.66 12.18 -4.52*** 1971-1981 -0.53 1.72 5.70 -3.98*** 1981-1991 -0.42 1.44 4.59 -3.14*** 1991-2001 -0.35 0.96 3.87 -2.92*** Net migration per 1,000 capita (annualized): 1951-1961 -7.22 -7.49 -4.78 -2.71* 1961-1971 -2.48 -2.69 -1.34 -1.34 1971-1981 0.69 -1.28 -0.61 -0.67 1981-1991 2.38 -0.16 0.80 -0.97 1991-2001 3.95 0.82 0.93 -0.11 Housing shortage 1951 2.39 6.87 11.99 -5.12*** Note: Pre-WW2 data are in italic letters, post-WW2 data are in regular letters. Column (1) shows shares and growth rates for districts that have been located in the SOZ. Columns (2) and (3) show shares and growth rates for districts in the WOZ separated in districts which capital city is located within 75 kilometers (column (2)) and further away than 75 kilometer (column (3)) to the temporary intra-Austrian border. Column (4) takes the mean differences of the district wise sub groups by distance within the WOZ by applying a t-test for mean comparisons. Means that are significantly different from border districts with control districts are denoted by the following system: *10%, **5%, and ***1%.

19 Second, there are several cases of relocations due to war destruction. Many heavy fights in the last days of WW II in Spring 1945 took place at the eastern front line in Lower Austria, Styria, Burgenland and Vienna. The electrical industry concern Elin AG shifted its war destroyed place of production from Vienna to the Upper Austrian city of Gmunden near Salzburg. The Vereinigte Färbereien AG (textile) relocated their head quarter from Vienna to Reutte in Tyrol. The metal working firm Redtenbacher & Co. relocated its production within Upper Austria from Losenstein to the 50 km western located municipality of Scharfenstein. As another example, the metal working Ãsterreichis- chen Siemens-Schuckert-Werke AG relocated their production temporarily from the Vienna to Styria (Deuchendorf) during the occupation period up to 1955. Lastly, also massive deconstruction and long-term occupation of Eastern Austrian firms by the Red Army affects production and employment. After WW II, Enzesfelder ordnance factoryâs workforce (Lower Austria) shrinks dramatically from 16,000 to only 300 workers and has been put under Soviet administration until 1955. The same was true for the textile producing Erste österreichische Glanzstoff-Fabrik AG (EÃG) which lost more than two-third of its 1938 workforce (2,000 workers) under Soviet occupation in Lower Austriaâs capital St Pölten. As a last example, the Steyr-Daimler-Puch AG lost places of production in Eastern Austria due to Red Armyâs confiscation. To sum up, firm level evidence supports the findings of a heavy shift in business to the west of Austria in the aftermath of WWII. In stark contrast to the examples of âgoing westâ outlined above, we do not find a single example of relocating production to the eastern Austria albeit the study of Mathis (1987) gives a very comprehensive overview on all large Austrian companies. Thus, we conclude relocating businesses of high economic significance from the eastern to the western part of Austria in the aftermath of WWII to be an important channel of long-lasting effects.

6.3 Industrial dynamics and Marshall Aid

20 Figure 5: Relacation of Headquarters and establishments

Czech Republic

0 75 km

Place of origin New location

Note: The map shows relocation of headquarters and production facilities towards (or within) the Western occupation zones. The map is based on a survey about Big Business in Austria (see Mathis, 1987). We get informations only for firms that have had once at least 1,000 employees during the entire twentieths century until the mid 1980s.

Table 6: Sectoral shares and sectoral dynamics on district level

SOZ WOZ Distance to temporary intra-Austrian Border Difference < 75 km > 75 km = (3) - (2) (1) (2) (3) (4) Pre-WW2: Sectoral shares in 1934 Agriculture 43.09 43.98 42.04 1.94 Industry 28.06 27.20 25.64 1.55 Others (Services) 28.85 28.83 32.31 -3.49 Pre-WW2 vs. Post WW2: Dynamics of industrial shares 1934-1961 (annualized) Agriculture -1.39 -1.50 -1.30 -0.20* Industry 1.08 2.18 3.88 -1.70*** Others (Services) -1.87 -1.74 -1.16 -0.58** Note: Column (1) - (3) show the district wise work occupation of residents based on the Census in 1934 (Rows in italic letters) and their annualized growth rates from 1934 to 1961 (Rows in regular letters). Column (2) and (3) show the district wise occupation of residents for districts within WOZ based on district‘s capital distance to the temporary intra-Austrian border. Column (2) does it for districts which capital cities are located within 75 kilometers, column (3) with districts which capital cities are located further away than 75 kilometers to the temporary intra-Austrian border. Column (4) takes the mean differences of the occupational shares by applying a t-test for mean comparisons. Means that are significantly different from border districts with control districts are denoted by the following system: *10%, **5%, and ***1%.

21 7 Robustness checks of the temporary border effects

In this section, we run a series of robustness checks to support our findings so far. We explore several robustness checks in which we address several issues that might challenge our findings.

Main Robustness Checks

Columns (1) and (2) in table 7 show the results of our main robustness checks. First, we use a randomly assignment of municipalities into the border and control group. In column (1) the first 600 municipalities in the WOZ according to alphabetic ordering (from Abfaltersbach to Krispl) are assigned to the pseudo border group. The result shows no relative shrinkage of the randomly as- signed municipalities into (pseudo-) border and (pseudo-) control group. Column (2) controls for the liberation and temporary occupation of Central and Eastern Styria by the Red Army from May to July 1945. This episode of approximately 2.5 months makes Central and Eastern Styrian municipali- ties less attractive for internal refugees. In addition, the Red Army does deconstructions of German owned WW2 production facilities during their temporary occupation of Styria (Iber et al. (2008)). We control for this issues by reducing our sample to municipalities that are located in the West of the 15.15◦ East longitude. This is the demarcation line between Köflach and Voitsberg, where on May 8 1945 the Red and the Royal Army met. Therefore, we drop 314 municipalities (5,236 observations) form the overall sample. The relative shrinkage of the remaining border municipalities remain highly statistically significant and the overall post-WW2 shrinkage effect is only slightly smaller than in the entire sample.

Western integration

We face the risk that our findings of the relative decline of temporary border municipalities in the WOZ of around 29% are mainly caused by worse market access conditions to Western European core markets. In fact, municipalities within a 75 kilometer strip to the temporary intra-Austrian border are in general further away from core markets in Southern Germany and Northern Italy. However, we already show two arguments that the relative decline of temporary border municipalities is not caused by worse market access condition: First, the relative shrinkage of border municipalities inside the US and UK zone stops in the 1980s and relative municipal growth rates return to their pre- WW2 steady state (see table 2). Hence, even though the control group municipalities in the US and UK zones have in general a closer geographic proximity to Germany and Italy, their relative growth rates are not sustainable different from those municipalities further away from these core markets. Second, we even observe a relative shrinkage of border municipality by considering municipalities in the US zone only. There, geographic proximity to Germany is equivalent for both, border and control group municipalities. However, to further address the potential influence of Western integration and

22 Table 7: Randomization, the liberation of Styria and Western integration

Annualized Population Growth Geographic proximity to Germany Main WOZ US Zone Alpha <15.15◦ < 25 km < 50 km < 25 km < 50 km (1) (2) (3) (4) (5) (6) Pseudo-Border x postWW2 -0.04 (0.04) Border x postWW2 -0.40*** -0.48*** -0.38*** -0.37*** -0.23** (0.05) (0.06) (0.05) (0.13) (0.12) Constant 0.24*** 0.18*** 0.15*** 0.13*** 0.28*** 0.19*** (0.03) (0.03) (0.04) (0.03) (0.04) (0.04) Year Effects YES YES YES YES YES YES Municipality FE YES YES YES YES YES YES No. of Obs. 20,818 15,568 6,958 10,052 3,206 5,054 R-Squared (adj.) 0.13 0.16 0.20 0.22 0.29 0.32 Note: The dummies Border and postWW2 are defined as in table 2 . Column (1) uses alphabetic ordering of municipal- ities to assign them as (pseudo-) border municipalities. Pseudo-Border equals one for the 600 first municipalities in the WOZ according to alphabetic ordering (from Abfaltersbach to Krispl) and zero otherwise. Column (2) shows the effect of the temporary intra-Austrian border for municipalities that are located in the West of 15.15◦ East longitude only. Columns (3) - (6) show the temporary intra-Austrian border effects for municipalities that have a geographic proximity to Germany of less 25 (columns (3) and (5)) and less than 50 kilometers (columns (4) and (6)). Border is a dummy that equals one (zero) when a municipality is located within (outside) 75 kilometers to the temporary intra-Austrian border. The dummy postWW2 is defined as in previous tables. Columns (3) and (4) include municipalities within the WOZ with a geographic proximity to Germany, columns (5) and (6) include such municipalities in the US occupation zone only. The geographic restrictions in column (2) - (6) are illustrated in figure 6. The Fixed Effect estimation uses time and municipality fixed effects. Standard errors are heteroskedasticity robust. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%. geographic proximity to core markets, we estimate the temporary border effect for municipality within a certain distance strip to Germany. Columns (3) -(4) in table 7 replicate our baseline estimation for municipalities that have a geographic proximity to the German border of less than 25 (column (3)) and less than 50 kilometers (column (5)). We do this for municipalities in the entire WOZ and in the US zone respectively. Figure 6 gives geographical illustration of these robustness checks. Our findings for these subsamples show strong and statistically significant shrinkages of municipalities close to the temporary border in in relation with municipalities that are comparable in terms of distance to the German border. Surprisingly, the effects in column (5) is much stronger than the effect for the entire US zone. When we enlarge our sample in column (6) we obtain results that are more similar to the entire US zone results – this is mainly because the relative prosperous agglomerations of Linz and Wels belong now partially to our border group. This reduces the temporary border effect towards the overall US zone level. To sum up, we conclude that market access to Western core markets can not explain the relative shrinkage of temporary border municipalities.

23 Matching on pre-WW2 municipal characteristics

As shown in the summary statistics in table 1 , time-variant as well as time-invariant municipal characteristics differs for municipalities in the WOZ. Even though our Fixed Effects method would control for time-invariant characteristics, matching increases the credibility of our results. Therefore, we match border municipalities within 75 kilometers to the temporary intra-Austrian border with WOZ municipalities further away than 75 kilometers that are – in terms of observable characteristics – as similar as possible. We apply a kernel matching approach based on the propensity score. Table 13 summarizes the treatment effects based on several samples of observable pre-WW2 characteristics. Column (1) contains border and matched-control municipalities in the entire WOZ, column (2) does the same procedure for the US and UK zone. We find results that are mostly comparable or even similar with the unmatched results in tables 2 and 9. Hence, most municipal pre-WW2 characteristics does not influence our results. In other words, our Fixed Effects strategy does its job very well. However, one time-invariant municipal characteristics leads to massive larger relative shrinkage of border municipalities. When we match on municipal shares of amenable land (share of settlement area) we get much higher negative temporary border effects. Amenable land is defined as land that can be used for buildings and arable land whereas alpine zones does not belong to. We conclude that the largest deviations between border and control municipalities might be observed for alpine municipalities.

Effects on district level

Further concerns that might challenge our main findings are potential differences in average municipal size between border the control group as well as spatial spill-over effects between regional economic centers and its surrounding municipalities. We address this issue by estimating relative population growth on the district level. Table 14 reports these results. Columns (1) - (2) contain different district distance cut-off thresholds to the temporary border intra-Austrian border. The relative post-WW2 shrinkage effect of districts that are located closer to the temporary border is smaller than in our main findings for municipalities. This is not surprising, since smaller municipalities relatively shrink more which leads to upward biased results in our main specification. Nevertheless, a district close to the temporary border relatively shrinks of about 23% (-0.37% p.a.) during the overall post-WW2 period. This finding is robust for different district distance cut-off thresholds. Column (4) in table 14 shows the relative border district shrinkage in different post-WW2 time periods. On the district level, the statistically significant decline of border districts does not start before 1951. Last but not least, the relative shrinkage came to an end in the last decade under consideration. This implies for the overall sample, that the spatial shock of a potential division of Austria became invisible in recent years but affects district’s growth pattern for more than 40 years.

24 WOZ contra SOZ

Finally, we compare the relative shrinkage of temporary border municipalities in the WOZ with the relative shrinkage of temporary border municipalities within the SOZ. We expect that the expectation of a market based economy in the WOZ leads to relocation of production facilities and inhabitants based on market access considerations. In opposition, we should not observe such a pattern within the SOZ because market access considerations of firms and inhabitants would not pay off in a (expected) planning economy.

Column (1) and (2) in table 15 show the relative municipal population growth for the entire post- WW2 period in the WOZ and SOZ with a distance threshold of 75 kilometers to the temporary border. Municipalities in the SOZ that are located within 75 kilometers face no statistically significant relative decline in population compared to SOZ-municipalities further away than 75 kilometers to the temporary border. However, the 75 kilometer distance threshold approximately halve the SOZ and also divides its booming regions in Lower Austria along the Danube. We therefore repeat our estimations in column (3) and (4) by taking a border distance cut off threshold of 50 kilometers for municipalities in the WOZ and SOZ. This reduces the estimates for the WOZ. But again, we can not find any statistically significant relative decline of temporary border municipalities in the SOZ. This findings support our considerations with respect to the importance of market access for the WOZ, but not for the SOZ.

8 Conclusion

25 References

Erickson, E. L. (1950). The zoning of austria. Annals of the American Academy of Political and Social Science, 267, 106–113.

Iber, W. M., Karner, S., Knoll, H., Pickl, O., Prügger, H., Ruggenthaler, P., Stern, S. and Wonisch, A. (2008). Die Rote Armee in der Steiermark: Sowjetische Besatzung 1945, vol. 1. Ludwig Boltzmann-Institut für Kriegsfolgen-Forschung, Graz – Wien – Klagenfurt).

Redding, S. J. and Sturm, D. M. (2008). The Costs of Remoteness: Evidence from German Division and Reunification. American Economic Review, 98 (5), 1766–1797.

—, — and Wolf, N. (2011). History and industry location: Evidence from german airports. The Review of Economics and Statistics, 93 (3), 814–831.

26 9 Appendix

9.1 Appendix: Historical facts

Table 8: Historical dates in post-war Austria

Date/Year Effects on the intra-Austrian Demarcation Line 08. Mai 1945 Capitulation of Nazi Germany and preliminary occupation 09. July 1945 Decision on the definite occupation zones 23./24. July 1945 Allied troops withdraw to their assigned occupation zones Crossing the zone border was impossible 01. Sept. 1945 Official occupation zones enter into force 30. Sept. 1945 Postal and train services over zone borders Cars and horse-drawn wagons are allowed to cross borders, but were strictly controlled 25. Nov. 1945 First general (parliamentary) election 28. June 1946 2. control agreement by the allies to remove existing border restrictions 14. Oct. 1946 First preliminary national census based on food vouchers 11. April 1947 Free movements of goods but still border controls 09. Oct. 1949 Second general (parliamentary) election 06. May 1951 First general (presidential) election 01. June 1951 First regular national census after WW2 09. Sept. 1953 Abolition of border controls between SOZ and WOZ 22. Febr. 1953 Third general (parliamentary) election 25. Oct. 1955 Austrian State Treaty to re-establish Austria’s Independence (jointly signed with the four allies (US, GB, France and USSR) and the Austrian Government) Abolition of zone border and official end of the occupation period 13. May 1956 Fourth general (parliamentary) election 1961 second regular national census after WW2 28. Nov. 1962 Fifth general (parliamentary) election 1971 Third regular national census after WW2 10. Oct. 1971 Sixth general (parliamentay) election

Note: Columns with regular letters indicates dates where political agreements have been made or where cross border restriction have been abolished. Columns with italic letters are dates where we get specific information about municipal population data.

27 9.2 Appendix: Data description

Population data (incl. electorate)

Population data for the municipality and district level based on 2014 territorial status are taken from Statistic Office of Austria (http://www.statistik.at). For the years 1946 and 1948 we obtained municipal population data based on food vouchers (Gemeindeverzeichnis von Österreich 1946, 1948 ). For the years 1949, 1951, 1953 and 1956 we approximate municipal population based on the change of total municipal electorate for parliamentary (1949, 1953, 1956) and presidential elections (1951). These hand-collected municipal data are transformed to the 2014 territorial status based on a survey about mergers and dissolutions of municipalities since 1945 (Auflösungen bzw.Vereinigungen von Gemeinden ab 1945 ; see: http://www.statistik.at). The number of municipalities was reduced from 4,118 in 1946 to 2,357 municipalities in 2014.

Municipal characteristics

Municipal characteristics such as the municipal total area, share of land that is amenable to settle- ment (settlement area) and municipal landscape characteristics are taken from the Blick auf die Gemeinde from the Statistic Office of Austria (http://www.statistik.at/blickgem/index.jsp). In- formation of the latitude and longitude of Austrian municipalities are taken from Lücking (2014) (http://opengeodb.org/wiki/OpenGeoDB_Downloads). Municipal distance to the temporary intra- Austrian border is measured as the shortest great-circle distance between each municipality and any municipality that is located within the WOZ direct at the temporary intra-Austrian border.

Age cohorts, birth rates and migration

Age cohorts on the district level for the year 1939 are taken from the population census 1939 of the German Reich (Volks-, Berufs- und Betriebszählung vom 17. Mai 1939: Alpen- und Donau- Reichsgaue (ohne Wien), Band 557, Heft 28). Age cohorts for 1961 are obtained from the Austrian population census 1961 (Die Ergebnisse der Österreichischen Volkszählung vom 21. März 1961, Heft 2–10). Pre-WW2 district wise gross birth rates are taken from the Statistical Yearbook for the natural population dynamics and marriage behavior (Die natürliche Bevölkerungsbewegung in Österreich in den Jahren 1933 bis 1936 & Österreichische Heiratsordnung 1933/34). Post-WW2 gross birth rates on the district level are taken from the Statistical Yearbook of the Republic of Austria (Statistisches Jahrbuch für die Republik Österreich) for the year 1948 and from the Statistical Yearbooks for the natural population dynamics (Die natürliche Bevölkerungsbewegung) for the years 1951 to 2001. Dis- trict wise net migration per decade from 1951 to 2001 are obtained from the Statistic Office of Austria (http://www.statistik.at).

28 Resident’s occupation and workforce

Pre-WW2 industrial occupation shares on the district level are taken from the Austrian population census 1934 (Die Ergebnisse der Österreichischen Volkszählung vom 22. März 1934, Heft 3–11). Occupation shares for 1961 are obtained from the Austrian population census 1961 (Die Ergebnisse der Österreichischen Volkszählung vom 21. März 1961, Heft 2–10). District wise workforce for 1939 in industrial sub sectors are taken from the population census 1939 of the German Reich (Volks-, Berufs- und Betriebszählung vom 17. Mai 1939: Alpen- und Donau-Reichsgaue (ohne Wien), Band

557, Heft 28).

European Reconstruction Program (ERP)

Industrial specific ERP investment are obtained from the Austrian Institute of Economic Research (Monatsberichte des österreichischen Instituts für Wirtschaftsforschung (WIFO), 1953, Heft 6). ERP funds per worker on the provincial level are from Hofbauer (1992).

29 9.3 Appendix: Results

Table 9: Intra-Austrian border effects in different Western occupation zones for the entire post-WW2 period

Dependent Variable: Annualized Population Growth WOZ US and UK Zone US Zone UK Zone (1) (2) (3) (4) Border x postWW2 -0.47*** -0.24*** -0.28*** -0.18* (0.04) (0.05) (0.05) (0.11) Constant 0.24*** 0.30*** 0.39*** 0.19*** (0.03) (0.03) (0.04) (0.03) Year Effects YES YES YES YES Municipality FE YES YES YES YES No. of Obs. 20,804 16,016 9,856 6,174 R-Squared (adj.) 0.14 0.14 0.07 0.29 Note: Data are the same as in table 2 . Border and postWW2 are defined accordingly. Column (1) replicates the baseline results with all municipalities in the Western occupation zone. Column (2) includes municipalities in the US and GB occupation zones. Columns (3) and (4) include municipalities in the GB and US occupation zones separately. The Fixed Effect estimation uses time and municipality fixed effects. Standard errors are heteroskedasticity robust. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

Table 10: Intra-Austrian border effects during the period of occupation until 1955 (1956)

Dependend Variable: Annualized Population Growth WOZ US and UK Zone US Zone UK Zone (1) (2) (3) (4) Border x Occupation -0.38*** -0.47*** -0.68*** -0.12 (0.06) (0.06) (0.07) (0.10) Constant 0.25*** 0.25*** 0.37*** 0.05 (0.03) (0.03) (0.03) (0.05) Year Effects YES YES YES YES Municipality FE YES YES YES YES No. of Obs. 13,374 10,296 6,336 3,969 R-Squared (adj.) 0.17 0.18 0.11 0.33 Note: Data are the same as in table 4 . Border is defined accordingly. Occupation is a dummy that equals 1 for the 3 periods of occupation (1939-1946, 1946-1951, 1951-1956) and 0 for the periods prior to 1939. Column (1) indicates the overall border effect during occupation for the entire WOZ (from 1939 until 1956). Column (2) includes municipalities in the US and UK occupation zones. Columns (3) and (4) include municipalities in the UK and US occupation zones separately. Standard errors are heteroskedasticity robust. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

30 9.4 Appendix: Channels of the long-lasting border effects

Table 11: Investments of the European Reconstruction Program (ERP) from 1948 until March 1953 in industrial sub sectors

ERP Funds Share of workforce in 1939 Distance to temp. border Difference Mill. US-$ in % < 75 km > 75 km = (3) - (2) (1) (2) (3) (4) (5) Industry (w/o construction) 5131.6 100.00 100.00 100.00 0.00 Electricity Production 1962.1 38.24 na na - Mining 302.3 5.89 6.38 3.32 3.07 Iron, Metal and Steal 1339.0 26.09 29.96 21.57 8.39** Chemicals 197.2 3.84 1.07 1.68 -0.61 Textiles 244.4 4.76 16.56 26.16 -9.60*** Paper 734.1 14.31 3.15 3.90 -0.74 Other sub sectors 352.8 6.88 42.88 43.38 -0.50 Note: Other sub sectors contains industrial sub sectors that receive less than 3% of industrial ERP funds. These are: Electric industry, Glass, Wood, Food, Leather and Shoes and other industries. Column (1) and (2) shows investment flows into industrial sectors in Millions 1953-US-Dollar and as percentage. Column (3) and (4) show the district wise shares of workforce in 1939 in the industrial sector (without construction). Column (3) does it for districts which capital cities are located within 75 kilometers, column (4) with districts which capital cities are located further away than 75 kilometers to the temporary intra-Austrian border. Column (5) gives the significance level of a t-test comparing the means of district wise workforce in the industrial sub sectors. Means that are significantly different from border districts with control districts are denoted by the following system: *10%, **5%, and ***1%.

Table 12: Total ERP funds per province until 1952 per industrial worker

Province ERP-Funds per in- dustrial worker Provinces in WOZ with border municipalities Styria 14,600 Upper Austria(US Zone 21,800 only) Provinces in WOZ without border municipalities Carinthia 28,000 Salzburg 66,000

Provinces in SOZ Lower Austria 5,500 Vienna < 5,000 Note: Numbers are from Hofbauer (1992) in Austrian-Schilling based on „Jahr- buch der Handelskammer Niederösterreich für das Jahr 1952“(issue cannot be found).

31 9.5 Appendix: Robustness checks

Table 13: Matching on pre-WW2 municipal characteristics

Dependend Variable: Annualized Population Growth WOZ US and UK Zone Matching on: (1) (2) Border x postWW2 Population 1939 -0.47*** -0.25*** (0.04) (0.05) Population growth 1869-1939 -0.48*** -0.25*** (0.04) (0.05) Population growth 1923-1939 -0.57*** -0.25*** (0.05) (0.06) Share of settlement area -0.74*** -0.60*** (0.09) (0.13) Density of settlement area 1939 -0.48*** -0.23** (0.09) (0.09) Workforce in 3 sectors 1939 -0.47*** -0.48*** (0.10) (0.13) Population 1939 -0.47*** -0.25*** and population growth 1869-1939 (0.04) (0.05) Population 1939 -0.56*** -0.26*** and population growth 1923-1939 (0.05) (0.06) Population 1939 -0.77*** -0.69*** and share of settlement area 1939 (0.09) (0.13) Population 1939 -0.48*** -0.21** and density of settlement area (0.09) (0.09) Population 1939 -0.44*** -0.44*** and workforce in 3 sectors 1939 (0.10) (0.12) Note: The dependent as well as the explanatory variables are the same as in table 2 . We match municipalities within 75 kilometers to the temporary intra-Austrian border with municipalities further away than 75 kilometers that are similar in terms of various pre-WW2 characteristics by applying a kernel matching approach. Column (1) does it for the entire WOZ, column (2) for the US und UK zones only. Pre-WW2 characteristics are: municipal population in 1939; population growth from 1969-1939 (whole period); population growth from 1923-1939 (interwar period); share of settlement (time invariant area that is amenable for settlement as share of total municipal area) as a proxi for geographical landscape (alpine, hilly, flat); the pre-WW2 population density of settlement area and the district wise share of workforce in the 3 industrial sectors (agriculture, industry, other (mainly services)). The Fixed Effect estimation uses time and district fixed effects. The table reports interaction-term coefficients and the corresponding heteroskedasticity robust standard errors. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

32 Table 14: The effects of the temporary intra-Austrian border on districts

Dependend Variable: Annualized Population Growth Capital’s distance to temp. border (in km) Time Effects < 75 vs > 75 < 50 vs > 75 until 1961 until 2001 (1) (2) (3) (4) Border x postWW2 -0.37*** -0.37*** -0.38*** (0.11) (0.12) (0.10) Border x Year 1939-46 -0.22 (0.32) Border x Year 1946-51 -0.29 (0.22) Border x Year 1951-61 -0.62*** (0.12) Border x Year 1961-71 -0.60*** (0.15) Border x Year 1971-81 -0.37*** (0.13) Border x Year 1981-91 -0.31** (0.16) Border x Year 1991-01 -0.21 (0.16) Constant 0.52*** 0.53*** 0.41*** 0.44*** (0.06) (0.06) (0.06) (0.06) Year Effects YES YES YES YES Municipality FE YES YES YES YES No. of Obs. 714 630 510 714 R-Squared (adj.) 0.51 0.52 0.54 0.52 Note: Border is a dummy that equals one (zero) for a district when its capital city is located within (outside) a certain strip to the temporary intra-Austrian border. The dummy postWW2 is defined as in table 2 . Column (1) compares districts within the WOZ which capital cities are located within 75 kilometers to the temporary intra-Austrian border with districts which capital city further away. Column (2) compares districts which capital cities are located within 50 kilometers with districts which capitol cities are further away than 75 kilometers from the temporary intra-Austrian border. Column (3) and (4) show the temporary border effects with a district’s capital cut-off distance of 75 kilometers. Column (3) show the overall border effects until 1961. Column (4) show the relative decline of districts that are located closer to the temporary intra-Austrian border for each sub period until 2001. The Fixed Effect estimation uses time and district fixed effects. Standard errors are heteroskedasticity robust. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

33 Table 15: Compare intra-Austrian border regions in WOZ and SOZ

Dependend Variable: Annualized Population Growth Strips < 75 km vs > 75 km Strips < 50 km vs > 50 km WOZ SOZ WOZ SOZ (1) (2) (3) (4) Border x postWW2 -0.47*** -0.38 -0.33*** -0.29 (0.04) (0.35) (0.04) (0.21) Constant 0.24*** 0.57*** 0.24*** 0.57*** (0.03) (0.06) (0.03) (0.06) Year Effects YES YES YES YES Municipality FE YES YES YES YES No. of Obs. 20,804 12,138 20,804 12,138 R-Squared (adj.) 0.14 0.01 0.14 0.01 Note: Border is a dummy that equals one (zero) when a municipality is located within (outside) a certain strip to the temporary intra-Austrian border. The dummy postWW2 is defined as above. Columns (1) and (2) compare municipalities for the WOZ and SOZ that are located within 75 kilometers to the temporary intra-Austrian border with municipalities further away than 75 kilometers from the border in the respective occupation zones. Column (1) replicates the main result for the WOZ. Column (2) shows the intra-Austrian border effects within the SOZ with a cut-off distance of 75 kilometers. Columns (3) and (4) compares municipalities for the WOZ and SOZ that are located within 30 kilometers with municipalities further away than 50 kilometers from the temporary border in the respective occupation zones. The Fixed Effect estimation uses time and municipality fixed effects. Standard errors are heteroskedasticity robust. Coefficients that are significantly different from zero are denoted by the following system: *10%, **5%, and ***1%.

Figure 6: Geographic determination of our robustness checks Robustness Checks

25 km

50 km

15.15° East

25 km Köflach 50 km Voitsberg

15.15° East

France United States United Kingdom Soviet Union

Note: The map show the geographic restrictions of our robustness checks in columns (2) - (6) in table ?? . The vertical line shows the 15.15◦ East longitude. The 15.15◦longitude divides Köflach and Voitsberg where a checkpoint between the Royal Army and the Red Army has been in place from May 9 to July 24 1945.Czech Municipalities Republic in the East of the the 15.15◦ longitude are not considered in our robustness check in column (2). The horizontal lines shows 25 and 50 kilometers distances to the German border. The continuous and dotted lines show the location of municipalities in column (3) and (4) whereas the continuous lines show the location of municipalities in column (5) and (6).

34

0 75 km

Place of origin New location