Historical Roots of Racism Working Paper.pdf

                                    Constitutional
                                      Economics
                                       Network

                                    Working Paper
                                        Series
                                  ISSN No. 2193-7214

                                      CEN Paper
                                     No. 06-2018




         The Reversal of Fortune, Extractive Institutions
            and the Historical Roots of Racism

          Matthew Bonick* and Antonio Farfán-Vallespín**

          *Department of Economic Policy and Constitutional Economic Theory,
                   University of Freiburg, Germany.
               E-Mail: matthew.bonick@vwl.uni-freiburg.de

               ** Department of International Economic Policy,
                   University of Freiburg, Germany.
                 E-Mail: antonio.farfan@vwl.uni-freiburg.de



                      October 25th , 2018


University of Freiburg
Institute for Economic Sciences
Department of Economic Policy and Constitutional Economic Theory
Platz der Alten Synagoge / KG II D-79085 Freiburg
www.wipo.uni-freiburg.de
   The Reversal of Fortune, Extractive
Institutions and the Historical Roots of Racism
                     Matthew Bonick∗†

                    Antonio Farf´n-Vallesp´ ‡
                          a     ın

                      October 25, 2018


                         Abstract

      We show that differences in present levels of racism within a sample of former European
    colonies can be traced back to historical institutions. Our identification strategy relies on
    the reversal of fortune, a historical shock capturing the exogenous establishment of differ-
    ent institutions during the onset of European colonization. Using both OLS and multilevel
    analysis, we find extractive historical institutions to be a strong predictor of higher levels of
    racism independent of present and other explanatory factors at the inpidual and country
    levels. We argue and provide evidence this relationship is causal and operates through per-
    sistent internal norms, beliefs and values, resilient to changes in institutional and economic
    circumstances.



1   Introduction
Recent research demonstrates that racism or racial intolerance is a very relevant
phenomenon linked to a number of economic and political outcomes important for
economists. Racism has been shown to affect political preferences such as views on
policies targeting minorities, reduced support for the welfare state, more support
for residential segregation and restrictive immigration policy (Bobo 1991; Charles
2000, 2003; Ford 2006; Dustmann and Preston 2007). It has also been connected
to health disparities, worse labor market outcomes and educational inequalities
for minority groups (Ashraf 1994; Lang, Manove, and Dickens 2005; Goldsmith,
Hamilton, and Darity 2006; Charles, Kofi, and Guryan 2008; Lang and Manove
2011; Lang and Lehmann 2012,; Dickerson and Jacobs 2013; Feagin and Bennefield
2014). Because of the relevance of the results of these many studies, it is prudent
to attempt to understand the factors causing the emergence of racism.
  ∗ University of Freiburg : email: matthew.bonick@vwl.uni-freiburg.de
  † Corresponding author : email : matthew.bonick@vwl.uni-freiburg.de
  ‡ University of Freiburg : email: antonio.farfan@vwl.uni-freiburg.de




                            1
  We are interested in whether racism is a phenomenon determined by short-term
environmental circumstances or whether it is a long-term cultural value that per-
sists over generations. Nowadays, we observe a rise in the number of expressions of
political and social attitudes that could be considered racist. This has lead many
to investigate current societal changes explaining this phenomenon. For example,
Johnston and Lordan (2016) show, mainly among high skilled middle aged work-
men, a 1 percentage point increase in unemployment leads to a 4 percentage point
increase in racial prejudice. However, other studies like Voigtlander and Voth
(2012) show that antisemitic preferences and violence can persist over centuries.
A more robust understanding of the short or long-term origins of racism is impor-
tant if policy makers want to develop effective strategies to mitigate consequences
racism. The effectiveness of different policies could vary if racial attitudes are
mostly motivated by current circumstances than if they are cultural values deeply
ingrained in societies.
  We contribute to this debate by utilizing an empirical strategy that allows us
to overcome important identification problems. Any attempt to determine the
direction of causality between culture, institutions and economic outcomes faces
an important problem of endogeneity since all these variables influence each other.
We handle these issues by using the well-known phenomenon of the reversal of for-
tune which represents the historical reversal of prosperity from pre-colonial times
to the present caused by the exogenous establishment of different forms of in-
stitutions brought by European colonial powers. This strategy is first used by
Acemoglu, Johnson, and Robinson (2002) who show that higher levels of urban-
ization and population density in the 1500s, both proxies for economic prosperity,
to be associated with worst institutions and economic development in the present.
The authors explain this result by arguing the areas with higher population den-
sity were ideal targets for European colonizers to establish extractive institutions
while areas of lower population density received inclusive institutions. The imple-
mentation of these different forms of institutions is the driving factor behind the
reversal of prosperity from 1500s to the present. Given this event is a quasi-natural
experiment capturing the exogenous establishment of institutions, it allows us to
disentangle the causal relationship between institutions and culture. Our results
show, former colonies that historically had more extractive institutions have higher
levels of racism today compared to colonies with inclusive institutions.
  To test our hypotheses, we combine data on colonial institutions with a inpid-
ual and country level measure for racism taken from the World Value Survey. We
operationalize racism by using responses to the question, ”On this list are various
groups of people. Could you please mention any that you would not like to have as
neighbors?”, with racism capturing those who select ”other races” as one of their
answers. For measuring the reversal of fortune and proxying for historically ex-
tractive institutions, we use log of population density and technology in the 1500s.
For current prosperity we use the variables standard in the literature for economic


                     2
development and institutional quality.
  For our baseline results we use two subsamples: former colonies and former non-
colonies. We separate our analysis into these two samples because the literature has
shown the phenomenon of the reversal of fortune is restricted to post-European
colonies and in other regions of the world, historical prosperity is generally a
predictor of higher modern economic development which indicates a persistence of
fortune, not a reversal (Nunn 2014 ; Spolaore and Wacziarg 2013). Thus, if our
identification strategy is valid, we expect to find the opposite relationship between
historical prosperity and our outcomes of interest across these two samples. We
first confirm that a reversal of the fortune occurred in former colonies but not
non-colonies in our data set. Next, we examine if our proxies for historically
extractive institutions predict racism today. The results support our hypothesis
as we find that colonies with more historically extractive institutions have higher
levels of racism today. Additionally, the results, as predicted, do not extend to
non-colonies.
  In order to discard potential omitted variable biases, we control for fraction-
alization, genetic persity, ancestry-adjusted variables, proportion of Europeans
in 1900, proportion of descent from indigenous population, colonial origin, legal
origin, religion, absolute latitude, trust, respect, obedience and control. As a final
step to account for omitted features, we utilize an instrumental variable approach
that produces consistent results. Overall, the outcomes of these control exercises
support our initial findings that historically extrative institutions predict higher
levels of racism in the present.
  Further, we attempt to disentangle the two possible channels of causality be-
tween historical institutions and racism. The first possible mechanism is that
the establishment of historically extractive institutions was a shock which perma-
nently altered cultural norms, beliefs and values in a society toward higher levels of
racism. However, it is also possible that extractive historical institutions only have
an indirect effect on racism. Acemoglu, Johnson, and Robinson (2001,2002) show
that the colonial experience is the main determinant of current levels of quality of
institutions. Further, Berggren and Nilsson (2013) show that better institutional
quality results in less racism. Ergo, extractive colonial institutions shaped levels
of racism via its effect on modern institutions.
  To disentangle whether racism is the product of current circumstances or a per-
sistent internal norm dating back to the colonial experience we implement several
strategies.
  First, we re-run our baseline regressions while controlling for current variables
for institutions, economic prosperity and human capital. If these current measures
are strong mediators and thus, make our historical variables lose significance, there
is evidence colonial institutions only have an indirect effect on racism through
present day features. When accounting for these factors, we find consistent results
compared to our baseline.


                     3
  Next, we argue that if racism is caused by the lower quality institutions, then
those inpiduals with a lower opinion of his countries’ institutions should be more
likely to be racist caeteris paribus. For this, we estimate, at the inpidual level, if
the negative effect of historically extractive institutions remains when we control
for measures of an inpiduals’ confidence in the government and other inpidual
and country level characteristics. Utilizing multilevel analysis, we find, for our
colonial sample, extractive historical institutions predict a greater probability an
inpidual will possess a racist attitude.
  In our final step, we examine inpiduals facing a change of institutional envi-
ronment to see whether their levels of racism depend on their new environment or
on the historical factors of their country of origin. For this, we use a sample of
immigrants in Europe taken from the European Social Survey and test whether
their levels of racism respond to the economic and institutional characteristics of
their countries of residence or rather to the historical level of extractiveness of
institutions in their country of origin. The reasoning behind this approach is when
inpiduals relocate, they bring their internal beliefs with them, a factor which
is independent of their surrounding environment. As a result, if the historical
institutions of an inpidual’s origin continue impact their level of racism after
immigration, we have evidence racism is an internal believe value and norm which
persists even when an inpidual faces a new institutional and economic setting.
  We find, even when accounting for the institutions of an inpidual’s country of
destination, historically extractive institutions predict higher levels of racism for
immigrants coming from a former European colony. When we examine the same
relationship for inpiduals who have migrated from non-European colonies, we
see, consistent with other sections, a reversed or nonexistent relationship.
  The conclusion of our paper is that historically extractive institutions, within
the geographical context of former European colonies, have a causal impact on cur-
rent levels of racism at both the inpidual and cross-country level. Moreover, the
paper also identifies racism as an internal norm, cultural value or belief persistent
to changes in current institutional and cultural environments. While we do not
argue historical institutions are the only factor, past or present, affecting racism,
the robustness of the connection we find cannot be explained by other variables.
  This paper contributes to both the literature investigating the determinants
of racism and the links between history, culture and institutions by finding one
particular case in which a historical change of institutions causally shaped cul-
tural values in the present. These results add to the recent studies showing that
cultural beliefs and values are rooted in historical factors, such as institutions
(Tabellini 2010; Nunn and Wantchekon 2011; Nunn 2012; Alesina, Giuliano, and
Nunn 2013; Spolaore and Wacziarg 2013; Nunn 2014; Alesina and Giuliano 2015;
Guiso, Sapienza, and Zingales (2016)). Finally, we add to the literature evaluating
the long-term impact of colonization on current societies by showing extractive
colonial institutions effect extends beyond current institutions to culture. A cul-


                      4
ture of racism could have helped perpetuate the lack opportunities for prosperity
and democracy in former extractive colonies.
  After the introduction in section 1, section 2 presents the theoretical context,
section 3 describes the identification strategy, section 4 presents the results of the
empirical analysis identifying the link between racism and historical institutions.
Section 5 examines if the effect of historical institutions on racism operates through
internal norms, beliefs or values and section 6 concludes.


2   Theoretical Context
In this section, we define racism, describe the reversal of fortune and present the
two alternative explanations for how historical institutions shaped by the reversal
of fortune, could impact levels of racism in the present.

2.1  Definition of Racism
Racism is a decision-making heuristic or rule of thumb for decisions on interactions
with and treatment of inpiduals belonging to other racial groups in a world of
uncertainty. These decision-making heuristics appear as values, beliefs or social
norms which evolve and are passed to the next generation through a process of
natural selection shaped by the comparative benefits of using alternative rules of
thumb. (Boyd and Richerson 1985, 2005, Nunn 2012).
  The heuristic is based on the belief that all inpiduals can be classified in
racial groups and someones identity to that group conveys fundamental informa-
tion about them. Thus, for a racist, the application of this heuristic guides their
treatment, choices, preferences, beliefs and what social norms are applied when
engaging in interactions with people of another race in an economic, political or
social setting. This does not mean the use of such a heuristic always leads to
optimal behavior but does reduce the cost of obtaining information on the proper
course of action in a given situation (Nunn 2012). There are several potential
motivations for inpiduals to be a racist and so, how racism manifests will not
be homogenous across inpiduals, circumstances, communities or even countries.
As a result, independent of the motivation for racism, there is justification for the
unequal treatment of someone from a different racial group and that this discrim-
ination is defensible or even preferable. For example, this can cover labor market
decisions driven by statistical discrimination, i.e. assumed differences in racial
productivity, all the way to views on racial superiority and acceptance of social
interactions such as interracial marriage. Although inter-racial violence is not a
necessary part of our definition, racist beliefs, preferences and rules can, in cer-
tain cases, justify or even demand aggressive or inhuman treatment of inpiduals
belonging to certain racial groups if that particular group is considered to be a
threat or hostile. Given all these factors, our understanding of racism fits within

                     5
the definition of culture by Guiso, Sapienza, and Zingales (2006), as ”those cus-
tomary beliefs and values that ethnic, religious, and social groups transmit fairly
unchanged from generation to generation”.

2.2  The Reversal of Fortune
Engerman and Sokoloff (1997), henceforth ES, and Acemoglu, Johnson, and Robin-
son (2002), henceforth AJR, argue and empirically show that European coloniza-
tion played a vital role in shaping the path of development of a number of countries
through historical political, economic and educational institutions. They hypothe-
size, where European colonizers found mineral resources, a large and concentrated
native population, agricultural land suitable for large scale plantations or when
they came upon a diseased environment, a lower number of Europeans would set-
tle in these areas leading to the emergence of extractive institutions. When the
opposite conditions were present, more Europeans immigrated resulting in inclu-
sive institutions. This hypothesis is further corroborated by Easterly and Levine
(2016), who indicate the pattern of European immigration was exogenously de-
termined by similar geographical factors. Extractive institutions are described as
areas where the elite established rules, laws and other government policies to in-
stitutionalize their economic and political advantage. According to AJR and ES,
such policies included: restricted access to democracy, lack of political rights for
most segments of society, unequal enforcement of property rights, a lower provision
of public schools, unequal access to financial institutions and a general lack of eco-
nomic opportunity for all. AJR empirically demonstrates, territories which were
historically more prosperous saw the establishment of extractive institutions with
these institutional characteristics persisting to the present. In turn, extractive
institutions hindered economic development manifesting in lower GDP per capita
in 1995. Areas which had a less concentrated native population saw a large share
of European immigration. In this scenario, inclusive institutions were established,
meaning, a set of institutions providing greater political access, legal protection
and education to a larger share of the population. These institutions persisted
to the present and were a driving factor leading to greater economic development
in these countries. Thus, through the establishment of different forms of histor-
ical institutions, the pattern of more historical prosperous societies now being
less economically developed compared those who were historically less prosperous
emerged. A pattern AJR refers to as the reversal of fortune.

2.3  The Reversal of Fortune, Institutions and Racism
In this section, we present two possible hypotheses for how the reversal of fortune
affects current levels of racism in former European colonies. The first hypothesis is
that historical institutions impacted levels of racism indirectly through their per-


                     6
sistent effect on modern institutions 1 . The second hypothesis is the establishment
of historically inclusive or extractive institutions was a shock which permanently
altered equilibrium level of racism within societies which persists across genera-
tions.

2.3.1  Indirect Impact Through Modern Institutions

There are several reasons to think that the levels of racism in a society might be fun-
damentally determined by contemporary conditions. This conclusion is supported
by the literature linking racial attitudes to education and the modern functioning
of institutions (Hello, Scheepers, and Gijsberts 2002; Berggren and Nilsson 2013).
  According to Berggren and Nilsson (2013, 2014), a society with the rule of law
will have less racism because properly enforced laws ensure legal rules apply equally
to everyone. As a result, there is no need to fear the actions of other racial groups
since violators will be punished independently of their racial group which will foster
interracial interaction. Economic freedom and economic prosperity ensuing from
good institutions might also foster interaction between members of different racial
groups leading to less racism.
  Berggren and Nilsson (2013) stipulate economic institutions promoting market
exchange, provide incentive structures which can lead to less racism through three
possible mechanisms. First, because of repeated successful transactions with other
races in the market place, inpiduals begin to internalize positive beliefs of other
races 2 . Second, in seeking mutually beneficial exchange a racist inpidual might
have incentives to substitute his racially-based heuristic by another more accu-
rate heuristics that would make better use of available information and would not
discriminate based on race. Becker (1957), for instance, under the condition of
competitive markets, suggested that non-discriminating firms possess a compet-
itive advantage over discriminatory firms in terms of productivity and access to
a wider set of customers and workers. Discriminatory firms will hence be forced
to change their racially-based heuristic or be driven out of the market. Third,
markets provide a mechanism for taking fragmented societies consisting of small
set of closed groups and transforming them into to a set of multiple interconnected
inpiduals. This can lead to the expansion of social capital once particularized
by racial groups to more a generalized application. These interconnected networks
are expected to lead to more trust across inpiduals belonging to different racial
groups facilitating the generation of values and social norms shared by larger pro-
  1 In our case, we use a broad definition of institutions, meaning it encompasses political, legal, economic and

educational institutions.
  2 This argument is further supported by the contact theory. Allport (1954) developed the Intergroup Contact

Theory, which proposes that, under the appropriate conditions, interpersonal contact is one of the most prominent
ways to reduce prejudice. The theory argues, when there are encounters across racial and ethnic lines, the majority
group members can communicate with minority group members and are then better able understand these groups
resulting in a diminishment of their previously held prejudice. Additionally, other theories suggest interracial
interaction might increase trust across groups and reduce the perception of other groups as a threat (Blumer
1958)



                            7
portion of the population. When social capital expands across racial groups, the
expression of racist attitudes may become a violation of social norms coming at
a social or economic cost. This phenomenon changes the incentive structure for
exhibiting a racist attitude which will making it less attractive to possess.
  The literature on racism also shows more educated inpiduals tend to be more
tolerant. The relationship indicates educated inpiduals keep less social distance
from ethnic minorities, meaning they possess a greater willingness to engage with
other ethnic groups, are less likely to be prejudiced and ethnocentric (Selznick
and Steinberg 1969; Hyman and Wright 1979; Jackman and Muha 1984; Hello,
Scheepers, and Gijsberts 2002). Some of the main explanations of these findings
are that education makes inpiduals more liberal, open-minded and less likely
to view other races as a potential competitor for valuable resources (Hyman and
Wright 1979; Hello, Scheepers, and Gijsberts 2002; Hello, Scheepers, and Sleegers
2006). As we can see, there is an abundance of evidence linking higher levels of
education to less social distance and fear of other races, meaning a more educated
society is likely to have more interracial interaction and cooperation leading to a
lower racist equilibrium.
  In our case, it has been already shown in the literature that former colonies with
more extractive institutions have lower quality political, economic and educational
institutions today (AJR and ES), therefore, it is to be expected to find an indirect
positive correlation between the degree of extractiveness of colonial institutions
and the levels of racism nowadays via modern institutions.

2.3.2  Direct impact of historical institutions on racism

Another explanation is that racism is a deeply ingrained cultural value in societies
and is persistent to external changes which is supported by increasing number of
studies that show a link between historical events and current cultural values exist
3
 For this hypothesis, we argue the establishment of extractive or inclusive insti-
tutions by colonizers was a shock that permanently altered the equilibrium level
of racism in societies and that this cultural heuristic persists across generations
independent of changes to other environmental factors. In this section, we present
two arguments for how the shock to institutions brought by the European colonial
powers would shift the equilibrium level of racism.
  In our first argument, members of the European colonial elites who controlled
extractive institutions purposefully shaped beliefs, cultural norms and established
social hierarchies which promoted racism as a mechanism pide to the popula-
tion in order to maintain political and economic power and extract resources 4 .
  3 See Buggle(2016) Becker et al. (2014) Guiso, Sapienza and Zingales(2016), Nunn and Watchekon (2011) or

Alesina, Giuliano and Nunn (2013) among others.
  4 This argument presupposes that colonial powers were able to influence cultural values against the interest of

the value-holders, which is contrary to the standard views that inpiduals choose their own values in order to
optimize their own utility and that of their offspring (Tabellini, 2008). We believe that it was feasible for colonial
elites to influence cultural values given the extent of their control over the administration, civil organizations,



                             8
Acemoglu, Robinson, and Verdier (2004) stipulate kleptocrats can be successful
in extracting resources of the greater population if they can prevent coordination
among the exploited. The utilization of such a strategy is commonly known as di-
vide et impera 5 . There is evidence that colonial powers made deliberate use of this
tactic in their interaction with different ethnic entities in colonized countries. For
example, as Acemoglu et al. (2001) argue, extractive institutions established or
perpetuated ethnic hierarchies within their institutional structures providing spe-
cial benefits and power to certain ethnic groups. This, in turn, gave incentives for
groups higher in the economic and political hierarchy to maintain the status quo
in fear of losing these advantages even after independence from colonial powers.
If implemented effectively, this strategy could encourage rivalries, grievances and
sow distrust and animosity across ethic groups hindering the cooperation of local
political entities against the colonizing powers 6 . Additionally, it is conceivable,
colonial elites applied these tactics to the education system they controlled. For
example, if they can instill fear, hatred and notions of racial or ethnic superiority,
it will make coordination among these different communities against the ruling
elites more difficult due to cultural barriers to cooperation.
  There is further evidence suggesting that regimes can actually instill certain val-
ues on their population, in particular, there is an increasing literature showing that
communist regimes affected the preferences of their citizens. For instance, Alesina
        ˜1
and Fuchs-SchA 4 ndeln (2007) show that communism affected the economics pref-
                     ˜
erences of Eastern Germans. Fuchs-SchA 1 ndeln and Masella (2016) try to identify
                      4
the channel of this effect and show that an additional year of exposure to socialist
education in Eastern Germany had a significant impact on education and labor
market outcomes. Cantoni et al. (2017) show that a change of curriculum in
P.R. China ”led to more positive views of China’s governance, changed views on
democracy, and increased skepticism toward free markets” .
  In the specific application to racism, when extracive institutions were present,
powerful elites had a greater incentive and opportunity to instill racist attitudes
in society. The reason extractive elites had a greater incentive is because, as ES
highlights, the majority of their income was derived from the exploitation of slaves
or the indigenous population through forced labor. As a result, by promoting a
racist cultural heuristic which inhibits cooperation and interaction across these
racial lines, they were protecting their interests. Additionally, by establishing and
institutionalizing racial beliefs and hierarchies, they provided incentives for the
maintenance of the cultural and institutional status quo even by certain members
education, art, and religion. For instance, Acemoglu et al. (2014) suggests that elites could take control of civil
societies organizations and use them as a mechanism to shape and control culture in their own advantage.
  5 Posner, Spier, and Vermeule (2010) provide a taxonomy of different game theoretical settings demonstrating

the logic of pide et impera and cites different historical examples of its implementation.
  6 An example of this tactic can be seen in the case of Rwanda. Nunn (2014) argues, while they do point

out there is still debate on this theory, colonial rules exacerbated the already present tensions along ethnic lines
by implementing policies to purposefully aggravate the already present class differences between the Hutus and
Tutsis.



                             9
of the non-elite as they fear losing the advantages they possess compared to other
races below them in society. For the case of inclusive institutions, these forces
would not be present leading to a different equilibrium level of racism compared
to societies that faced extractive institutions. Acharya, Blackwell, and Blackwell
(2016) provide empirical support for this argument as they show white inpiduals
in counties with a higher intensity of slavery in 1860 in the South of the United
States are more likely to express both policy preferences and attitudes against
black inpiduals today 7 . They argue in counties with a higher intensity of slavery,
Southern whites had a greater political and economic incentives to perpetuate the
prevailing racist norms and institutions to maintain control over freed black slaves.
In turn, this lead to the pergence of county level political and personal racial
attitudes seen in the present. Overall, it is plausible that the elites controlling
colonial institutions had the incentive, ability and opportunity to establish, or
at least exacerbated racism in extractive colonies. These actions contributed to
both the preservation of these institutions and the pergence in levels of racism
compared to societies with a different institutional environment seen today.
  The second argument stipulates, while not a deliberate act, the differential
shock to institutional quality endogenously lead to perging equilibrium levels
of racism. The establishment of inclusive institutions created an environment
conducive for the formation of a more educated, open, cooperative, and thus, a
more racially tolerant society. For the case of extractive institutions, the opposite
forces would be present. The argumentation for why well-functioning political,
economic and educational institutions would endogenously lead to lower level of
racism applies the logic from the section indirect effect through modern institutions
with one important difference. In this case, the effect of institutions is rooted in
history not the present. Since we argue racism is a internal norm, value and belief,
once this cultural heuristic was established, it becomes persistent even if quality of
institutions changes at a later point. Such a hypothesis is consistent with the work
of Guiso, Sapienza, and Zingales (2016), Putnam, Leonardi, and Nanetti (1993)
and Tabellini (2010).
  Jha (2013) identifies such a phenomenon by showing cities and towns in me-
dieval India participating in overseas trade generally saw less Hindu-Muslim riots
in the late 19th and early 20th centuries. Using the presence of natural harbors as
an instrument for trading cities, Jha stipulates, interaction across religious lines
were incentivized because Muslims could facilitate Hindu’s access to markets in
the Middle East. Thus, the benefits of cooperation across religions were greater
in locations with access to international trade. Jha argues, the institutional envi-
ronments which supported exchange and interaction between Hindus and Muslims
had a higher likelihood of peaceful coexistence in the future.
  There are additional studies showing that certain cultural values probably arose
and persisted as an adaption to weak institutions. Grosjean (2014) finds that in
 7 This  includes opposing affirmative action, and having racial resentment and colder feelings toward blacks



                            10
the United States ”a culture of violence was transmitted to subsequent generations,
but only in the South and, more generally, where historical institutional quality
was low. The interpretation is that the Scots-Irish culture of honor prevailed
and persisted as an adaptive behavior to weak institutions.” Anderson, Johnson
and Koyama (2017) show that persecution of Jews in pre-modern Europe was
correlated with weather shocks but also that the intensity of this effect was stronger
in weaker states. We assume that a similar effect might have taken place in colonies
with extractive institutions. Those colonies with worst institutions might have
experienced more tensions between racial groups compared to those with better
institutions which would lead to a higher levels of racism.
  Clearly, these two arguments, whether racism is deliberate or not, are not mu-
tually exclusive. They both might have taken place and whether it was one or the
other does not affect our results or alter our methodology.

2.3.3  Persistence of racism after the initial shock

The next question is how did the level of racism during colonial times, either
instilled via colonial authorities or emerged as an adaption to colonial institutions,
managed to persist over the Centuries and decades after independence. There is an
extensive literature covering the issue of transmission of cultural values, both from
a theoretical and from an empirical perspective. See for instance an exhaustive
overview of the literature on transmission of cultural values Bisin and Verdier
(2011) or Nunn (2012).
  The literature has identified two main channels of transmission: horizontal
and vertical. Vertical transmission takes place from parents to their offspring.
See for instance Tabellini (2008), Fernandez and Fogli (2009) or Giavazzi, Petkov
and Schiantarelli (2016) among many others. Horizontal transmission takes place
among peers via imitation and socialization (see again Bisin and Verdier (2011) for
a survey.) Oblique transmission is considered by Giavazzi, Petkov and Schiantarelli
(2016) as a case of horizontal transmission in which the transmission stems from
persons of a previous generation but different than the parents. This is the case
of institutional transmission of values via teachers, preachers or public figures.
  Some papers have also addressed the question of whether cultural values per-
sist due to the persistence of institutions determining them or due to the internal
persistence of cultural values. One example is Voigtlnder and Voth (2012) who
show that the different local levels of anti-Semitism in Germany in medieval times
persisted until modern times, controlling for all potential economic and institu-
tional confounding effects, hence supporting the hypothesis that cultural values,
particular those concerning the interaction with other racial or ethnic groups, can
persist over 600 years. Therefore, there is sufficient evidence to believe that a
transmission of racism over generations, relatively independently from the partic-
ular institutional and economic setting, might be possible.


                     11
3   Data and empirical strategy
3.1   Main variables of interest
Our measure for racism is based on one question in the World Value Survey: ”On
this list are various groups of people. Could you please mention any that you would
not like to have as neighbors?” The answer is coded 1 if the inpiduals mention
people of a different race their response and 0 if they do not 8 . For the country-
level, we average this variable from inpidual level responses, by country, over all
waves 9 .
  To capture historical institutions, we use 2 proxies, the log of population density
and technological development in the 1500s. The measure for log of population
density in the 1500s is taken from Acemoglu, Johnson, and Robinson (2002). One
question may be, why are we not using the variable, as Acemoglu, Johnson, and
Robinson (2002), urbanization in the 1500s? Given the already limited size of our
sample for racism and the sparseness for the data points for urbanization, we chose
to use another proxy that results in a larger set of observations. Additionally, while
population density is a widely used measure for pre-colonial development, some
argue there are drawbacks with regards to the measures quality and theoretical
correctness in capturing historical prosperity. As a result, we use another variable
which captures levels of development in the 1500s (Chanda, Cook, and Putterman
2014).
  The second proxy is the level of technology in the 1500s, taken from Comin,
Easterly, and Gong (2010) via Chanda, Cook, and Putterman (2014). The mea-
surement index is based on the presence of 24 technologies across 5 different sectors
within a given territory. The measure captures the availability of such technology
around the 1500s in sectors which includes communication, agriculture, military,
industry and transportation before European contact and colonization. Studies
have demonstrated that this variable predicts levels of current income and higher
population density in the 1500s (Comin, Easterly, and Gong 2010; Ashraf and
Galor 2011). In terms of our sample selection, in order to replicate AJR’s (2002)
results, we use the same classification for what is and what is not a post-colonial
country. As a result, only territories which were colonized by Western European
powers are considered in our colonial sample. See the appendix for the list of
  8 Given our measure for racism is based on the question,”On this list are various groups of people. Could you

please mention any that you would not like to have as neighbors?” and, this question or other follow up questions
do not ask the motivation for such an answer, we cannot determine the reasoning behind their choice. However,
the answer to this question does indicate an inpidual attaches some negative utility to having someone of another
racial group as a neighbor in a broad sense. This means, we can identify inpiduals attaching different benefits
to proximity or interaction with people based on racial distinctions which we interpret as the use of a racist base
heuristic. As a result, we can identify an inpidual possessing a racially based heuristic without determining its
underlying motivation. Our measure of racism is consistent with our definition, meaning, the identification of the
presence of a broadly defined racist heuristic.
  9 The waves include: 1981-1984, 1990-1993, 1995-1998, 1999-2004, 2005-2009 and 2010-2014. Additionally,

since many countries only have one data point, we choose to utilize averages across all waves. This is a common
strategy used in the literature on generalized trust (Bjornskov and M`on 2013).
                                    e



                            12
countries considered as colonial and non-colonial. Additionally, we restrict our
analysis to countries that always have data points for log of population density
in the 1500s, technology index in the 1500s, log of GDP per capita in 2000 and
technology index in the year 2000. One note on the use of our proxies for histori-
cal institutions is that these measures are not precise and thus, when we refer to
historical institutions, we are speaking about institutions in very broad terms.

3.2  Identification strategy
The estimation of the causal impact of institutions on racism must handle several
econometric problems, the first of which is reverse causality. We have already
mentioned that racism can affect policy preferences which, in turn, affect the
choice of institutions. Therefore, simply estimating the correlation between current
institutions and current levels of racism would not allow us to claim causality. In
order to avoid the problem of reverse causality, we use the reversal of fortune, or the
imposition of extractive institutions on colonies that were more prosperous in the
1500s and inclusive ones on less prosperous colonies in the 1500s. The reversal of
fortune acts as a quasi-natural experiment that allocated institutions of different
quality to former colonies based on observable characteristics. This observable
variable, namely prosperity in the 1500s, is known to be negatively correlated
with prosperity in the present among former colonies (Acemoglu, Johnson, and
Robinson 2002). This factor is thus a good proxy for the degree of extractiveness of
colonial institutions. Another important component of using the reversal of fortune
as an identification strategy is that the literature has shown it is a phenomenon
limited to former European colonies. In other geographical contexts population
density in the 1500s is a predictor of higher prosperity, not lower (Nunn 2014). If
we are indeed capturing the impact of historically extractive colonial institutions
on racism, we should not find the same relationship between racism and log of
population density or technology in the 1500s across our colonial and non-colonial
samples.
  Second, the relationship between institutions and racism is likely to be affected
by several other factors, hence the results could be biased. Therefore, we will
control for potential confounding factors that the literature has identified. We are
interested in testing whether our proxies for extractiveness of colonial institutions
remain significant to the inclusion of possible controls and given the limited size
of our sample, we choose to use parsimonious specifications and only add controls
one by one.
  Third, our goal is to differentiate whether racism is a cultural value that reacts
to environmental characteristics or it is a persistent internal norm. For this, we
will conduct three tests to see whether our proxy for extractive colonial institutions
remains significant in predicting racism even after controlling for different proxies
for current institutions.


                     13
4   Empirical Analysis
4.1  Reproduction of the Reversal of Fortune
The first step in our empirical analysis is to confirm that a reversal of fortune
took place in our former European colonial sample and that such a relationship is
reversed or non-existent for non-colonies.
  In table 1, using OLS, we examine the association between historical institutions
and modern political institutions, human capital, technological advancement and
economic prosperity for colonial and non-colonial samples. We use 6 different
measures for modern outcomes which includes: technology in 2000, two measures
for log of GDP per capita, one measure for overall institutions, one measure for
the rule of law and one for human capital proxied by average schooling.
  Table 1 shows for our sample of ex-colonies, those who had higher pre-colonial
prosperity faced a reversal of fortune across all outcome variables. Panel A re-
ports the relationship between log of population density in the 1500s, for both our
colonial and non-colonial samples and our 6 measures for present day outcomes.
For our colonial sample, we see a negative and statistically significant relationship,
at the 1 percent level, between log of population density and all the outcomes of
interest. For the non-colonial sample, we find the opposite relationship with 5 of
the 6 variables tested displaying a positive and statistically significant correlation
with population density in the 1500s at the 1 percent level. The factor average
schooling does not display a significant connection, however, it is positive in sign.
  Panel B displays the results of the relationship between historical institutions,
proxied by technological progress in the 1500s and our variables for modern pros-
perity. The results indicate that, for our colonial sample, technology in the 1500s is
a weaker predictor for modern prosperity compared to log of population density as
only 4 of the 6 coefficients display a statistically significant relationship. However,
all the point estimates are negative in sign. When examining our non-colonial
sample, all variables see reversal in sign with all coefficients being significant at
least at the 10 percent level.
  We conclude from this test that a reversal of fortune took place in our sample
of former colonies but not in our non-colonies. Therefore, we can use prosperity
in 1500s as a proxy for the degree of extractiveness of institutions.

4.2  Racism and Historical Institutions: Baseline results
Table 2 presents our baseline results. We examine if historical institutions predict
present day levels of racism for both colonial and non-colonial samples. As an
additional control, we test the relationship between current levels of racism and
modern measures for prosperity and technological advancement. According to the
literature Berggren and Nilsson (2013), we also expect to find a negative rela-
tionship between greater modern prosperity and racism across both samples. The

                     14
              Table 1: Testing The Reversal of Fortune

                 1       2      3       4      5       6
Panel A
                      Dependent Variable: Present Outcomes Below
                   Colonial Sample          Non-Colonial Sample
              Technology  Log GDP Log GDP Technology Log GDP Log GDP
              Index 2000  pc 2000   pc     Index 2000 pc 2000   pc

Log pop density 1500 CE   -0.140***   -0.642*** -0.976***    0.161***   0.549***   1.029***
               (-0.025)   (-0.121)  (-0.149)   (-0.034)   (-0.136)   (-0.27)
Observations           36      36     36      24      24      24
R-squared           0.537     0.399    0.513     0.376    0.271    0.365
                    Colonial Sample            Non-Colonial Sample
              WGI 1996    Rule Law  Avg Sch    WGI 1996   Rule Law  Avg Sch

Log pop density 1500 CE   -0.569***   -0.535***   -1.713***  0.708***   0.727***    0.969
               (-0.093)   (-0.115)   (-0.275)   (-0.2)   (-0.206)   (-0.568)
Observations           36      36      33      24      24      22
R-squared           0.549     0.439     0.51    0.275     0.279     0.131
Panel B
                      Dependent Variable: Present Outcomes Below
                   Colonial Sample          Non-Colonial Sample
              Technology  Log GDP Log GDP Technology Log GDP Log GDP
              Index 2000  pc 2000   pc     Index 2000 pc 2000   pc

Technology 1500 CE      -0.082*   -0.211   -0.497*     0.212***   0.696*    1.395*
               (-0.041)  (-0.21)   (-0.288)    (-0.072)   (-0.387)   (-0.728)
Observations           36     36     36       24      24      24
R2               0.116    0.027   0.084      0.178    0.118     0.181
                   Colonial Sample             Non-Colonial Sample
              WGI 1996   Rule Law  Avg Sch     WGI 1996   Rule Law  Avg Sch

Technology 1500 CE      -0.304*    -0.201   -1.336***  1.256***   1.135***   1.817**
               (-0.169)   (-0.189)   (-0.476)  (-0.296)   (-0.393)   (-0.779)
Observations           36      36      33     24      24      22
R2               0.099     0.039    0.201    0.234     0.184    0.091
Notes : All historical variables have been standardized . All regressions contain a constant. OLS coefficients
are reported in each column. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1




                          15
confirmation of this test reassures us that our identification strategy is correct.
If the relationship between current racism and current prosperity would have the
same sign as the relationship between current racism and pre-colonial prosperity,
we would have an important identification problem.
  The coefficients in table 2 confirm our hypothesis as all our measures for the
reversal of fortune within the colonial sample predict higher levels of racial intoler-
ance, while they display no effect or the opposite sign for the non-colonial sample.
Column 1 indicates a 1 standard deviation increase in the log of population den-
sity in the 1500s is correlated with an increase in the average level of racism by
7.3 percent. Given the mean value of racism for the colonial sample is around 18
percent, this represents a change of around 39 percent. Additionally, the log of
population density in the 1500s accounts for 40 percent of the variation of racism
and displays a statistically significant coefficient at the 1 percent level. Technology
in the 1500s, shown in column 3, displays a similar and slightly stronger relation-
ship with racism in terms of sign, magnitude and R2 as population density. When
examining the non-colonial sample, we see, as hypothesized, the opposite relation-
ship. All coefficients are negative in sign with technology in the 1500s displaying
a statistically significant correlation, albeit at the 10 percent level.
  In columns 2 and 4, we conduct the same analysis as in columns 1 and 3 for
log of GDP per capita and the technology index in the year 200010 . We observe
that when our proxies for prosperity are measured in 2000, there is a negative
and significant relationship with racism across both samples. This means, for our
colonial sample, more prosperous countries in the 1500s are now more prone to be
racist, while the opposite is true for the association between modern prosperity
and racism. Such a result supports that using the reversal of fortune is a valid
identification strategy to capture the affect of historically extractive intuitions on
racism in former European colonies. There is no other reason for explaining the
change in sign for the relationship of prosperity and racism across time. Our
interpretation is further supported by the outcomes in the non-colonial sample as
both modern and historical measures for prosperity have a consistent and negative
connection with racism.

4.3   Controlling for Fractionalization, Migration and Genetic Diversity
To make sure we are capturing the right causal channel and that the relationship
is not due to omitted factors which the literature has shown to be correlated with
racial preferences and institutions, we run a number of robustness checks in the
 10 We compare log of population density in 1500s with log of GDP in the present because we subscribe to the

Malthusian theory. The Malthusian theory stipulates that in the pre-industrial periods, unlike the post-industrial
period, an advancement of technology or increases in land availability did not result in long-term increases in
income per capita but was reflected in the rise in population density. Thus, more technologically sophisticated
societies had a denser population but not necessary a higher standard of living or greater income per capita. As a
result, comparing log of population density in the 1500s and log of GDP per capita in 2000 is appropriate because
they are the measures that best capture societal prosperity given their time periods (Ashraf and Galor 2011)



                            16
            Table 2: Historical Institutions and Racism

                    1      2     3        4
   Panel A
                        Dependent Variable: Racism
                   Colonial Sample    Non-Colonial Sample

   Log population   density,  0.073***        -0.034
   1500
                  (0.015)         (0.25)
   Log of GDP per cap 2000          -0.034***         -0.031**
                        (0.021)          (0.013)
   Observations           36      36     24       24
   R2               0.402    0.157   0.083      0.230




17
   Panel B
                        Dependent Variable: Racism
                   Colonial Sample    Non-Colonial Sample

   Technology Index, 1500 CE   0.093***        -0.105*
                   (0.020)        (0.59)
   Technology Index, 2000 CE         -0.254***         -0.257***
                        (0.056)          (0.079)
   Observations           36      36     24       24
   R2               0.406    0.176   0.210      0.321
   Notes : All historical variables have been standardized . All regressions contain a
   constant. OLS coefficients are reported in each column. Robust standard errors
   are in parentheses. *** p<0.01, ** p<0.05, * p<0.1
following three sections. Additionally, in all the following country level regressions
in our control exercises, we account for absolute latitude as a proxy for geographical
characteristics.
  Alesina et al. (2011) argue colonialism had a pivotal role in the establishment
of artificial national boundaries leading to many ethnic groups being separated
into bordering countries without their consent. Ethnic persity has also been
linked to lower levels of trust across ethnic groups and worse institutional outcomes
(Alesina et al. 2003; Hodler 2006; Putnam 2007; Stolle, Soroka, and Johnston
2008; Koopmans and Veit 2014b, 2014a). Thus, former colonies with extrative
institutions could have higher levels of fractionalization due to the establishment
of artificial states which, in turn, is the actual driving factor behind the relationship
between higher levels of racism and institutions. To account for this possibility, we
control for three forms of fractionalization that are ethnic, linguistic and religious.
All measures are extracted from Alesina et al. (2003). The point estimates in
table 4, columns 1 to 3, indicate fractionalization has a negligible effect on the
relationship between racism and historical institutions.
  Another omitted variable for a similar reason as fractionalization is modern
migration. To account for the potential contact effect of modern migration on
racism, we create our own variable which is the net migration over total popula-
tion averaged across 1984-2012. Both measures are taken from the World Bank
database with the created variable being labeled migration ratio. When we control
for modern migration, we see little effect on our coefficients of interest compared
to our baseline.
  Genetic persity could also be an omitted factor related to our variable of in-
terest. Ashraf and Galor (2013) argue high levels of genetic heterogeneity have
several consequences including; disarray, mistrust, a reduction in cooperation and
a disruption of the socioeconomic order of society. In conjunction with this ar-
gument, Spolaore and Wacziarg (2009) and Spolaore and Wacziarg (2013) claim
genetically rooted differences can create mistrust, a lack of communication and
racial or ethnic biases across group distinctions. Galor and Klemp (2018) also find
an important determinant of pre-colonial autocratic institutions which is genetic
persity. As a result, it is plausible that genetic persity is an important omitted
variable connecting racism and historical institutions due to the fact societies with
higher levels of genetic persity could have both worse historical institutions and
higher levels of racism. When controlling for ancestry adjusted and unadjusted
genetic persity, both taken from Ashraf and Galor (2013), we see a slight drop in
coefficient size but consistent outcomes in terms of sign and statistical significance.
  From table 3, we see that controlling for different features of a populations’
persity and modern migration cannot account for the relationship between his-
torical institutions and racism.




                     18
               Table 3: Controlling for Fractionalization, Migration and Genetic Diversity

                    1        2        3        4         5        6
   Panel A
                                  Dependent Variable: Racism
   Control Variables       Ethnic Frac  Lingustic Frac  Religious Frac Migration  Genetic Diversity   AA Gentic
                                         Ratio                Diversity

   log population density, 1500  0.072***    0.072***     0.068***    0.079***     0.055***     0.077***
                   (0.017)    (0.017)     (0.019)     (0.019)     (0.019)     (0.021)
   Observations            35       35        35       36        36       30
   R2                0.421     0.415      0.417      0.422      0.446      0.424




19
   Panel B
                                  Dependent Variable: Racism
   Control Variables       Ethnic Frac  Lingustic Frac  Religious Frac Migration  Genetic Diversity   AA Gentic
                                         Ratio                Diversity
   Technology Index, 1500 CE    0.094***    0.090***     0.083***   0.089***    0.071***       0.097***
                   (0.024)    (0.023)     (0.024)    (0.020)    (0.022)       (0.024)
   Observations            35       35        35      36       36         30
   R2                0.436     0.425      0.440     0.502     0.457        0.462
   Notes : All historical variables have been standardized . All regressions contain a constant. OLS coefficients are reported in
   each column. All specifications control for absolute latitude. AA Genetic Diversity stands for Ancesrty Adjusted Genetic
   persity. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1
4.4   Controlling for Ancestry Adjusted Variables, Share of Europeans
    and Indigenous Population
Another potential confounding factor that could shape the degree of racism and
institutions is the cultural characteristics migrants have brought with them from
their origin country. Putterman and Weil (2010) created a variable incorporating
the historical prosperity of descendants of populations who migrated to new coun-
tries after the 1500s. They argue, ancestry adjusted prosperity captures cultural
characteristics migrants bring when immigrating and is a factor which shapes eco-
nomic outcomes independently from historical institutions11 . It is easy to extend
this line of reasoning to racism as inpiduals who migrated to the new nations
would have brought norms, beliefs and values regarding other races and institu-
tions from their origin country.
  To isolate the effect of historical institutions and account for historical pop-
ulation movements, we control for ancestry-adjusted variables for the respective
measures for the reversal of fortune. This we means we control for ancestry ad-
justed log of population density and technology when regressing log of population
density and technology index in the 1500s on racism respectively. The weights
used to calculate our ancestry-adjusted variables are migration-weighted factors
over the time frame 1500 to 2000 CE and are extracted from Putterman and Weil
(2010) via Chanda, Cook, and Putterman (2014).
  In addition to the specific ancestry-adjusted variables (log of population den-
sity and technology in the 1500s), we control for two other ancestry adjusted
components. These measures include millennia of agriculture and the length of
state history12 . From table 4, columns 1, 2 and 3. we can see, ancestry adjusted
features have little to no impact on the size or significance of our coefficients of
interest.
  Additionally, Easterly and Levine (2016) show that a large population share
of Europeans during colonialization is related to greater economic development in
the present. Similar to the arguments for ancestry-adjusted variables, Europeans
may have brought cultural characteristics, technology and human capital with
them during migration to the new world. As a result, the cultural traits of the
Europeans themselves could be one of the factors that are shaping both racism and
the colonial institutions they created. To account for this possibility, we control
 11 Putterman and Weil (2010) and Chanda, Cook, and Putterman (2014) show ancestry adjusted historical

prosperity has a positive relationship with economic development. The positive relationship is present even in
their former European colonial sample which generally has a negative relationship between unadjusted historical
prosperity and current economic development.
 12 Ancestry adjusted millennial of agriculture measures the years since the population started to utilize agricul-

ture to a greater degree than foraging as the primary source of food after being adjusted for historical population
movements. Ancestry adjusted state history, is the proportion of time in which present-day countries had, first,
a supra-tribal polity, second, how large the area in which the polity covered, and, finally, if it was internally
developed or imposed by an external source adjusted for population migration. Chanda, Cook, and Putterman
(2014), Putterman and Weil (2010) and Spolaore and Wacziarg (2013) find the ancestry adjusted variables of
millennia of agriculture and state history to be positive and significant predictors of present income levels and
thus, we find it important to control for these factors for additional robustness.



                            20
for the share of Europeans in the 1900s, a variable taken from AJR (2002). As
an additional control for historical characteristics of the population, we account
for the share of indigenous population descent from the 1500s, a measure which
is extracted from Chanda, Cook, and Putterman (2014) via Putterman and Weil
(2010).
  When controlling for the percentage of Europeans in the population in the
1900s, displayed in column 4, we see a drop in the size and level of significance
of our independent variables of interest. Log of population density sees a drop in
its point estimate from 7.3 to 4.6 with a reduction in the level of its statistical
significance from 1 to 5 percent. Technology in the 1500s sees a similar change.
When controlling for the percentage of indigenous population in the 1500s, shown
in column 5, the coefficient for log of population density and technology sees a
reduction in its magnitude and statistical significance, from the 1 to 5 percent
level but otherwise displays consistent results.

4.5  Controlling for Religion, Colonial Origin and Other Cultural Fac-
   tors
Another potential confounding factor which could affect both racism and historical
institutions is religion. Religion has been shown to affect levels of social capital and
tolerance and the development of political and educational institutions (Fukuyama
2001; Guiso, Sapienza Zingales 2003; Becker and Woessmann 2009; Woodberry
2012; Acemoglu, Gallego, and Robinson 2014). To account for religion, we take
variables from La Porta (1999) that measures the proportion of different religions
in societies in the 1980s. When including religion in our specification, in table 5
column 1, we see little influence on our variables of interest in terms of magnitude,
sign and statistical significance.
  We also control for colonial identity in column 2, legal origin in column 3 and
regional fixed effects in column 4. Variables on colonial identity and legal origin are
extracted from La Porta (1999). The inclusion of colonial identity or legal origin
into our specification results in little to no change to our coefficients compared to
the baseline. However, the inclusion of regional fixed effects has a large impact on
the size and significance of our coefficients, indicating the importance of regional
characteristics in shaping levels of racism. Even so, historical institutions remain a
significant factor in predicting racism. This result is not surprising given we argue
that racism is persistent and hence, a large part of its variation from the mean is
time-invariant and disappears with regional fixed effects.
  Another important possibility is that racism maybe acting as a proxy for a
broader set of cultural features. As shown by Tabellini (2010), trust, control,
respect and obedience are shaped by historical institutions and education. To
account for this possibility, we control for all four of these variables in four different



                      21
   Table 4: Controlling for Ancestry Adjusted Variables, Share of Europeans, and Indigenous Pop

                        1       2       3       4       5
    Panel A
                               Dependent Variable: Racism
    Control Variables         Ancestry   Millenia of State His- Share Euro     Indiginous
                     Adjusted   Agriculture tory AA    1900      Pop
                     Variables   AA

    log population density, 1500    0.068***   0.072***   0.070***    0.046**    0.066**
                      (0.016)    (0.018)    (0.017)    (0.021)    (0.031)
    Observations              36      36      35      36      36
    R2                  0.414     0.420     0.415     0.445     0.412




22
    Panel B
                               Dependent Variable: Racism
    Control Variables         Ancestry   Millenia  State His- Share Euro     Indigenous
                     Adjusted   Agriculture tory AA    1900      Pop
                     Variables   AA

    Technology Index, 1500 CE      0.104***   0.097***   0.107***    0.061**   0.067***
                      (0.021)    (0.022)    (0.023)    (0.026)    (0.021)
    Observations              32      36      35      36      36
    R2                  0.390     0.459     0.485     0.455     0.496
    Notes : Our proxies for historical institutions have been standardized . All regressions contain a
    constant. OLS coefficients are reported in each column. All specifications control for absolute latitude.
    AA stands for ancestry adjusted. Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, *
    p<0.1
             Table 5: Controlling for Religion, Colonial Origin, Legal Origin and Other Cultural Factors

                     1       2      3      4      5      6       7    8
   Panel A
                                 Dependent Variable: Racism
   Control Variables       Religion   Colonial   Legal   Regional    Trust   Respect  Obedience  Control
                         Origin    Origin  Effects

   log population density, 1500  0.062***   0.062***   0.073***   0.049**   0.074***  0.071***  0.076***  0.076***
                   (0.013)    (0.018)   (0.018)   (0.021)   (0.016)   (0.020)   (0.017)   (0.023)
   Observations            36       36     36      36     36     36      36     36




23
   R2                0.495     0.526    0.459    0.523    0.439    0.412    0.428    0.416
   Panel B
                                 Dependent Variable: Racism
   Control Variables       Religion   Colonial   Legal   Regional    Trust   Respect  Obedience  Control
                         Origin    Origin  Effects
   Technology Index, 1500 CE    0.083***   0.083***   0.087***  0.049*   0.092***   0.106***  0.095***  0.093***
                   (0.022)    (0.027)   (0.024)  (0.025)   (0.023)   (0.028)   (0.021)   (0.025)
   Observations            36       36     36     36     36      36      36     36
   R2                0.449     0.488    0.433   0.506    0.436    0.469    0.468    0.430
   Notes : Proxies for historical institutions have been standardized . All regressions contain a constant. OLS coefficients are
   reported in each column. All specifications control for absolute latitude. Robust standard errors are in parentheses. *** p<0.01,
   ** p<0.05, * p<0.1
specifications, columns 5 to 8 13 . When controlling for these 4 features we see
little change to the point estimates as they remain consistent in size, sign and
significance.

4.6   Controlling for Different Samples, Outliers and IVs
In the literature, some specific countries are known to be outliers related to the
reversal of fortune. Therefore, it is usual in this strand of literature14 to check that
the results are not driven certain sub-samples. To do so, we first run our baseline
regression excluding the US, Canada, Australia, and New Zealand in column 1
and additionally excluding Singapore and Hong Kong in column 2. We also detect
that Libya and Bangladesh present unusually high rates of racism, therefore we
exclude them in column 3. The results presented in table 6 remain robust to the
exclusion of these sub-samples.
  Another method to address the impact of potential outliers is to run a MM
estimation, which we conduct in column 4 (Yohai 1987 and Aelst, Hubert and
Rousseeuw 2008). The outcome of this estimation displays consistent coefficients
in terms of size, sign and significance. Overall, from this table, we see our analysis is
robust to removing potentially important sub-samples and controlling for outliers.
  As a final step to account for potential omitted factors and issues of measure-
ment error with our historical data, similar to the strategy as AJR (2002), we
utilize an instrumental variable approach using two different sets of instruments.
In the first set of instruments, we use non-adjusted measures for millennia of agri-
culture and state history. Millennia of agriculture is a measure capturing the
timing of the Neolithic revolution. Ashraf and Galor (2011), Putterman (2008)
and Spolaore and Wacziarg (2013) empirically show that the earlier the Neolithic
transition of a country, the higher their population density in the 1500s. Con-
cerning state history, Chanda, Cook, and Putterman (2014) show that an early
emergence of the state allowed for innovative technologies and larger historical
populations. Hence, both millenia of agriculture and state history can be used as
an instrument for prosperity in the 1500s that might be prone to measurement
error. For our second set of instruments, we use log of population density in 1000
CE and technology in the year 0 CE taken from AJR (2002) and Comin, Easterly,
and Gong (2010) respectively.
  Columns 5 and 6 in table 6 demonstrate that the outcomes of utilizing the
instruments for log of population density and technology in 1500s produce similar
results compared to the OLS analysis displayed in table 2. In all cases there is
little change in the size of the coefficients, with all outcomes being statistically
 13 All measures are extracted from the world value survey and are created in the same manner as our measure

for racism. Trust captures generalized trust, control is the degree to which inpiduals feel they have control over
their life, and respect and obedience measure respondents’ answers to the questions, how important it is to teach
children tolerance and respect and obedience.
 14 See AJR (2002) or Chanda, Cook, and Putterman (2014) among others




                            24
significant at the 1 percent level. The p−value for the Hanson j−statistic and the
F −statistic for weak identification are also reported. The results of these tests
support the validity of our instruments. The first stage of these regressions are
reported in the appendix. Overall, we can see the use of an instrumental variable
approach further supports the notion that historical institutions have a causal
impact on racism and are not due to other omitted factors.


5   Racism as a Persistent Internal Norm or a Result of
   Present Institutions?
To this point we have provided evidence decedents in former European colonial
societies that had extractive institutions are more likely to be racist and this
relationship is causal. However, the question still remains, through what channels
do extractive institutions alter the evolution of the cultural norm of racism? Is it
indirectly through reducing the quality of modern institutions or directly through
a permanent shift in internal norms? In this section, we utilize three different
empirical strategies disentangle these two possible channels.

5.1  Cross-Country Analysis Controlling of Present Day Prosperity
The first step in trying to determine the channel of causality is to re-examine our
baseline specification from table 2 while simultaneously controlling for prosper-
ity and different measures for current institutional quality at the cross country
level to see if these present day factors mediate the relationship between histor-
ical institutions and racism. If they do, we have evidence historically extractive
that institutions shape racism via modern institutions and or prosperity. Table 7,
columns 1 and 2 show the correlations between racism and historical institutions
while controlling for GDP per capita and technology in the year 2000 respectively.
In both columns, the coefficients for historical institutions are consistent in sign,
size and significance compared the baseline model in table 2. In column 3, we
account for the general institutional setting, measured by the average of the world
governance indicators in 1996, and in column 4, we control specifically for the rule
of law. In column 3, we see a general drop in the magnitude of our coefficients
of interest, however, both maintain their sign and level of significance. Column 4
shows that controlling for the rule of law results in almost no change to the point
estimates for historical institutions.
  In column 5 of table 7, we control for average schooling which also results in
little variation in the results. Column 6 controls for economic institutions, proxied
by economic freedom following Berggren and Nilsson (2013)15 . The results of table
7 provide evidence that historically extrative institutions have a direct impact on
 15 We choose economic freedom as our proxy for economic institutions because Berggren and Nilsson (2013)

empirically show, different aspects of economic freedom can foster tolerance.


                          25
               Table 6: Controlling for different outliers, samples and IVs

                      1       2       3      4      5      6
   Panel A
                      Dependent Variable: Racism
   Control Specifications      Excluding Excluding  Excluding     MM      IV 1     IV 2
                   Neo-    Sing, HK  Bangladesh
                   Europe   and Neo-  and Libyia
                         Europe

   log population density, 1500   0.075***   0.090***   0.053***   0.062***  0.085***   0.070***
                    (0.019)    (0.019)    (0.014)    (0.018)  (0.029)   (0.020)
   Observations             32      30      34      36     35     32
   R2                 0.325     0.378     0.406          0.399    0.432
   p-value of Hansen J statistic                              0.955    0.326
   F stat for weak identification                              16.909    79.459




26
   Panel B
                      Dependent Variable: Racism
   Control Specifications      Excluding Excluding  Excluding     MM      IV 1     IV 2
                   Neo-    Sing, HK Bangladesh
                   Europe   and Neo- and Libyia
                         Europe

   Technology Index, 1500 CE     0.085***   0.100***   0.069***   0.083***  0.092***   0.122***
                    (0.022)    (0.022)    (0.018)    (0.023)  (0.027)   (0.033)
   Observations             32      30      34      36     35     32
   R2                 0.333     0.422     0.452          0.428    0.401
   p-value of Hansen J statistic                              0.686    0.869
   F stat for weak identification                              52.948    27.958
   Notes : Proxies for historical institutions have been standardized . All regressions contain a constant. OLS
   coefficients are reported in each column. We control for absolute latitude in all regressions. IV 1 entails
   unadjusted state history and millennia of agriculture. IV 2 entails log of population density in 1000 and
   technology index in 0 CE . Robust standard errors are in parentheses. *** p<0.01, ** p<0.05, * p<0.1
                    Table 7: Controlling for Present Outcomes Variables

                     1        2         3      4      5        6
   Panel A
                               Dependent Variable: Racism
   Control Variables       Technology  Log GDP pc 2000 WGI 1996 Rule Law Avg Sch         Economic Freedom
                  Index 2000

   log population density, 1500  0.088***     0.083***     0.072***   0.076***  0.082***     0.064***
                   (0.026)      (0.024)     (0.019)   (0.017)   (0.020)     (0.018)
   Observations            36        36        36      36     33       36
   R2                0.431       0.429      0.412    0.415    0.452      0.417




27
   Panel B
                               Dependent Variable: Racism
   Control Variables       Technology  Log GDP pc 2000 WGI 1996 Rule Law Avg Sch         Economic Freedom
                  2000

   Technology Index, 1500 CE    0.081***     0.086***     0.078***   0.084***  0.082***     0.078***
                   (0.023)      (0.022)     (0.019)   (0.020)   (0.023)     (0.019)
   Observations            36        36        36      36     33       36
   R2                0.454       0.465      0.500    0.490    0.469      0.515
   Notes : Proxies for historical institutions have been standardized . All regressions contain a constant and contol for
   absolute latitude. OLS coefficients are reported in each column. Robust standard errors are in parentheses. ***
   p<0.01, ** p<0.05, * p<0.1
shaping modern levels of racism independently of present day institutional quality
and prosperity .

5.2  Inpidual Level Analysis WVS
Our second step in identifying if racism is an internal norm is to account for an
inpidual’s view on the quality of institutions in their home country. Inpiduals
may have more animosity towards other races not because they posses an internal
norm of racism but because they lack confidence in their government to operate
effectively. A lack in government quality may lead societies to rely more heavily
on racial groups as a means of social, political and economic protection leading to
greater racial biases. If this hypothesis holds, we expect inpiduals who consider
their institutions to be of low quality will be more likely to be racists compared
to those inpiduals with a better opinion about the quality of government. If the
connection between racism and historical institutions operates through this the
channel, controlling for a person’s confidence in the government should make the
coefficient for extractive historical institutions lose its significance.
  To test this, we re-estimate our baseline regression at the inpidual level while
controlling for a person’s confidence in the government along with other inpidual
and country level characteristics for the same sample of countries.
  For our measure of racism, we use the same question as the country level variable
before it was aggregated. As a result, racism now takes the value 0 or 1, with 1
indicating an inpidual does not want to have someone of a different race as a
neighbor. For our proxies for historical institutions, we use the same country level
variables, log of population density and technology in the 1500s.
  Given we combine data from the inpidual and country levels, our data is
hierarchical and clustered and thus, if we use simple OLS it will increase the
probability of a type 1 error due to the underestimation of standard errors because
they do not possess a normal distribution (Klein et al. 2000). To account for the
nature of our data, we use hierarchical linear modeling methods. To estimate the
effects of historical institutions, country level data, on racism, inpidual level data,
we use a linear multilevel random effects model. In multilevel methods, random
effects refer to group-specific factors, in our case historical institutions that are
assumed to influence the dependent variables. In using random effects, we assume,
unobserved country-specific effects are randomly distributed with a mean of zero,
have constant variance and are uncorrelated to the predictor variables. These
assumptions allow the constant term to vary randomly across countries (Autio,
Pathak, and Wennberg 2013). Using multilevel analysis has two advantages: first,
it allows us to test if the connection between racism and historical institutions
extends to the inpidual level, with the additional benefit of the examination being
in a far larger sample size, ranging from 101,356 to 50,694 inpidual observations.
Second, it allows us to more precisely control for an inpidual’s perception on


                     28
the functioning of their institutions and other inpidual characteristics such as
education and income which are not captured in country level analysis.
  Table 8 column 1 reports the relationship between racism and our 2 different
measures for historical institutions with only inpidual level controls. Consistent
with our baseline results in table 2, there is a robust, positive and statistically
significant relationship between racism and historically extractive colonial institu-
tions. These results indicate worse historical institutions are predictors of a higher
probability an inpidual will be racist and that this relationship does not oper-
ate through confidence in the government. However, given we are not utilizing
a model with country fixed effects, we need to account for omitted country level
factors which could be biasing the results. To account for the level of prosperity
and the institutional environment an inpidual is embedded, in columns 2 to 4,
we also control separately for country level measures for Log of GDP per capita in
2000, institutions and average schooling. In column 5, we account for all three of
these country level factors in the same specification. Controlling for these features
inpidually or together has little affect on the coefficient of interest. As a placebo
test, in column 6, we run the same specification as column 5 for our non-colonial
sample resulting in a similar outcome as table 2, in that, there is a reversal in the
sign of the point estimates and, in some cases, a statistically significant negative
relationship between historical institutions and racism. We also check the robust-
ness of our results to alternative estimation methods. Using OLS with robust
standard errors clustered by country and a multilevel-logit model produce nearly
identical outcomes as table 8. To further control for other potential omitted coun-
try level factors, we reproduce the regressions from tables 3 to 5 at the inpidual
level while controlling for the same country level characteristics. The outcomes of
these regressions produce consistent results compared to their country level coun-
terparts. Both the alternative estimation methods and additional control tables
are shown in the appendix.

5.3  Examining Inpiduals Who Have Immigrated to Europe
Our last test to determine whether racism is an internal norm is to examine in-
piduals who are from former European colonies and have immigrated to a new
environment, specifically Europe. The logic behind this analysis is if racism is in-
deed an internal norm and not a result of one’s present context, people will bring
such an attitude with them to a new environment. To test this, using data from
the European Social Survey, we examine, at the inpidual level, if our proxies for
historical institutions from a person’s origin country predict their level of racism
even after they have immigrated to Europe. To capture an inpiduals level of
racism, we use four different measures which are generally taken from a set of
rotating questions asked in the first (2002) and seventh (2014) waves of the Euro-
pean Social Survey. We do not examine questions with regard to race that do not


                     29
                               Table 8: Multi-level Analysis

                      1         2         3          4         5           6
   Panel A
                                        Dependent Variable: Racism
                  Colonial Sample  Colonial Sample   Colonial Sample Colonial Sample    Colonial Sample  Non-Colonial Sample

   log population density, 1500   0.034***      0.039***      0.031***      0.037***      0.034***        0.006
                    (0.008)      (0.014)       (0.011)      (0.012)      (0.012)        (0.011)
   Observations            101,346      101,346       101,346      93,869       93,869        50,694
   Number of groups/clusters      36         36         36         33         33          20
   Panel B
                                        Dependent Variable: Racism




30
                  Colonial Sample  Colonial Sample   Colonial Sample Colonial Sample    Colonial Sample  Non-Colonial Sample

   Technology Index, 1500 CE     0.275***      0.263***      0.238***      0.255***      0.300***       -0.488**
                    (0.063)      (0.066)       (0.060)      (0.069)      (0.053)        (0.227)
   Observations            101,346      101,346       101,346      93,869       93,869        50,694
   Number of groups/clusters      36         36         36         33         33          20
   Specifications
   Inpiudal Controls         Yes        Yes        Yes        Yes         Yes         Yes
   Control Log GDP Per Cap       No         Yes        No         No         Yes         Yes
   Control Insitituions         No         No         Yes        No         Yes         Yes
   Control Avg Schooling        No         No         No         Yes         Yes         Yes
   Wave Fixed Effects          Yes        Yes        Yes        Yes         Yes         Yes
   Notes : All regressions contain a constant. The unit of observation is the inpidual. All coefficients are reported in each column separated by
   panels. Inpidual level controls include: confidence in the government, income, level of education, life satisfaction, subjective health, a dummy
   for sex and age. Our dependent variables of interest are measures at the country level. Other controls are at the country level, which include:
   institutions, measured by the average of the world governance indicators index in 1996, log of GDP per capita in 2000 and average schooling
   from 1985-1995. Additionally, to account for time, we use wave fixed effects. Robust standard errors clustered by country are in parentheses
   *** p<0.01, ** p<0.05, * p<0.1
appear on both these waves. The first measure is the response to the question,
”To what extent do you think your country should allow people of a different race
or ethnic group from most people?”. The answers are on a scale from 1 to 4 with 1
corresponding to allowing many into the country and 4 indicating allowing none,
we refer to this variable as, race immigration. The second variable is in response
to the question, ”Thinking of people who have come to live in your country from
another country who are of a different race or ethnic group from most people. How
much would you mind or not mind if someone like this was appointed as your boss”.
The response is on a scale from 0 to 10 with 0 indicating do not mind at all and 10
indicating mind a lot. We refer to this feature as, race boss. The third measure is
derived from the question, ”Now thinking of people who have come to live in your
country from another country who are of a different race or ethnic group from most
people. How much would you mind or not mind if someone like this married a close
relative of yours”. The variable is also on a scale from 0 to 10 and has the same
responses as race boss. We call this variable, race marriage. The final variable is
in response to the question, ”How good or bad is it for a country to have a law
against racial or ethnic discrimination in the workplace?”. The answers are on a
scale from 0 to 10 with 0 being extremely bad and 10 extremely good. We refer
to this feature as laws against discrimination. Given these questions do not refer
to a specific race, only those who are not the majority race or ethnicity, similar to
the WVS survey question, it provides us the flexibility to examine views on race
from inpiduals in a variety of countries and cultures without being bogged down
by exactly what race they identify with or how they feel about certain races.
  We control for a number of inpidual level controls included in all specifications
which are: years of education, life satisfaction, feelings towards household income,
gender, age, health, dummies for religious denomination, trust in the legal system,
how long one has lived in the country and the number of people of minority race
and ethnicity in their current living area. Controlling for trust in the government
is important for identical reasons as the previous section. We account for views
on household income because economic insecurity has been shown to be a driving
factor of higher levels of racial intolerance and since immigrants from specific
countries could be more economically insecure and thus, more racist (Johnson
and Lordan 2016). Religion dummies are utilized to account for the role different
religious denominations may have in shaping racism (Guiso et al 2003). We control
number of people from minority races or ethnic groups in the current living area for
two reasons. First, it is possible people who have more racist views self-select into
more homogeneous neighbourhoods and second, the fact someone lives in a racially
heterogeneous or homogeneous area could increase or reduce their level of racism
through contact and socialization with other groups. Finally, we account for the
how long someone has lived in the country to ensure that the connection between
racism and the historical institutions of their origin country does not dissipate the
longer an inpidual has lived in the country.


                     31
                            Table 9: Immigrants to Europe ESS

                    1      2      3      4       5     6      7         8
   Panel A
                        Colonial Origin Sample               Non-Colonial Origin Sample
                  Race    Race    Laws    Race Immi-   Race    Race   Laws     Race Immi-
                  Boss    Marriage Discrimi- gration      Boss    Marriage Discrimi- gration
                             nation                     nation

   log population density, 1500  0.168***  0.159***   -0.143***  0.024**    -0.059   -0.012   0.099*     -0.036***
                   (0.042)  (0.037)   (0.026)   (0.011)    (0.043)   (0.055)   (0.051)     (0.011)
   Observations           1,241   1,252    1,248    1,238     2,940    2,943   3,055      3,064
   R2                0.180   0.328    0.089    0.198     0.148    0.205    0.087      0.175
   Panel B




32
                        Colonial Origin Sample               Non-Colonial Origin Sample
                  Race    Race    Laws    Race Immi-   Race    Race   Laws     Race Immi-
                  Boss    Marriage Discrimi- gration      Boss    Marriage Discrimi- gration
                             nation                     nation

   Technology Index, 1500 CE   1.329***  1.645***   -.774**   0.245**    -0.739  -1.757*    -0.635      -0.405
                   (0.405)  (0.323)   (0.246)   (0.107)    (0.705)  (1.009)   (0.574)     (0.274)
   Observations           1,106   1,116    1,113    1,106     1,940   1,941    2,027      2,034
   R2                0.182   0.336    0.080    0.197     0.161   0.205     0.205      0.189
   Specifications
   Inpidual controls       Yes     Yes     Yes     Yes     Yes     Yes     Yes        Yes
   Country Fixed Effects       Yes     Yes     Yes     Yes     Yes     Yes     Yes        Yes
   Notes: All regressions contain a constant. The unit of observations is the inpidual. All coefficients are reported in each column
   separated by panels. Controls include: years education, life satisfaction, feelings towards household income, dummy for gender, age,
   health, 8 dummies for religious denomination, trust in legal system, years living in country and people of minority race and ethnicity in
   current living area. We restrict our analysis to countries that in general have at least 15 observations for all the specifications.
   Additionally, we only examine people who do not identify as being part of a minority race or ethnic group. Robust standard errors
   clustered by country are in parentheses . *** p<0.01, ** p<0.05, * p<0.1
  In table 9, we examine the relationship between racism and an inpidual’s
ancestral historical institutions at the inpidual level using OLS with country
fixed effects. We use country fixed effects to account for the institutional and
cultural environment in which an inpidual has immigrated and to control for
potential selection bias. It is possible immigrants coming from different countries
will choose to migrate to alternative locations, thus using county fixed effects
should account for this possibility (Alesina and Giuliano 2010).
  The results of table 9 indicate, for those inpiduals from a post-colonial Eu-
ropean nation, worse historical institutions from their birth country have a sta-
tistically significant relationship with higher levels of racism across all measures
even when accounting for inpidual and country of destination characteristics.
All coefficients display a statistically significant correlation with both measures
for historical institutions at least at the 5 percent level. When testing the same
relationship for people from a non-colonial society, we find, consistent with our
previous results, the opposite and a generally weaker relationship in almost all
cases. These outcomes support the validity of our previous findings that historical
institutions shape racism through internal norms, beliefs and values and supports
the notion that our identification strategy is capturing the reversal of fortune for
post-European colonies.
  In table 10, we control for modern economic prosperity, institutional quality and
education levels of an inpidual’s country of origin across different specifications.
As we have shown in section 4.2, economic prosperity is correlated with lower lev-
els of racism, thus, it is feasible, if racist inpiduals are more likely to immigrate
from poorer societies and since worse historical institutions predict lower economic
prosperity, we could be capturing the relationship between economic prosperity of
the country of origin and racism, not historical institutions and racism (Alesina
and Giuliano 2010). When accounting for log of GDP per capita, in all specifica-
tions, the results remain relatively unchanged in terms of the coefficient size, sign
and statistical significance. For a similar reason as economic prosperity, current
institutions could be an omitted variable and thus we also control for institutional
quality of an inpidual’s origin country. When additionally accounting for origin
country institutions, columns 2,3,5,6,8,9,10 and 12, historical institutions continue
to predict that inpiduals will mind to a greater degree if people who are a dif-
ferent race or ethnicity as the majority is their boss, if they marry a relative and
less inpiduals from a different race or ethnic group should be let into the country
independent of adjusting for origin country institutions. A person’s views on laws
against discrimination and its connection to historical institutions is dependent on
the proxy used. When historically worse institutions are proxied by log of pop-
ulation density in the 1500s, they are associated with an increased likelihood an
inpidual will think such laws are bad for the country which is significant at the
1 percent level but when institutions are measured by technology in the 1500s,
while having the same sign as log of population density, it is no longer significant.


                     33
Another important omitted variable related to our outcomes of interest is unob-
served human capital. Though we adjust for an inpidual’s level of education,
it is possible that our results are driven by a lower level of human capital in the
country of origin. When checking the robustness of our results while adjusting
for human capital of an inpidual’s origin country, columns 3,6,9 and 12, the
outcomes remain consistent compared the previous respective columns.
  The outcomes of table 10 provide some evidence that the effect of historical
institutions may have a different channel depending on the specific question asked
as they could be capturing different aspects of racist or discriminatory beliefs. For
example, questions about races with regard to being one’s boss, marrying a relative
or the number of immigrants that should be allowed into the country are always
significantly correlated with historical institutions. However, this is not always
the case for views on laws against discrimination. While the purpose of this paper
is to establish a connection between historical institutions and racism, which is
supported by these results, such findings open the door for a deeper exploration of
racism as different aspects of it may be historically rooted while others are shaped
to a greater degree by modern phenomenon. This further exploration is outside
the scope of the paper.
  One potential factor that could be biasing our results is the presence of large
number of inpiduals from a single origin country. When exploring this possibility,
we find there are a large number of people from Morocco (around 160) and the
United States (around 51) in our baseline specifications from table 9. To account
for this potential issue, we re-run our regressions from table 9 while removing
inpiduals from Morocco and USA in table 10. When excluding the sub-sample
of Morocco and the United States, we see consistent results compared to table 9.
  As an additional robustness check, we examine if the historical institutions
of an persons father and mothers’ origin country predicts their level of racism.
Table 11 reports the outcomes using nearly the same specification as table 916
with columns 1 to 4 displaying the results for the fathers side, 5 to 8 the mothers
side and 9 to 12 when both the father and mother have the same country of
origin. Overall, we see consistent results compared to table 9, meaning extractive
historical institutions continue to predict higher levels of racism. There are two
exceptions to this outcome seen in column 4 panel A and column 12 panel B as
the coefficients are no longer statistically significant though they display the same
sign. The analysis from table 11 provides further support for the long term impact
of historically extractive institutions and the existence of the inter-generational
transmission of racist attitudes.
 16 We do not control for how long someone has lived in the country because second generation immigrants are

included in our sample of analysis and for them this question is not asked or applicable.




                           34
                   Table 10: Immigrants to Europe ESS Dropping Morocco and USA

                    1      2      3      4       5     6      7        8
   Panel A
                         Dropping Morocco                    Dropping USA
                  Race    Race   Laws     Race Immi-  Race    Race   Laws    Race Immi-
                  Boss    Marriage Discrimi-   gration    Boss    Marriage Discrimi- gration
                             nation                      nation

   log population density, 1500  0.155***  0.150***  -0.127***   0.022**   .184***   .217***  -0.131***     -0.027**
                   (0.039)  (0.038)   (0.032)   (0.010)   (0.039)   (0.042)   (0.035)      (0.012)
   Observations           1,028   1,040    1,048    1,031    1,163    1,173    1,169       1,158
   R2                0.168   0.299    0.100    0.180    0.187    0.327    0.082       0.207
   Panel B




35
                         Dropping Morocco                    Dropping USA
                  Race    Race   Laws     Race Immi-  Race    Race   Laws    Race Immi-
                  Boss    Marriage Discrimi-   gration    Boss    Marriage Discrimi- gration
                             nation                      nation

   Technology Index, 1500 CE   1.413***  1.874***   -0.662**   0.246**   1.326***  1.872***   -0.570*     0.271**
                   (0.404)  (0.299)   (0.295)   (0.103)    (0.394)   (.345)   (0.306)     (0.097)
   Observations           893    904     913     899     1,028    1,037    1,034      1,026
   R2                0.175   0.316    0.093    0.180     0.190    0.355    0.074      0.207
   Specifications
   Inpidual controls       Yes     Yes     Yes     Yes     Yes     Yes     Yes        Yes
   Country Fixed Effects       Yes     Yes     Yes     Yes     Yes     Yes     Yes        Yes
   Notes: All regressions contain a constant. The unit of observations is the inpidual. All coefficients are reported in each column
   separated by panels. Controls include: years education, life satisfaction, feelings towards household income, dummy for gender, age,
   health, 8 dummies for religious denomination, trust in legal system, years living in country and people of minority race and ethnicity in
   current living area. Robust standard errors clustered by country are in parentheses . *** p<0.01, ** p<0.05, * p<0.1
                   Table 11: Immigrants to Europe Using Father and Mother Origin Country ESS

                 1      2      3      4      5      6      7      8      9      10     11      12
   Panel A
                       Fathers Side                   Mothers Side                Fathers and Mother same country
               Race    Race    Laws     Race Immi-  Race    Race   Laws     Race Immi-  Race     Race    Laws     Race Immi-
               Boss    Marriage Discrimi-   gration    Boss    Marriage Discrimi-   gration    Boss     Marriage Discrimi-   gration
                          nation                      nation                        nation

   log population      0.145**  0.223***  -0.118***   0.025    0.114**  0.176***  -0.151***   0.048***   0.166***   0.229***  -0.139**    0.039*
   density, 1500
               (0.064)   (0.043)   (0.025)   (0.023)   (0.039)   (0.036)   (0.049)   (0.014)   (0.045)    (0.050)  (0.051)    (0.018)
   Observations       1,093    1,093    1,088    1,094    1,078    1,078    1,074    1,078     852      851    848      849
   R2            0.196    0.361    0.073    0.241    0.193    0.351    0.067    0.236    0.198     0.358   0.073     0.263
   Panel B




36
                       Fathers Side                   Mothers Side                Fathers and Mother same country
               Race    Race    Laws     Race Immi-  Race    Race   Laws     Race Immi-  Race     Race    Laws     Race Immi-
               Boss    Marriage Discrimi-   gration    Boss    Marriage Discrimi-   gration    Boss     Marriage Discrimi-   gration
                          nation                      nation                        nation

   Technology        1.777***  2.429***  -0.585***   0.255*   1.072***  1.628***   -1.003**   0.355***   1.554***   2.247***  -0.847**    0.256
   1500 CE
               (0.400)   (0.238)   (0.187)   (0.132)   (0.356)   (0.452)   (0.383)   (0.096)   (0.432)    (0.391)  (0.381)    (0.175)
   Observations       1,003    1,003    998     1,006     993     993     989     995     774      773    770      773
   R2            0.201    0.360    0.067    0.242    0.194    0.348    0.064    0.237    0.203     0.353   0.071     0.264
   Specifications
   Inpidual controls    Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes      Yes
   Country Fixed Effects    Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes     Yes      Yes
   Notes: All regressions contain a constant. The unit of observations is the inpidual. All coefficients are reported in each column separated by panels. Controls include: years
   education, life satisfaction, feelings towards household income, dummy for gender, age, health, 8 dummies for religious denomination, trust in legal system and ethnicity in
   current living area. Robust standard errors clustered by country are in parentheses. *** p<0.01, ** p<0.05, * p<0.1
6  Conclusion
In our study, we show, first, there is a reversal of fortune in the countries compos-
ing our sample of former European colonies and that this reversal does not occur
in non-colonies. Following the literature, we claim this is a consequence of the
establishment of extractive institutions which were exogenously determined. We
use this phenomenon as an identification strategy to show that former colonies
with more extractive institutions exhibit higher levels of racism today. We argue
this could be caused by both a deliberate instillation of racism in the population
of extractive colonies, and/or that extractive institutions created an environment
conducive for the endogenous formation of lower levels of racism vis-a-vis former
colonies that had more inclusive institutions. We then show that this relationship
is robust to controlling for several potential confounding factors. Finally, we ex-
amine the mechanism for how historically extractive institutions shape levels of
racism using three different strategies. The results of analysis indicate the effect
of historically extractive institutions on modern levels of racism mainly operates
through the persistence of racism as an internal belief, value and norm.
  Our results go beyond the identification of one of the determinants of racism
and contribute to the literature in several other ways. First, we show that the
impact of colonial institutions not only affects modern institutional quality but
also left an imprint on cultural values which persists until today. Second, the
paper contributes to the understanding of how racism persists across time by
identifying that it is an internal norm, belief and value. Linking this result with
the recent findings in the literature on the negative economic and political impact
of racist attitudes, we can claim that the persistence of racism might be one factor
contributing to the persistence of extractive institutions across time, an outcome
which has been shown to have consequences on economic development. Our results
also support the hypothesis that abrupt changes of institutions might dramatically
alter cultural values in ways that are not easily reversed. This research opens new
avenues of study investigating of the impact of extractive colonial institutions on
other cultural values beyond racism. Finally, there is room for further research
expanding on our current findings with a deeper exploration in to what aspects
of institutions lead to more racism and how this mechanism operates. This will
likely require micro level studies to provide the required contextual detail.




                    37
Data Appendix Country Level
Racism: The question in the survey is: ”On this list are various groups of people.
Could you please mention any that you would not like to have as neighbors?”
Answer is coded (1) if people of a different race is mentioned and 0 if not. For the
country level, the variable is averaged by country. Source: World Value Surveys.

  Ancestry Adjusted variables: Ancestry adjusted covers population density,
millennia of agriculture, state history, and technology. The measures are migration
weighted measures for the period 1500 to 2000 CE. Source: Putterman and Weil
(2010) via Chanda, Cook, and Putterman (2014).

  Colonial Origin: Colonizer dummies are for the identity of the European col-
onizer country which include British, French, German, Spanish, Italian, Belgian,
Dutch, and Portuguese. Source: La Porta (1999)

  Control: Derived from the question ”Some people feel they have completely
free choice over their lives, while other people feel that what they do has no real
effect on what happens to them”. Please use this scale where (1) means ”no choice
at all” and (10) means ”a great deal of choice” to indicate how much freedom of
choice and control you feel you have over the way your life turns out. The mea-
sure is averaged by country over all available waves. Source: World Value Surveys.

  Economic Freedom: Measure for the degree of economic freedom with in a
society. The measures encompasses 5 sub-components which are the size of gov-
ernment, legal structure and security of property rights, access to sound money,
Freedom to trade internationally and Regulation of credit, labor and business. The
variable is averaged over all years available. Source: Gwartney and Joshua (2011)

  Fractionalization: Measures for ethnic linguistic and religious fractionaliza-
tion are taken from Alesina et al. (2003)

  Genetic Diversity and Ancestry Adjusted Genetic Diversity: Measures
capture the migration adjusted and unadjusted levels of predicted genetic persity
on the modern country level. Source: Ashraf and Galor (2013).

 Latitude: Absolute value of latitude scaled between zero and one. Source: La
Porta (1999).

 Legal Origin: Dummy variables that indicate the legal tradition of a country
which includes British, French, German or Scandinavian. Source: La Porta (1999)


                    38
  Log of GDP per capita in 2000 CE: log of real GDP per capita, in constant
2000 international dollars, as reported by the Penn World Table, version 6.2 taken
via Ashraf and Galor (2013)

  Log of GDP per capita: log of GDP per capita in constant 2005 US dollars
averaged over the period 1984-2013. Source: World Bank Development Indicators

  Millennia of Agriculture: The quantity of millennia a country has utilized
agriculture until 2000 CE. Source: Putterman and Trainor (2006) via Chanda,
Cook, and Putterman (2014).

  Obedience: Derived from the question ”Here is a list of qualities that children
can be encouraged to learn at home. Which, if any, do you consider to be espe-
cially important?” If obedience is mentioned, it is coded as (1), if not mentioned
(0). The measure is averaged by country over all available waves. Source: World
Value Surveys.

  Population Density in 1500 and 1000: Total population in relation to
arable land in 1500 and 1000 CE. Source: McEvedy and Jones (1978) via Ace-
moglu, Johnson, and Robinson (2002)

 Regional Dummies: Latin America, Europe and Central Asia, South Asia,
SubSaharan Africa, East Asia and the Pacific and Western Europe. Source: World
Bank .

  Religion: Religion measures the percent of a country in which the population
identifies with a specific religion which includes Roman Catholic, Protestant, Mus-
lim, and Other in 1980. Source: La Porta (1999)

  Respect: Derived from the question ”here is a list of qualities that children can
be encouraged to learn at home. Which, if any, do you consider to be especially
important?” If tolerance and respect for other people is mentioned, it is coded as
(1,) if not mentioned (0). The measure is averaged by country over all available
waves. Source: World Value Surveys.

  Rule of Law: Our proxy for legal institutions is the rule of law. This measure
captures the level of confidence agents have in and abide by the rules of soci-
ety. Specially, the quality of contract enforcement, property rights, the police, the
courts and finally, the probability of crime and violence. The measure is averaged
over the years 1996-2013: Source Kaufmann, Kraay, and Mastruzzi (2013)

  Share of European 1900s: Percentage of settlers of European decent 1900s.


                     39
Source: Acemoglu, Johnson, and Robinson (2002)

 Share of indigenous population decent 1500: Share of people in the population
who are of indigenous decent from 1500 CE. Source: Putterman and Weil (2010).

  State History in 1500 CE: An index of state antiquity for the period 1 CE
to 1500 CE. Source : Putterman (2007) via Chanda, Cook, and Putterman (2014).

  Technology in 0 AD, 1500 and 2000 CE: These measures, though they
are constructed differently, capture the level of technology in a country around
the year 0, 1500 and 2000 respectively. Source: for technology in 1500 was taken
from Comin, Easterly, and Gong (2010) via Chanda, Cook, and Putterman (2014)
. Current technology and technology in 0 AD was taken directly from Comin,
Easterly, and Gong (2010)

  Trust: Derived from the question ”Generally speaking, would you say that
most people can be trusted or that you need to be very careful in dealing with
people?” Answer is coded (1) if people can be trusted and (0) if you cannot be
too careful. The measure is averaged by country over all available waves. Source:
World Value Surveys.

  World Governance Indicators 1996 average: The average of all world gov-
ernance indicators for 1996. It captures voice and accountability, political stability,
government effectiveness, regulatory quality, rule of law, and control of corruption.
Source: Kaufmann, Kraay, and Mastruzzi (2013).

  Years of Schooling, Average 1985 to 1995: The country-level average
years of schooling for the population above 15 years of age over the period 1985
to 1995. Source: Barro and Lee (2010).




                     40
Data Appendix WVS Inpidual Level
Racism: The question in the survey is : ”On this list are various groups of peo-
ple. Could you please mention any that you would not like to have as neighbors?”
Answer is coded (1) if People of a different race is mentioned and 0 if not.

  Education: The question in the survey is: What is the highest educational level
that you have attained? (1) Inadequately completed elementary education, (2)
Completed (compulsory) elementary education, (3) Incomplete secondary school:
technical/vocational type, (4) Complete secondary school: technical/vocational
type, (5) Incomplete secondary: university-preparatory, (6) Complete secondary:
university-preparatory, (7) Some university without degree/higher education, (8)
University with degree/higher education.

  Income: A scale of incomes in which the household falls into, before taxes
and other deductions. This variable takes values from 1 to 10, 1 being the lowest
decile and 10 the highest. The data is recollected in local currency, scaled and
then aggregated so the deciles represent a country level income ranking.

  Life Satisfaction:The question in the survey is : All things considered, how
satisfied are you with your life as a whole these days? Using this card on which
1 means you are ”completely dissatisfied’ and 10 means you are ”completely sat-
isfied” where would you put your satisfaction with your life as a whole? Source :
World Value Surveys.

 Age and Gender: Respondent’s age. Gender of the respondent. (0) Female
and (1) Male

  Health: Respondent’s Subjective health. This variable takes values from 1 to
5, 1 being very good health and 5 being very poor.

  Confidence Government: Respondent’s Confidence in the Government. This
variable takes values from 1 to 4, 1 a great deal of confidence and 4 being non at
all.




                    41
Data Appendix ESS Inpidual Level
Race Boss: Response to the question, thinking of people who have come to live
in your country from another country who are of a different race or ethnic group
from most people. How much would you mind or not mind if someone like this was
appointed as your boss. The response is on a scale from 0 to 10 with 0 indicating,
do not mind at all and 10 indicating mind a lot.

  Race Marriage: Response to the question, thinking of people who have come
to live in your country from another country who are of a different race or ethnic
group from most people. How much would you mind or not mind if someone like
this married a close relative of yours. The response is on a scale from 0 to 10 with
0 indicating, do not mind at all and 10 indicating mind a lot.

  Race Immigrant: Response to the question,to what extent do you think your
country should allow people of a different race or ethnic group from most people?
The answers are on a scale from 1 to 4 with 1 corresponding to allowing many into
the country and 4 indicating allowing none.

  Laws Discrimination: Response to the question, how good or bad is it for
a country to have a law against racial or ethnic discrimination in the workplace?
The answers are on a scale from 0 to 10 with 0 being extremely bad and 10 ex-
tremely good.

  Immigrants Same Race: Response to the question,to what extent do you
think your country should allow people of the same race or ethnic group from
most people? The answers are on a scale from 1 to 4 with 1 corresponding to
allowing many into the country and 4 indicating allowing none.

  Education: Years of full-time education completed.

  Feelings Income: Feeling about household’s income nowadays. This variable
takes values from 1 to 4, 1 being living comfortably on present income and 4 very
difficult living on present income.

  Trust Legal System: Respondent’s trust in the legal system. This variable
takes values from 0 to 10, 0 being non at all and 10 complete trust.

 Age and Gender: Respondent’s age. Gender of the respondent. (2) Female
and (1) Male

  Life Satisfaction: Respondents life satisfaction. This variable takes values

                    42
from 0 to 10, 0 being extremely dissatisfied and 10 extremely satisfied.

  Living Minorities: If people of a minority race/ethnic group are in the re-
spondents current living area. This variable takes values from 1 to 3, 1 almost no
one and 3 many.
  Religious Denomination : Fixed effects for religious denomination. Includes:
Roman Catholic, Protestant, Eastern Orthodox, Other Christian denomination,
Jewish, Islamic, Eastern religions and Other non-Christian religions.
  Lived in Country: Years lived in the county. Includes : with last year, 1 to
5 years, 6 to 10 years, 11 to 20 years and more than 20 years.




                    43
Appendix Figures and Tables


            .6
                                                                              LBY




                                                                                    BGD
            .4

                                                                                    IND
        Racism




                                                        ECU                                EGY
                                                                    VNM
                                                                IDN

                                           ZMB


                                                 MYS
                                                                NGA
                                                                      DZA
                     HKG
            .2




                                              MLI     PHL          GHA

                                                                       UGA

                                                       TZA              ETH  MAR TUN
                                     ZAF
                                                           MEX



                     SGP                                         BFA                   PAK
                                           CHL        PER


                     USA
                  AUS                                GTM
                        URY
                         BRA
                 CAN      ARG       NZL            COL
            0




               -4          -2                    0                        2                4
                              Log Population Density 1500s
                              Racism                          Fitted values



 Figure 1: Scatter Plot Racism and Log of Population Density 1500s Colonial Sample
            .6




                                                                   LBY




                                           BGD
            .4




                                                 IND
        Racism




                                                       ECU
                                                        EGY
                                              VNM
                                                     IDN

                                 ZMB


                                                                   MYS
                                    NGA
                                                           DZA
                                                                              HKG
            .2




                                    MLI   GHA            PHL

                                    UGA

                                TZA
                               ETH                    MAR        TUN
                                                               ZAF
                                                               MEX



                                 BFA            PAK                               SGP
                                                       PER           CHL


                                                                                 USA
                                                     GTM                      AUS
                                                             BRA     URY
                                                           COL       ARG      NZL CAN
            0




               4             6                     8                         10              12
                              Log of GDP per capita 2000
                              Racism                          Fitted values



   Figure 2: Scatter Plot Racism and Log of GDP Per Capita Colonial Sample




                                           44
         .6
                                                                                      LBY




         .4                                                                BGD



                                                                                   IND
     Racism



                             ECU                                                        EGY
                                                                                       VNM
                                                                             IDN

                                ZMB


                                                                                MYS
                                                             NGA
                                                                                      DZA
                                                                                          HKG
         .2




                                                   GHA           MLI     PHL

                                         UGA

                                      TZA                 MAR         ETH                  TUN
                             ZAF
                                         MEX



                                                                BFA               SGP
                                                                                PAK
                             PER
                             CHL


                          USA
            AUS                         GTM
            URY             BRA
               ARG          CAN
                          NZL
                          COL
         0




            0                     .2                     .4                .6               .8
                                               Technology 1500s
                                          Racism                   Fitted values



Figure 3: Scatter Plot Racism and Technology Index 1500s Colonial Sample
         .6




                                                   LBY




                       BGD
         .4




                           IND
     Racism




                                   EGY  ECU
                      VNM
                             IDN

                                ZMB


                                                         MYS
                       NGA
                                DZA
                                                                   HKG
         .2




            MLI             GHA        PHL

                      UGA

                  ETH   TZA            TUN
                                   MAR
                                                     ZAF
                                            MEX



                   BFA         PAK                                        SGP
                                    PER             CHL


                                                                                          USA
                                   GTM                                               AUS
                                             BRA       URY
                                             COL ARG                                      CAN
                                                                                       NZL
         0




               .2                       .4                     .6                .8             1
                                                Technology 2000
                                          Racism                   Fitted values



Figure 4: Scatter Plot Racism and Modern Technology Colonial Sample




                                                     45
 Appendix 1 : Samples for Cross Country and Multi-Level Analysis
 Colonial Sample   Non-Colonial Sample
 Argentina      Bosnia and Herzegovina
 Australia      China
 Algeria       Czech Republic
 Bangladesh      Finland
 Brazil        France
 Burkina Faso     Germany
 Canada        Hungary
 Chile        Iran
 Colombia       Iraq
 Ecuador       Italy
 Egypt        Japan
 Ethiopia       Lithuania
 Ghana        Netherlands
 Guatemala      Norway
 Hong Kong, China   Poland
 India        Romania
 Indonesia      Saudi Arabia
 Libya        Spain
 Malaysia       Sweden
 Mali         Switzerland
 Mexico        Thailand
 Morocco       Turkey
 New Zealand     Uganda
 Nigeria       Ukraine
 Pakistan       Uzbekistan
 Peru
 Philippines
 Singapore
 South Africa
 Tanzania
 Tunisia
 Uganda
 United States
 Uruguay
 Vietnam
 Zambia




      Appendix Table 2 : Summary Stats Country Level
Variable             Obs Mean Std. Dev. Min        Max
Racism                 60  0.177  0.116  0.024  0.540
Log Population Density 1500s      60  1.195  2.064  -3.831  5.643
Technology Index 1500s         60  0.575  0.305  0.000  1.000
Log GDP Per Capita in 2000       60  8.807  1.102  6.587  10.445
Technology Index 2000          60  0.508  0.220  0.174  1.012
Log of GDP Per Capita          60  8.298  1.597  5.079  10.933
Rule of Law               60  0.185  1.013  -1.649  1.941
World Governance Indicators 1996    60  0.173  0.972  -1.806  1.836
Average Years of Schooling 1985-1995  55  6.861  2.658  0.902  12.319




                    46
        Appendix Table 3 : Sum Stats Inpidual Level WVS
  Variable            Obs   Mean   Std. Dev. Min        Max
  Racism              220,605   0.169   0.375   0.000  1.000
  Log Population Density 1500s   220,604   0.930   2.148   -3.831  5.643
  Technology Index 1500s      220,605   0.521   0.322   0.000  1.000
  Satisfaction           218,190   6.834   2.376   1.000  10.000
  income              197,452   4.619   2.345   1.000  10.000
  education             189,566   4.606   2.223   1.000  8.000
  age                219,001   40.345   16.121   13.000  99.000
  sex                215,970   0.491   0.500   0.000  1.000
  Health              218,174  2.12876  0.872367     1    5
  Lack of Confidence Government   192,162  5.411137  2.887504     1    10




            Appendix Table 4 : First Stage IVs from Table 6
               (1)          (2)          (3)         (4)
Dependent Variables    Log Population Density 1500s         Technology Index 1500s
           Column 5 Panel A   Column 6 Panel A  Column 5 Panel B    Column 6 Panel B
Millennial of Agr     0.236***                  0.168***
              (0.075)                  (0.056)
State History        1.252*                  1.402***
              (0.623)                  (0.408)
Technology 0 CE                 1.293***                  1.641***
                         (0.436)                  (0.472)
Log Pop 1000 CE                 0.415***                   0.095
                         (0.070)                  (0.081)
Absolute Latitude      -1.380        0.788         -0.681       0.387
              (0.940)        (0.694)        (0.606)       (0.810)
Observations         35          32          35          32
R2             0.503        0.848         0.629        0.519
Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1




                      47
                      Appendix Table 5 : Replication Table 8 Multi-level Logit
                      1       2        3        4              5          6
   Panel A
                                       Dependent Variable: Racism
   Samples            Colonial Sample  Colonial Sample   Colonial Sample Colonial Sample    Colonial Sample  Non-Colonial
                                                                  Sample
   log population density, 1500   0.291***      0.296***      0.272***      0.285***      0.279***       0.014
                    (0.059)      (0.089)       (0.095)      (0.097)      (0.106)       (0.103)
   Observations            101,346      101,346       101,346      93,869       93,869       50,694
   Number of groups/clusters      36         36          36        33         33         20
   Panel B
                                       Dependent Variable: Racism




48
   Samples            Colonial Sample  Colonial Sample   Colonial Sample Colonial Sample    Colonial Sample  Non-Colonial
                                                                  Sample

   Technology Index, 1500 CE     2.384***      2.229***      2.065***      2.116***      2.598***      -3.970**
                    (0.432)      (0.418)       (0.435)      (0.480)      (0.404)       (1.584)
   Observations           101,346      101,346       101,346      93,869       93,869       50,694
   Number of groups/clusters      36         36          36        33         33         20
   Specifications
   Inpiudal Controls         Yes        Yes        Yes        Yes         Yes         Yes
   Control Log GDP Per Cap       No         Yes        No         No         Yes         Yes
   Control Insitituions         No         No         Yes        No         Yes         Yes
   Control Avg Schooling        No         No         No         Yes         Yes         Yes
   Wave Fixed Effects          Yes        Yes        Yes        Yes         Yes         Yes
   Notes : All regressions contain a constant. The unit of observation is the inpidual. All coefficients are reported in each column separated by
   panels. Inpidual level controls include: confidence in the government, income, level of education, life satisfaction, subjective health, a dummy
   for sex and age. Our dependent variables of interest are measures at the country level. Other controls are at the country level, which include:
   institutions, measured by the average of the world governance indicators index in 1996, log of GDP per capita in 2000 and average schooling
   from 1985-1995. Additionally, to account for time, we use wave fixed effects. Robust standard errors clustered by country are in parentheses
   *** p<.01, ** p<0.05, * p<0.1
      Appendix Table 6 : Reproduction Table 8 OLS with clustered standard errors by country
                  1     2      3       4     5      6
   Panel A
                           Dependent Variable: Racism
   Samples         Colonial  Colonial   Colonial   Colonial  Colonial  Non-
               Sample   Sample    Sample    Sample   Sample   Colonial
                                              Sample
   log population density, 1500   0.041***    0.045***    0.043***    0.041***   0.044***     0.009
                    (0.007)     (0.012)    (0.011)    (0.010)    (0.012)    (0.015)
   Observations           101,346     101,346    101,346     93,869    93,869     50,694
   R2                 0.067      0.067     0.067     0.071     0.072     0.054
   Number of groups/clusters      36       36       36      33       33      20
   Panel B
                                 Dependent Variable: Racism




49
   Samples            Colonial    Colonial    Colonial   Colonial    Colonial   Non-
                  Sample     Sample     Sample    Sample     Sample    Colonial
                                                        Sample

   Technology Index, 1500 CE    0.330***    0.299***    0.291***    0.290***   0.313***    -0.692***
                    (0.056)     (0.058)    (0.056)    (0.058)    (0.056)    (0.201)
   Observations           101,346     101,346    101,346     93,869    93,869     50,694
   R2                 0.074      0.078     0.079     0.082     0.086      0.068
   Number of groups/clusters      36       36       36      33       33       20
   Specifications
   Inpiudal Controls        Yes       Yes       Yes      Yes      Yes      Yes
   Control Log GDP Per Cap      No       Yes       No      No       Yes      Yes
   Control Insitituions        No       No       Yes      No       Yes      Yes
   Control Avg Schooling       No       No       No      Yes      Yes      Yes
   Wave Fixed Effects         Yes       Yes       Yes      Yes      Yes      Yes
   Notes : All regressions contain a constant. The unit of observation is the inpidual. All coefficients are reported in
   each column separated by panels. Inpidual level controls include: confidence in the government, income, level of
   education, life satisfaction, subjective health, a dummy for sex and age. Our dependent variables of interest are
   measures at the country level. Other controls are at the country level, which include: institutions, measured by the
   average of the world governance indicators index in 1996, log of GDP per capita in 2000 and average schooling from
   1985-1995. Additionally, to account for time, we use wave fixed effects. Robust standard errors clustered by country
   are in parentheses *** p<0.01, ** p<0.05, * p<0.1
   Appendix 7 : Controlling for Fractionalization, Migration and Genetic Diversity at the Inpidual Level
                  1       2        3        4   5      6
   Panel A
                              Dependent Variable: Racism
   Control Variables    Ethnic Frac  Lingustic Frac Religious Frac Migration  Genetic  AA Gentic
                                       Ratio  Diversity Diversity
   log population density, 1500   0.032***     0.033***     0.030***    0.037***   0.025**   0.037***
                    (0.009)     (0.009)      (0.009)     (0.010)   (0.010)    (0.011)
   Observations           98,929      98,929      98,929     101,346    101,346    86,405
   Number of Groups          35        35        35       36      36      30
   Panel B




50
                                 Dependent Variable: Racism
   Control Variables       Ethnic Frac   Lingustic Frac Religious Frac Migration      Genetic   AA Gentic
                                           Ratio      Diversity  Diversity
   Technology Index, 1500 CE    0.277***     0.266***    0.245***    0.264***    0.233***   0.302***
                    (0.076)     (0.069)     (0.069)    (0.064)     (0.075)   (0.080)
   Observations           98,929      98,929     98,929     101,346    101,346    86,405
   Number of Groups           35        35       35      36        36     30
   Specifications
   Inpiudal Controls        Yes        Yes       Yes       Yes      Yes     Yes
   Wave Fixed Effects         Yes        Yes       Yes       Yes      Yes     Yes
   Notes : All regressions contain a constant. The unit of observation is the inpidual. All coefficients are reported in
   each column separated by panels. Inpidual level controls include: confidence in the government, income, level of
   education, life satisfaction, subjective health, a dummy for sex and age. Our dependent variables of interest are
   measures at the country level. Other controls are at the country level which includes controls for absolute latitude in
   specifications. Additionally, to account for time, we use wave fixed effects. Robust standard errors clustered by
   country are in parentheses *** p<0.01, ** p<0.05, * p<0.1
    Appendix Table 8 : Controlling for Religion, Colonial Origin, and Other Cultural Factors Inpidual Level
                 1     2     3     4    5    6     7       8
   Panel A
                     Dependent Variable: Racism
   Control Variables    Religion  Colonial  Legal   Regional  Trust  Respect Obedience    Control
                    Origin   Origin   Effects
   log population density, 1500  0.026***  0.026***  0.033***   0.021**  0.032***  0.031***   0.032***    0.032***
                   (0.007)  (0.008)   (0.009)   (0.010)   (0.009)  (0.009)   (0.009)     (0.009)
   Observations          101,346   101,346   101,346   101,346   96,393  101,346   101,346     99,576
   Number of Groups          36     36     36     36     36    36      36       36
   Panel B




51
                        Dependent Variable: Racism
   Control Variables       Religion  Colonial Legal    Regional    Trust   Respect  Obedience    Control
                        Origin  Origin   Effects
   Technology Index, 1500 CE   0.261***  0.239***  0.255***   0.160**  0.260***  0.260***   0.264***    0.264***
                   (0.075)  (0.075)   (0.072)   (0.073)   (0.067)  (0.068)   (0.067)     (0.067)
   Observations          101,346   101,346   101,346   101,346   96,393  101,346   101,346     101,346
   Number of Groups          36     36     36     36     36    36      36       36
   Specifications
   Inpiudal Controls       Yes     Yes     Yes     Yes    Yes     Yes     Yes       Yes
   Wave Fixed Effects        Yes     Yes     Yes     Yes    Yes     Yes     Yes       Yes
   Notes : All regressions contain a constant. The unit of observation is the inpidual. All coefficients are reported in each column
   separated by panels. Inpidual level controls include: confidence in the government, income, level of education, life satisfaction,
   subjective health, a dummy for sex and age. Our dependent variables of interest are measures at the country level. Other controls
   are at the country level which includes controls for absolute latitude in specifications. Additionally, to account for time, we use
   wave fixed effects. In this case we do not restrict our sample to countries that have at least 15 observations. Robust standard
   errors clustered by country are in parentheses *** p<0.01, ** p<0.05, * p<0.1
References

Acemoglu, D., S. Johnson, and J. A. Robinson. 2002. “Reversal of Fortune. Geography and Institutions
  in the Making of the Modern World Income Distribution.” The Quarterly Journal of Economics,
  117(4): 1231–94.
Acemoglu, Daron, Francisco A. Gallego, and James A. Robinson. 2014. “Institutions, human capital,
  and development.” Annual Review of Economics, 6(1): 875–912.
Acemoglu, Daron, Simon Johnson, and James A. Robinson. 2001. “The Colonial Origins of
  Comparative Development. An Empirical Investigation.” American Economic Review, 91(5): 1369–
  401.
Acemoglu, Daron, Tristan Reed, and James A. Robinson. 2014. “Chiefs: Economic Development and
  Elite Control of Civil Society in Sierra Leone.” Journal of Political Economy, 122(2): 319–68.
Acemoglu, Daron, James A. Robinson, and Thierry Verdier. 2004. “Alfred Marshall Lecture.
  Kleptocracy and Divide-and-Rule: A Model of Personal Rule.” Journal of the European Economic
  Association, 2(2-3): 162–92.
Acharya, Avidit, Matthew Blackwell, and Maya Sen. 2016. “The Political Legacy of American
  Slavery.” The Journal of Politics, 78(3): 621–41.
Alesina, Alberto, Arnaud Devleeschauwer, William Easterly, Sergio Kurlat, and Romain Wacziarg.
  2003. “Fractionalization. Journal of Economic Growth.” Journal of Economic Growth, 8(2): 155–94.
Alesina, Alberto, William Easterly, and Janina Matuszeski. 2011. “Artificial Sates.” Journal of the
  European Economic Association, 9(2): 246–77.
Alesina, Alberto, and Nicola Fuchs-Schündeln. 2007. “Good-Bye Lenin (or Not? ): The Effect of
  Communism on People's Preferences.” American Economic Review, 97(4): 1507–28.
Alesina, Alberto, and Paola Giuliano. 2010. “The power of the family.” Journal of Economic Growth,
  15(2): 93–125.
Alesina, Alberto, and Paola Giuliano. 2015. “Culture and Institutions.” Journal of Economic Literature,
  53(4): 898–944.
Alesina, Alberto, Paola Giuliano, and Nathan Nunn. 2013. “On the Origins of Gender Roles. Women
  and the Plough.” Quarterly Journal of Economics, 128(2): 469–530.
Allport, Gordon W. 1954. The Nature of Prejudice. Cambridge, MA: Persues Books.
Anderson, Robert W., Noel D. Johnson, and Mark Koyama. 2017. “Jewish Persecutions and Weather
  Shocks. 1100–1800.” The Economic Journal, 127(602): 924–58.
Ashraf, Javed. 1994. “Differences in Returns to Education: An Analysis by Race.” American Journal of
  Economics and Sociology, 53(3): 281–90.
Ashraf, Quamrul, and Oded Galor. 2011. “Dynamics and Stagnation in the Malthusian Epoch.” The
  American Economic Review, 101(5): 2003–41.
Ashraf, Quamrul, and Oded Galor. 2013. “The “Out of Africa” hypothesis, human genetic persity,
  and comparative economic development.” The American Economic Review, 103(1): 1–46.
Association, World V. S. 2015. World Values Survey 1981–2014 Longitudinal Aggregate v.
  20150418: Madrid, Spain: Aggregate File Producer: JDSystems.
Autio, Erkko, Saurav Pathak, and Karl Wennberg. 2013. “Consequences of cultural practices for
  entrepreneurial behaviors.” Journal of International Business Studies, 44(4): 334–62.
Barro, Robert J., and Jong-Wha Lee. 2010. A New Data Set of Educational Attainment in the World,
  1950-2010. NberWorking Paper.
Becker, Gary S. 1957. The Economics of Discrimination. Chicago & London: University of Chicago
  Press.
Becker, Sascha O., Katrin Boeckh, Christa Hainz, and Ludger Woessmann. 2016. “The Empire Is
  Dead, Long Live the Empire! Long-Run Persistence of Trust and Corruption in the Bureaucracy.” The
  Economic Journal, 126(590): 40–74.
Becker, Sascha O., and Ludger Woessmann. 2009. “Was Weber Wrong? A Human Capital Theory of
  Protestant Economic History.” Quarterly Journal of Economics, 124(2): 531–96.
Berggren, Niclas, and Therese Nilsson. 2013. “Does Economic Freedom Foster Tolerance?” Kyklos,
  66(2): 177–207.
Berggren, Niclas, and Therese Nilsson. 2014. “Market institutions bring tolerance, especially where
  there is social trust.” Applied Economics Letters, 21(17): 1234–37.
Bisin, Alberto, and Thierry Verdier. 2011. “The Economics of Cultural Transmission and
  Socialization.” In Handbooks in economics. Vol. 1, Handbook of social economics. Volume 1A, ed.
  Jess Benhabib, Matthew O. Jackson, and Alberto Bisin, 339–416. Amsterdam, Boston: North
  Holland.
Bjørnskov, Christian, and Pierre-Guillaume Méon. 2013. “Is trust the missing root of institutions,
  education, and development? Public Choice.” Public Choice, 157(3-4): 641–69.
Blumer, Herbert. 1958. “Race relations as a sense of group position.” Pacific Sociological Review, 1: 3–
  7.
Bobo, Lawrence. 1991. “Social responsibility, inpidualism, and redistributive policies.” Sociological
  Forum, 6(1): 71–92.
Boyd, Robert, and Peter J. Richerson. 1985. Culture and the evolutionary process. Chicago,
  London: University of Chicago Press.
Buggle, Johannes C. 2016. “Law and social capital. Evidence from the Code Napoleon in Germany.”
  European Economic Review, 87: 148–75.
Cantoni, Davide, Yuyu Chen, David Y. Yang, Noam Yuchtman, and Y. J. Zhang. 2017. “Curriculum
  and Ideology.” Journal of Political Economy, 125(2): 338–92.
Chanda, Areendam, C. J. Cook, and Louis Putterman. 2014. “Persistence of Fortune. Accounting for
  Population Movements, There Was No Post-Columbian Reversal.” American Economic Journal:
  Macroeconomics, 6(3): 1–28.
Charles, Camille Z. 2000. “Neighborhood Racial-Composition Preferences. Evidence from a Multiethnic
  Metropolis.” Social Problems, 47(3): 379–407. https://academic.oup.com/socpro/article-
  pdf/47/3/379/4601887/socpro47-0379.pdf.
Charles, Camille Z. 2003. “The Dynamics of Racial Residential Segregation.” Annual review of
  sociology, 29: 167–207.
Charles, Kerwin, Kofi, and Jonathan Guryan. 2008. “Prejudice and Wages: An Empirical Assessment
  of Becker’s The Economics of Discrimination.” Journal of Political Economy, 116(5): 773–809.
Comin, Diego, William Easterly, and Erick Gong. 2010. “Was the Wealth of Nations Determined in
  1000 BC?” American Economic Journal: Macroeconomics, 2(3): 65–97.
Dickerson, Niki T., and Jerry A. Jacobs. 2013. “Race differentials in college selectivity, 1981–2000.”
  Journal of African American Studies, 10(1): 3–18.
Dustmann, Christian, and Ian P. Preston. 2007. “Racial and Economic Factors in Attitudes to
  Immigration.” The B.E. Journal of Economic Analysis & Policy, 7(1).
Easterly, William, and Ross Levine. 2016. “The European origins of economic development.” Journal
  of Economic Growth, 21(3): 225–57.
Engerman, Stanley L., and Kenneth L. Sokoloff. 1997. “Factor endowments, institutions, and
  differential paths of growth among new world economies.” How Latin America Fell Behind: 260–
  304.
Feagin, Joe, and Zinobia Bennefield. 2014. “Systemic racism and U.S. health care.” Structural Stigma
  and Population Health, 103: 7–14.
Fernandez, Raquel, and Alessandra Fogli. 2009. “Culture. An Empirical Investigation of Beliefs,
  Work, and Fertility.” American Economic Journal: Macroeconomics, 1(1): 146–77.
Ford, Robert. 2006. “Prejudice and White Majority Welfare Attitudes in the UK.” Journal of Elections,
  Public Opinion and Parties, 16(2): 141–56.
Fuchs-Schündeln, Nicola, and Paolo Masella. 2016. “Long-Lasting Effects of Socialist Education.”
  Review of Economics and Statistics, 98(3): 428–41.
Fukuyama, Francis. 2001. “Social capital, civil society and development.” Third World Quarterly,
  22(1): 7–20.
Galor, Oded, and Marc Klemp. 2017. Roots of Autocracy. Cambridge, MA: National Bureau of
  Economic Research.
Giavazzi, Francesco, Ivan Petkov, and Fabio Schiantarelli. 2014. Culture. Persistence and Evolution.
  Cambridge, MA: National Bureau of Economic Research.
Goldsmith, Arthur H., Darrick Hamilton, and William Darity. 2006. “Shades of Discrimination. Skin
  Tone and Wages.” American Economic Review, 96(2): 242–45.
Grosjean, Pauline. 2014. “A HISTORY OF VIOLENCE. THE CULTURE OF HONOR AND
  HOMICIDE IN THE US SOUTH.” Journal of the European Economic Association, 12(5): 1285–316.
Guiso, Luigi, Paola Sapienza, and Luigi Zingales. 2003. “People’s opium? Religion and economic
  attitudes.” Journal of Monetary Economics, 50(1): 225–82.
Guiso, Luigi, Paola Sapienza, and Luigi Zingales. 2006. “Does Culture Affect Economic Outcomes?”
  Journal of Economic Perspectives, 20(2): 23–48.
Guiso, Luigi, Paola Sapienza, and Luigi Zingales. 2016. “LONG-TERM PERSISTENCE.” Journal of
  the European Economic Association, 14(6): 1401–36.
Hello, Evelyn, Peer Scheepers, and Mérove Gijsberts. 2002. “Education and Ethnic Prejudice in
  Europe. Explanations for cross-national variances in the educational effect on ethnic prejudice.”
  Scandinavian Journal of Educational Research, 46(1): 5–24.
Hello, Evelyn, Peer Scheepers, and Peter Sleegers. 2006. “Why the more educated are less inclined to
  keep ethnic distance. An empirical test of four explanations.” Ethnic and Racial Studies, 29(5): 959–
  85.
Hodler, Roland. 2006. “The curse of natural resources in fractionalized countries.” European Economic
  Review, 50(6): 1367–86.
Hubert, Mia, Peter J. Rousseeuw, and Stefan van Aelst. 2008. “High-Breakdown Robust Multivariate
  Methods.” Statistical Science, 23(1): 92–119.
Hyman, Herbert H., and Charles R. Wright. 1979. “Education's Lasting Influence on Values.”
Jackman, Mary R., and Michael J. Muha. 1984. “Education and Intergroup Attitudes. Moral
  Enlightenment, Superficial Democratic Commitment, or Ideological Refinement?” American
  Sociological Review, 49(6): 751–69.
Jha, Saumitra. 2013. “Trade, Institutions, and Ethnic Tolerance. Evidence from South Asia.” The
  American Political Science Review, 107(4): 806–32.
Johnston, David W., and Grace Lordan. 2016. “Racial prejudice and labour market penalties during
  economic downturns.” European Economic Review, 84: 57–75.
Kaufmann, Daniel, Aart Kraay, and Massimo Mastruzzi. 2013. “World governance indicators.” The
  World Bank.
Klein, Katherine J., Paul D. Bliese, Steve W. J. Kozolowski, Fred Dansereau, Mark B. Gavin, Mark
  A. Griffin, David A. Hofmann, Lawrence R. James, Francis J. Yammarino, and Michelle C.
  Bligh. 2000. “Multilevel analytical techniques. Commonalities, differences, and continuing
  questions.” In Multilevel theory, research, and methods in organizations: Foundations, extensions,
  and new directions, 512–53. San Francisco, CA, US: Jossey-Bass.
Koopmans, Ruud, and Susanne Veit. 2014a. “Cooperation in Ethnically Diverse Neighborhoods: A
  Lost-Letter Experiment.” Political Psychology, 35(3): 379–400.
Koopmans, Ruud, and Susanne Veit. 2014b. “Ethnic persity, trust, and the mediating role of positive
  and negative interethnic contact: A priming experiment.” Social Science Research, 47: 91–107.
La Porta, R. 1999. “The quality of government.” Journal of Law, Economics, and Organization,
  15(1): 222–79.
Lang, Kevin, and Jee-Yeon K. Lehmann. 2012. “Racial Discrimination in the Labor Market. Theory
  and Empirics.” Journal of Economic Literature, 50(4): 959–1006.
Lang, Kevin, and Michael Manove. 2011. “Education and Labor Market Discrimination.” American
  Economic Review, 101(4): 1467–96.
Lang, Kevin, Michael Manove, and William T. Dickens. 2005. “Racial Discrimination in Labor
  Markets with Posted Wage Offers.” American Economic Review, 95(4): 1327–40.
McEvedy, Colin, and Richard Jones. 1978. Atlas of world population history: Penguin Books Ltd,
  Harmondsworth, Middlesex, England.
Nunn, Nathan. 2012. “Culture and the Historical Process.” Economic History of Developing Regions,
  27(sup1): S108-S126.
Nunn, Nathan. 2014. “Historical Development.” In Handbook of Economic Growth, ed. Philippe
  Aghion, and Steven Durlauf, 347–402: North-Holland.
Nunn, Nathan, and Leonard Wantchekon. 2011. “The Slave Trade and the Origins of Mistrust in
  Africa.” The American Economic Review, 101(7): 3221–52.
Posner, E. A., K. E. Spier, and A. Vermeule. 2010. “Divide and Conquer.” Journal of Legal Analysis,
  2(2): 417–71.
Putnam, Robert, Robert Leonardi, and Raphaëlle Nanetti. 1993. “Making democracy work
  Princeton.” NJ: Princeton University Pi-ess.
Putnam, Robert D. 2007. “E Pluribus Unum. Diversity and Community in the Twenty-first Century The
  2006 Johan Skytte Prize Lecture.” Scandinavian Political Studies, 30(2): 137–74.
Putterman, Louis. 2008. “Agriculture, Diffusion and Development. Ripple Effects of the Neolithic
  Revolution.” Economica, 75(300): 729–48.
Putterman, Louis, and Cary A. Trainor. 2006. “Agricultural transition year country data set.” Brown
  University: 729–48.
Putterman, Louis, and David N. Weil. 2010. “Post-1500 Population Flows and the Long Run
  Determinants of Economic Growth and Inequality.” Quarterly Journal of Economics-, 125(4): 1627–
  82.
Richerson, Peter J., and Robert Boyd. 2005. Not by genes alone. How culture transformed human
  evolution. Chicago, Ill.: Univ. of Chicago Press.
Selznick, Gertrude J., and Stephen Steinberg. 1969. “The tenacity of prejudice. Anti-Semitism in
  contemporary America.”
Spolaore, Enrico, and Romain Wacziarg. 2009. “The Diffusion of Development *.” Quarterly Journal
  of Economics, 124(2): 469–529.
Spolaore, Enrico, and Romain Wacziarg. 2013. “How Deep Are the Roots of Economic
  Development?” Journal of Economic Literature, 51(2): 325–69.
Stolle, Dietlind, Stuart Soroka, and Richard Johnston. 2008. “When Does Diversity Erode Trust?
  Neighborhood Diversity, Interpersonal Trust and the Mediating Effect of Social Interactions.”
  Political Studies, 56(1): 57–75.
Tabellini, Guido. 2008. “Institutions and Culture.” Journal of the European Economic Association, 6(2-
  3): 255–94.
Tabellini, Guido. 2010. “Culture and Institutions: Economic Development in the Regions of Europe.”
  Journal of the European Economic Association, 8(4): 677–716.
Voigtlander, N., and H.-J. Voth. 2012. “Persecution Perpetuated. The Medieval Origins of Anti-Semitic
  Violence in Nazi Germany.” Quarterly Journal of Economics, 127(3): 1339–92.
Woodberry, Robert D. 2012. “The Missionary Roots of Liberal Democracy.” American Political
  Science Review, 106(02): 244–74.
Yohai, Victor J. 1987. “High Breakdown-Point and High Efficiency Robust Estimates for Regression.”
  The Annals of Statistics, 15(2): 642–56.