Altitude, Genes and Town Names: A Study of HistoricalIn�uence and Political Mobilization

Christopher PaikPrinceton Universitycpaik@princeton.edu

July 2011

Abstract

This paper investigates the historical in�uence of ethnic integration on current politicaloutcomes. It draws evidence from various regions in the Tibet Autonomous Region (TAR)and the surrounding counties in Tibet Autonomous Prefectures which recently witnessedwaves of protests in 2008. Using a novel town name index database that identi�es theethnolinguistic origin of each town name, the empirical �ndings show that regions witnessingChinese presence in the past also experienced fewer protests. This paper also introducesan altitude threshold variable to identify potential exogenous variations separating Chinesefrom Tibetans. Above the threshold elevation, altitude illness due to loss of oxygen becomesincreasingly common among Chinese but not Tibetans. The challenging environment acts asa physiological hindrance only for Chinese, while Tibetans�successful genetic adaptations tothe highlands are shown in their phenotypes distinct from their Chinese counterparts. Usingthe threshold as an instrument for the selective Chinese in�uence in certain regions, thispaper shows that the signi�cance of the historical presence of Chinese on current uprisingsremains robust to a series of both historical and contemporaneous variables, and discussesinstitutional mechanisms to explain sustained leniency towards the Chinese state. The�ndings con�rm the importance of long term settlement policies in reducing unrests, andsupport other works suggesting that history continues to in�uence current socioeconomicoutcomes.

1 Introduction

In 2008, ethnic Tibetan areas witnessed one of the largest waves of protest and social unrest inrecent decades. On March 16, 2008 hundreds of monks from Kirti monastery in Aba County,Sichuan Province, broke out of the cordon of security forces, and joined a mass demonstrationcalling for Tibet�s independence (ICT 2008, Pg.73). The protests led to violence coming fromboth protestors and Chinese security forces, and ended in mass imprisonment of the protestorsand heavy casualties. Just 200 kilometers southwest of Aba County is Heishui County, which islocated within the same prefecture. According to the latest population census in 2000, HeishuiCounty�s population, at 62312, was comparable to Aba County�s 57000. Furthermore, theproportions of Tibetan residents for both counties were the same at 91 percent. Yet over thecourse of the four months between March and June in 2008, when the protest movement sweptacross both Tibet Autonomous Region (TAR) as well as the surrounding prefectures in Sichuan,Gansu, Qinghai and Yunnan Provinces, Heishui County witnessed zero reported incident of anyprotest. Such contrasting patterns of violence were also found within TAR. Gongjue County inChamdo Prefecture had population of 43160 people, with 98 percent of the population identi�edas Tibetan, and the rest identi�ed as Han. During March and April of 2008, the countywitnessed three separate occasions of layperson protests and consequent detentions enforced bythe Chinese security forces (ICT 2008, Pg.73). In contrast Leiwuqi County, with comparablepopulation of 40895 people and only 214 kilometers away from Gongjue, experienced no protest.The county was also predominantly populated by Tibetans (98 percent) and Han Chinese (2percent).

A number of factors commonly included as determinants to civil con�ict, such as povertyand political instability (Fearon & Laitin 2003, Miguel, Satyanath & Sergenti 2004) may playimportant roles in the rise of protests in Tibet. The current encroachment of Han migrants,and disruption of cultural unity of the geographically salient Tibetans as a catalyst for protest,also �nd support in studies that have long suggested changing group settlement patterns to begood indicators for group con�ict behavior (Horowitz 1985, Posen 1993, Fearon 1998, Toft 2002,Toft 2003, Weidmann 2009). According to these works, contemporaneous diminishing groupconcentration and increased lenience towards the Chinese are factors that decrease politicalmobilization. However, the aggregate statistics of protest movements in Tibet fails to explainregional variations in Chinese-Tibetan relations. Even within TAR, di¤erences in counties suchas Gongjue and Leiwuqi in terms of their reaction to fellow Tibetan protestors suggests thatthere may be unexplained factors contributing to con�ict between the two population groups.

This paper argues that the varying levels of con�ict in regions of otherwise similar sociode-mography and political settings may be explained by a historical factor, in addition to ethniccomposition, population and policy changes in the modern period. Speci�cally, the paper analy-ses the impact of historical Chinese presence in various parts of Tibet, on the regions�presentattitude towards the central administration as represented by recent protests.1 It argues thatthe degree to which the Chinese historically in�uenced certain locales determined in the longrun the level of acceptance of the Chinese state dominance in the region. In tracing each town�shistory, this paper argues that the town name is a legacy of the region, and uses the origin ofthe town name to infer historical Chinese presence.2 It then provides empirical evidence that

1Here the term Chinese predominantly refers to the Han people but is inclusive of Manchus who founded theQing Dynasty, since Manchus and their language have been assimilated into the Han population. The term isexclusive of ethnic groups that have played major roles in political mobilization against the central administration,such as Yugurs, Mongolians, and Tibetans. The data in this paper only cover Tibetan protests, and thereforeonly include towns that have either Tibetan or Chinese heritage as indicated by their names.

2While some of these town names are recent outcomes of the central government�s e¤ort to establish Han-

the variation in the level of protests in 2008 is explained by this measure of Chinese heritageon Tibetan regions. The �ndings show that counties with more towns of Chinese names wereless likely to experience any protest.

Furthermore, this paper employs a novel instrumental variable approach to show that themain result is not driven by unobserved di¤erences, emerging from the Chinese� selection ofa place based on the level of the highlanders�cultural attitude towards outsiders. It uses thegenetic di¤erence between Chinese and Tibetans to argue that the former was physiologicallyless capable of settling in certain regions, while the latter encountered no such problem in theirnative lands. Altitude illness is a well-known medical condition that occurs when a personascends to high mountainous slopes and su¤ers from lack of oxygen (known as hypoxia). Nu-merous studies have shown that Tibetans and Chinese have di¤erent genetic compositions, andthat certain genes carried by Tibetans allow them to adopt better to the thin air surroundingsthan Chinese. A recent quasi-natural experiment has also shown that above 4500 meters in al-titude, only the Chinese subjects showed symptoms of altitude illness despite a previous periodof acclimatization, while the Tibetan counterparts showed no ailment. The genetic di¤erencebetween the two groups likely comes from long term biological adoption by Tibetans going backthousands of years, prior to the advent of the outsiders.

This paper suggests that the genetic adaptation by Tibetans to high altitudes allowed anatural separation between the two groups in highlands of Tibet, while those in lowlands haveexperienced more integration between the two groups in history. It then presents a potentialinstitutional mechanism under which ethnic integration led to the residents�acceptance of theChinese dominion over the region in the long run. Historically the most prominent institutionsin Tibet have been the Buddhist monasteries. Tibetan Buddhism and its monasteries not onlyhave been instrumental in developing Tibetan culture, but they have also been the social nucleiof political dissent and Tibetan nationalism. In towns with large monasteries, generally foundin the regions of Tibetan heritage, the teachings of past religious leaders have been passedon to strengthen Tibetan culture and solidarity. These values fostered Tibetan nationalismamongst monks and Tibetans, who have traditionally been connected to their local templesby serf obligations and familial ties. This meant that local Tibetans and their monasterieshad an invested relationship that diminished the members�subsequent incentives for becomingpart of other monasteries. Strict classi�cation of Buddhist colleges based on geographic originalso likely led to highly localized devotion of the monks to their monasteries. In addition tothe tendency of both laypeople and monks to devote themselves to local towns, migration ofpeople was restricted in Tibet throughout much of its history. The majority of the country�sland and people were organized into manorial estates, and most Tibetans were aristocrats andserfs bound to the lands. Thus any spread of culture and Tibetan identity through migrationwould have occurred only after 1959 when the manorial system was abolished. By this time,however, the Chinese state began to repress Buddhist monasteries such that the spread ofTibetan nationalism to traditionally Chinese regions would have been di¢ cult. All these factorscame together to create a setting in which Tibetan culture and identity were reinforced overgenerations only in historically Tibetan-dominated areas.

While this paper follows closely the literature in ethnic demography and con�ict (Montalvo& Reynal-Querol 2005, Fearon & Laitin 2003, Horowitz 1985, Posen 1993), by emphasizing thelong term consequences of ethnic policy choices and inter-ethnic relations, it also provides a newperspective on the studies that often focus on contemporary issues of ethnicity and civil con�ict.

dominated towns in Tibet, most go back in history at least to the 13th Century. This was the period when the�rst formal contact took place between the Qing Dynasty and Tibet, a kingdom then only known by pilgrimsand surrounded by almost insurmountable natural barriers restricting access for foreigners.

The paper furthermore presents supporting evidence of the importance of historical institutions,following the works of Acemoglu, Johnson & Robinson (2001), Acemoglu, Johnson & Robinson(2002), Acemoglu, Johnson & Robinson (2005) and Jha (2009). It suggests that inter-ethnicrelations of the past and their legacies should not be overlooked in considering modern politicaloutcomes. In arguing that genetic di¤erences cause a natural divide between the two groups,this paper also builds on recent works in economics using interdisciplinary approach to tacklethe origin of ethnolinguistic groups and endogeneity issues related to ethnicity and con�ict(Spolaore & Wacziarg 2009, Michalopoulos 2009, Miguel et al. 2004). Finally in the Tibetancontext, this paper complements seminal works by Goldstein (1997) and Barnett (2009) byemploying quantitative methods to study the region that has hitherto lacked such analysis. Itbuilds upon the authors�historical accounts and qualitative analysis of Tibet�s role on China�sdomestic stability and its territorial integrity, and the region�s importance in China�s diplomacyabroad.

The rest of the paper proceeds as follows: Section 2 gives a brief background history of therelation between Tibet and China. Section 3 discusses the empirical strategy. Section 4 presentsarguments for how town names may be appropriate measures of the regions�history. Section5 describes altitude sickness and genetic factors as suitable bases for exogenous variation andnatural selection of historical Chinese presence. Section 6 presents descriptions of the town nameindex database and other variables. Section 7 provides the empirical �ndings, and Section 8discusses potential institutional mechanisms which explain the variations in the recent politicalmobilization. Section 9 concludes with relevant policy implications.

2 Tibet and China: Past and Present

The history between Tibet and China has been written through the famous tea-horse trade be-tween Chinese tea merchants and Tibetan horse traders, and other relations that the neighbor-ing regions built with one another over millennia. The debate over Tibet�s sovereignty howeveris a phenomenon resulting from changes in its relatively recent political landscape. Accordingto the o¢ cial Chinese government stance Tibet became part of China during the Mongol YuanDynasty in the 13th century. In contrast, Tibetan nationalists allege the relationship betweenTibetans and Mongols was a �priest-patron�one, and Tibet was only subjugated to the Mon-gols in the same way as China in a Mongol empire centered in China (Goldstein 1997, Pg. 4).The Manchu Qing Dynasty from the 17th century managed to induce subordination in Tibet,and Qing emperors introduced a series of reforms to reorganize Tibetan religious and politicalinstitutions. Manchu imperial residents (amban) were also stationed in Lhasa to �keep a closewatch on the leaders of Tibet and oversee the garrison in Lhasa�(Goldstein 1997, Pg. 6). In1912, Qing collapsed, and the last amban was banned from Tibet. For the next four decades,Tibet enjoyed de facto independence from China, as the Republic of China (ROC) governmentwas unable to assert authority on Tibet, despite the ROC government�s claim of sovereignty(Lin 2006, Goldstein & Rimpoche 1989). Nevertheless, Tibet�s claim for independence duringthis period did not receive international recognition. The Silma Convention signed betweenTibet and Britain �declared that Tibet would be autonomous from China, but also acknowl-edged Chinese suzerainty over Tibet� (Goldstein 1997, Pg. 33). The ambiguous relationshipbetween Tibet and China was to be cleared when the Chinese Communist Party (CCP) cameinto power. After the CCP achieved victory in the Chinese Civil War in 1949, the new com-munist regime made clear its mission to reunify China according to the territory boundary ofthe Qing. The CCP ensured that the annexation of Tibet was one of the main tasks for thePeople�s Liberation Army (PLA), and in October 1950 the PLA captured Eastern Tibet. With

its demonstration of military prowess, the PLA however did not immediately march towardsLhasa. Instead, Mao Zedong intended to use negotiation to peacefully bring Tibet into fold.With the Tibetan government failing to secure international support through its �rst appeal tothe United Nations in 1950, the government decided the best option was to start serious nego-tiations with Beijing (Goldstein 1997, Pg. 95). On May 23, 1951, the famous Seventeen-PointAgreement was signed, which formally ended Tibet�s de facto independence.

A series of protests ensued after Tibet lost its independence. The �rst widespread revoltsin ethnic Tibetan regions occurred in Sichuan, Qinghai and Gansu provinces in 1956, and metwith severe repression from the PLA troops. In March 1959, a popular revolt among Tibetansin Lhasa followed after a rumor had spread of the Dalai Lama being kidnapped by the PLA.Afterwards Tibetan guerilla �ghters established a base in Mustang, Nepal and the clandestineresistance army continued for a decade, despite the limited success in making inroads into Tibetitself. Within Tibet, during the Cultural Revolution, another main revolt occurred in Nyemoin 1969, which was characterized by extreme brutality against the Chinese presence. Under theleadership of a Tibetan nun Thrinley Choedron, a small guerrilla war was waged against theChinese state, but was quickly paci�ed by the PLA (Goldstein, Jiao & Tanzen 2009). Between1987 and 1989, Tibet yet again witnessed mass protests and unrest in Lhasa , starting withmonks marching from Drepung monastery to the capital city Lhasa to stage a pro-independencedemonstration. Throughout the 1990s and 2000s, sporadic protests waged by Tibetans in Chinaoccurred, leading up to the protests in 2008.

In October 2007, when monks at Drepung Monastery in Lhasa attempted to celebrate theawarding of the US Congressional Gold Medal to the Dalai Lama, several monks were reportedlyarrested by the Chinese security forces (ICT 2008). On March 10, 2008, the 49th anniversaryof the Tibet uprising in 1959, monks in several monasteries in Lhasa marched through thecenter of the city, demanding the release of the previously arrested monks, but also shoutingpro-independence slogans and waving the Tibetan snow lion �ag (Topgyal 2011). The protestsin 2008 spread from Lhasa, the capital city of TAR, to other ethnic Tibetan areas in Sichuan,Gansu, and Qinghai Provinces. They continued for several months from March well into thesummer of 2008, right before the start of the Beijing Olympic Games. With calls for support forthe Tibetans and outcries against Beijing�s hard-line crackdown, the global Olympic torch rallyin many Western societies became the stage of pro-Tibet protestors clashing with pro-Chinanationalists, emboldening Tibetans�understanding of the amount of international support theywould garner. Severe crackdown from the Chinese state followed, which again led to furtherresistance protests targeting the sending of patriotic education teams or paramilitary troopsinto local monasteries (Barnett 2009, Pg. 14). Throughout the spring, there were more than100 protests in ethnic Tibetan areas.

The protests in 2008 are notable for their scope and pervasiveness, as the movement spreadfar beyond Lhasa and was sustained for an extended period of time involving both monks as wellas civilians and educated elites. To the author�s knowledge, it is also the very �rst time in Tibetwhen incident reports with detailed location information were recorded and made available,allowing for systematic analysis of the political mobilization and con�icts in the region. Unlikeprevious accounts of protests that were mainly anecdotal with limited resources for veri�cation,both the central Chinese government and the foreign press had reported detailed accounts ofincidents that occurred throughout the year.3 This paper introduces a dataset mainly basedon advocacy groups (the International Campaign for Tibet, TibetInfoNet and Central Tibetan

3Barnett (2009) for example notes that in 2008 the Chinese authorities enacted a new policy of having theo¢ cial media responding to any report in the foreign press of a Tibetan incident, in an e¤ort to gain control ofrepresentations. The media usually con�rmed the outlines of the report but characterized it di¤erently.

Administration) but with reports based on both foreign and Chinese media.

3 Empirical Strategy

This paper�s empirical approach is to use town names as indicators of historical Chinese presencein the region. The town names represent historical Chinese in�uence that in the long run led tothe lessening of tension between the two ethnic groups. As will be shown below, the existence ofChinese town name does not indicate current Chinese population dominance in the town, butrather suggests Chinese presence in history. The paper therefore does not take the di¤erencein the origin of names as an instrument for modern cultural di¤erences across regions, as thisapproach would likely fail the exclusion restriction. After all, what de�nes Tibetan culturetoday is probably correlated with at least part of the unobservables explaining the level ofuprisings. In the initial set up, the empirical strategy �rst makes the assumption that whenthe Chinese moved to the west towards Tibet, they made their marks in the lands close to theborder and attractive for trade and military purposes. That is, whether or not to sinicize aregion was a function of how much one could gain from trade or territorial takeover, as well ashow close it was to the Chinese-dominant regions. This assumption will be violated if regionsdi¤ered in important ways in their initial conditions that may also have had an e¤ect on theseareas, for example, through variations in the vegetation suitable for the Chinese livelihood. Themore attractive for agriculture a region was, for example, the more likely it would have beensettled by incoming Chinese migrants. Another factor in Chinese integration would have beenproximity and the ease with which certain towns could be travelled to. The ancient trade routesfor example would have been developed along the paths that were likely to be least strenuousfor travelers. In order to address these potential concerns, this paper introduces a rich set ofgeographic control variables as well as distance variables taking into consideration the cost oftravelling in the past.

A potential concern arises from assuming that the selection of certain regions for interactingwith Tibetans was uncorrelated with the cultural background of highlanders. It is plausible thatthe Chinese may have disliked geographically similar places for unobservable reasons correlatedwith current level of uprisings, such as certain regions having particularly hostile local Tibetanhighlanders, and this historically combatant population may have in�uenced their descendantsto develop strong Tibetan attitudes against Chinese. To assess whether this is the case, analternative approach relaxes the exogeneity assumption and uses the likelihood that the Chinesewould have become ill due to high altitudes as the instrument for their ability to settle inthe area. Speci�cally, in order to show that the results above are not driven by unobserveddi¤erences emerging from the selection of where to sinicize, the following set of regressions usesan altitude threshold drawn from Tibet�s unique topography as an instrument for historicalChinese presence in the region. The paper introduces the presence of altitude illness, muchmore common among Chinese than Tibetans, as a factor correlated with the propensity ofChinese live in lowlands but much less so up in higher altitudes. The initial Chinese presencein certain Tibetan towns may have been in�uenced by the group�s inability to adapt to newaltitudes, and this physiological hindrance to living in certain places led to di¤erent levels ofhistorical Chinese presence across Tibet and its surrounding regions. This variation in turnhas had resilient e¤ect on the local attitudes towards the central government, as re�ected onthe current levels of political mobilization. If a region was located at an altitude above thethreshold, the region remained predominantly Tibetan as outsiders became more exposed toaltitude illness. Assuming that the regions located above the threshold were not any more likelythan otherwise geographically similar ones (in terms of vegetation, elevation and proximity to

the borders and trade centers) to have attracted Tibetan population more hostile towards theChinese prior to the initial interaction with the outsiders, the instrumental variable approachcan be useful in assessing the impact of historical sinicization to the current trends in politicalmobilization in Tibet.

Altitude illness is a collective term that encompasses the major conditions caused directly byhypobaric hypoxia, or lack of oxygen due to falling in partial pressure oxygen in high altitudes(Murdoch, Pollard & Gibbs 2006). It is the only condition in which traditional technology is in-capable of mediating its e¤ects among the various environmental stresses that modern humanshave encountered and succeeded in overcoming (Yi & et al. 2011). Deprivation of adequateoxygen supply leads to severe physiological stress, leading to a number of symptoms includingheadache, loss of appetite, nausea, vomiting, fatigue, weakness, dizziness, light-headedness orsleep disturbance. Left untreated, severe forms of altitude illness may result; Acute MountainSickness (AMS) might progress to fatal high altitude cerebral oedema (HACE), or swelling ofextremities and face, that leads to brainstem herniation, or high altitude pulmonary oedema(HAPE), or swelling of lungs that lead to di¢ culty in breathing. Murdoch et al. (2006) describesthe general impact of oxygen decrease as one ascends mountain slopes; in the range between1500 and 2500 meters there is a decrease in exercise performance and ventilation, but arterialoxygen saturation remains above 90 percent. Between the 2500 and 3500 meter range, theillness becomes common, and past between 3500 and 5800 meters, arterial oxygen saturationfalls below 90%, below the normal range of 93 to 100 percent. Above 5800 meters, progres-sive physiological deterioration occurs because successful acclimatization, or adjustment to thehypoxia of high altitude, cannot be achieved, and permanent human habitation is impossible.Murdoch et al. (2006) argues that while there is considerable interpersonal variation in theability to acclimatize, there is a strongly direct correlation between altitude and the occurrenceof illness.4

Due to large inter-individual di¤erences in the ability to acclimatize, as well as lack of large-scale studies reporting the e¤ects of exposure to high altitude on the prevalence of altitudeillness, measuring the elevation at which the illness occurs for Chinese is di¢ cult. Fortunately,the period between 2001 and 2005 in Tibet provides an invaluable quasi-natural experiment onhow the population adopt to high altitudes. During this period, the construction of the highestrailroad in the world, the Qinghai-Tibetan Railroad, was completed (West 2006).5 Each yearmore than 20000 construction workers ascended to the various construction sites. The 5 yearperiod, during which a total of 78712 lowland workers ascended to high altitudes to work on therailroad, provides a unique study setting on a massive scale relevant to this paper�s empiricalanalysis. A randomized experiment taken during the period included 600 lowlanders, who wereall healthy, nonsmoking, male Han Chinese born and lived at sea level, who had no historyof prior exposure to high altitude, who were free from any pre-existing disease before ascentor using any medication, and who commuted for 5 years between near sea level and 4500meters. This group of subjects was compared to 600 other lowland workers, recruited eachyear upon their �rst ascent to high altitude as newcomers. The sample pool was also comparedto 200 Tibetan workers native to the average 4500-meter altitude environment. The incidence

4Murdoch et al. (2006) provides a series of examples, in which approximately 25% of visitors who ascendrapidly to altitudes between 2000 and 3000 meters in Colarado experience acute mountain sickness, while inNepal, about 50% developed AMS at altitudes above 4000 meters, even after 5 or more days of acclimatization.Acclimatization takes longer the higher the altitude, with an ascent rate of 300-400 meters per day above 3000meters.

5Wu & et al. (2009) describes the topography on which the railways were built on; the track connects Golmud(2808m) and Lhasa (3658m), and runs 1142 kilometers; about 85% of the track is situated above 4000 meters,culminating at 5072 meters when crossing the Tanggula Pass (Wu & et al. 2009).

and severity of AMS in commuters were lower upon each subsequent exposure, whereas theyremained similar in newcomers each year. AMS susceptibility was thus lowered by repeatedexposure to altitude, but still higher than the Tibetan counterparts who had no AMS; exposurestherefore increasingly protected lowlanders against AMS, but did not allow attaining the level ofadaptation of altitude natives. Importantly, Qi & et al. (2009) reports in a separate study thatafter 6 days�acclimatization at the altitude of 2800 meters, the adaptive Han Chinese railroadworkers were assigned to di¤erent workshop sections ranging from 4000 to 5072 meters abovesea level and participated in the railway construction. The onset of HAPE was present within2�6 days after ascending only to an altitude greater than 4500 meters above sea level. Thatis, while most healthy workers were easily adopted to the high altitude of 2500 meters, whichis the threshold commonly attributed as the altitude above which altitude sickness occurs, the4500-meter mark presented severe challenges with increased incidents of altitude illness.6 Inaccordance with these �ndings, this paper uses 4500 meters as the natural threshold dividingregions that can be easily acclimatized by Chinese, vs. those that can only be adopted bynative highlanders. The altitude level approximates Tibet�s average elevation (4438 meters),and the level above which hypoxia among Chinese becomes common even after a period ofacclimatization.

Another major issue to be addressed is that, as is common in studies looking at long-term in�uence of historical variables, certain regions were likely subject to di¤erent politicalin�uences subsequent to Chinese dominance. In the case of Tibet, each Buddhist sect had adi¤erent level of political involvement; the protest outcome may therefore have been derivedfrom in�uences of certain prominent sects over others. This paper introduces a set of historicalTibetan Buddhist sects data in order to control for this concern. Finally, A closely related study(Han & Paik 2011) also argues that after 1949, when the Chinese Communist Party(CCP) wonthe Chinese Civil War, it was committed to reunify China according to the territory boundaryof the Qing Dynasty. This meant that more than any other period in history, annexation ofTibet as part of China became a top concern for the newly established administration. As ameasure to ensure the "liberation" of Tibet, the Chinese government went on to implement bothrepressive and concessionary measures to control many aspects of Tibetan Buddhism, which inturn had important consequences to the protest movements. This paper therefore also controlsfor the potential impact that the policy changes on religious freedom had on di¤erent regions.

4 Town Names and Heritage

This paper uses the names given to each town as artifacts re�ecting historical Chinese in�uencein Tibet. The assertion that place names, or toponyms, are outcomes of the past culturaland political landscape �nds much support from the linguistics and geography literature. Forexample, Radding & Western (2010) shows that names are given intentionally to impart acertain meaning, and they have layers of meanings that people do not want to see erased;toponyms are deeply entrenched in culture, and their substitution come with consequences. Theauthors in Radding & Western (2010) go on to argue that even literal or phonetic translationis one type of name change that can be construed as a political act, since the meaning ofnames is embedded in cultures, and translating a name runs the risk of destroying its meaning.That is, the literal translation of a place name itself can make the place inconsequential, as the

6On a similar note, the key altitude ranges de�ned by the Indian Army for its acclimitisation schedules forits foot soldier is set up as follows: Stage I (2700m-3600m), Stage II (3600m-4500m), and Stage III (>4500m)(Tyagi & Malik 2008)

new language may not necessarily have a word with precisely the same meaning, since the storybehind the toponym may be connected to the very Tibetanness of the name. Furthermore, oftenin the case of phonetic translation, there is likely a mismatch of meanings of the Tibetan nameelements- inevitably leading to a new name purely arbitrary referent of a word. When a politicalregime takes over another, multiple place names may be created, each used by di¤erent groups.Even more dramatic shifts in political regimes are re�ected in the complete loss of originaltoponyms. The name changes can therefore be interpreted as outcomes of political shifts, oftenimposed by colonial authorities who use name changes as means to aid in the establishmentand legitimation of a state (Radding & Western 2010). Horsman (2006) similarly argues thatbestowing a name on a physical entity is an act of appropriation, and with each change inpolitical regime comes new names for a given geographical location. Place names are therefore"one of the most conservative elements in a language... surviving even repeated language shift"(Burenhult & Levinson 2008). Their resilience is re�ected often in cases when both internalsocietal changes and external threats from another language or society supplant the existingone, the place name may well remain (Herman 1999).

Kangding, the prefectural capital of Ganzi in Tibet Autonomous Prefecture (TAP), SichuanProvince, is a good example of how a town name represents the region�s history. The cityhas historically been known among Tibetans as Dartsedo. In an old Qing map publishedby Christian missionaries in 1717,7 Dartsedo is labeled as Tachienlu, apparently the phonetictranslation of Dartsedo. Since the Qing Dynasty, the Chinese kept a garrisoned o¢ ce in thisimportant trading mart for both Tibetan and Chinese merchants. The name Kanding representsthe historic Chinese presence in the region; in 1725 Qing military occupied Dartsedo area anda place called Kata near Dartsedo, which was made the military headquarters for the ChineseTroops. The capital was subsequently renamed Kangding by the Chinese. Xihai City onthe other hand is the prefectural capital of Haibei TAP, Qinghai Province with no Tibetanheritage. Xihai, also known as Two Two One Factor (Er Er Yao Chang), is a former site ofChina�s top-secret nuclear weapons research. Xihai means the western sea, the name of the�rst and only Chinese administrative unit established in the immediate locale before the 20thcentury. The name therefore re�ects a carefully-chosen historical parallel for the Chinese, andtakes after the original "Xihai Commandery," a military settlement whose ruins lie little morethan ten kilometers away. (Marshall & Cooke 1997, Pg. 1724). Similarly the town of Bayi inNyingchi County, Tibet, is a relatively new development site under the direction of the centralgovernment. Historically the town was known as Lhabagar and sparsely populated, but dueto its favorable climate and low altitude for Han Chinese, the town was heavily developedsince 1960s. Bayi literally means 81, named after the founding date of the Chinese Peoples�Liberation Army on August 1st, 1927. The town went through rapid development after thesigning of so called 17-points agreement with Tibetan government in 1951, which e¤ectivelyended the sovereignty of Tibet.

5 Altitude and Genes

Ancient records indicate that human di¢ culty with the rare�ed air at high altitudes has existedfor long time. Writings attributed to Aristotle (384-322 BC) for example describe travel onMount Olympus in Macedonia: "Also, because the rarity of the air which was there did not �llthem with breath, they were not able to survive there unless they applied moist sponges to their

7d�Anville, Jean Baptiste Bourguignon. "Nouvel atlas de la Chine, de la Tartarie chinoise et du Thibet",1737.

noses" (West 1998). Ward (1990) documents the �rst recorded case of altitude sickness comingfrom the Karakorums, with Chinese sources dating to 37-32 BCE describing the route fromYarkand to Afghanistan as passing the aptly named �Great Headache�and �Little Headache�mountains. McKay (2007) also presents the account of the well known Chinese Buddhist travelerFa-Hsien, who took the western route to India in 399-414 CE. He describes a death from whatwas probably high-altitude pulmonary oedema, or swelling of lungs. In contrast, Tibetan sourcesare mainly reticent about their landscape and the di¢ cult journeys undertaken through it;furthermore, accounts of Tibetans su¤ering altitude sickness are extremely uncommon (McKay2007).

The overall number of people impacted by either AMS or CMS is relatively few (Leon-Velarde & et al. (2005) reports that about 5-10 percent of population in highlands are a¤ected),but the number naturally increases past di¤erent stages of high altitudes, and appears to a¤ectChinese much more than Tibetans. The psychological impact of la-drak, or "The Poison of thePass," or the "exhalations of mischievous Gods" (West 1998) would have been greater in theancient times, when the scienti�c cause of altitude illness was not known. The apparent lowincidence of altitude illness in Tibetans cannot be ascribed to simple geographic di¤erences. TheQinghai-Tibetan plateau, where Tibetans live, is bounded by the Himalayas in the southwestand the Kunlun and Aljin mountains in the northeast. On the northern side of the Himalayanwatershed, Tibet, now divided among the Tibetan Autonomous Region of the Peoples�Republicof China and several other Chinese provinces, has an average elevation of 4380 meters. Coveringmore than 2.5 million kilometers-squared, the Qinghai-Tibetan plateau is the highest and largestplateau in the world. Wu (2001) reports that in 1990 about 4,594,188 Tibetans lived on theplateau, with 53 percent living at an altitude over 3500 meters. Furthermore, about 600,000 liveat an altitude exceeding 4500 meters in the Chantong-Qingnan area. Furthermore, Tibetansare lifelong high-altitude residents and cannot easily move to higher or lower elevations8. Themajority of the population group living at the altitudes comparable to the Andeans in SouthAmerica therefore suggests that Tibetan�s ability to adapt to their landscape is unique to thepeople.

Di¤erences in the physical attributes of Chinese and Tibetans �nd most support from thegenetics literature, which suggest genetic di¤erences stemming from long-term biological adap-tation to high-altitude hypoxia. In a review of accumulating data linking various genes toaltitude illness, MacInnis & et al. (2010) reports 58 genes to have been investigated, of which17 have shown some association with the sickness, and that the accumulating data are consistentwith a polygenic condition with a strong environmental component.9 A comparison between Ti-betans and Han Chinese shows that there is a much lower frequency of acute mountain sicknessamong the highlanders owing to genetic adaptation (Wu & et al. 2005), and chronic mountainsickness, a disease that tends to occur later in life among natives or long-life high-altitude resi-dents, is also less common in Tibetans than in similarly exposed Andeans (Moore 1998), witha number of highly selected genes playing a role in maintaining hematocrit levels (the ratio ofred blood cells to the total volume of blood) that may contribute to this apparent resistance(Bigham & et al. 2010). Simonson & et al. (2010) similarly �nds among Tibetans, who havelived at very high altitudes for thousands of years, frequent occurrences of genes whose products

8Wu reports that over 90 percent of the population are engaged in farming and herding; in farming thealtitude limit of crops is around 4500 meters, while the nomads reside above 4800 meters and 5500 meters. Eventhe recent shift of some livelihood into mining has not changed the permanent residence of Tibetans at altitudesbetween 3700 and 6000 meters.

9 In addition, Beall & et al. (1998)�s analysis reveals that the proportion of phenotypic variance in hemoglobinconcentration attributable to genetic factors (i.e., the heritability of hemoglobin concentration) was 86 percentin the Tibetans and 87 percent in the Andeans.

are likely involved in high-altitude adaptation and associated with the decreased hemoglobinphenotype that is unique to this highland population. Tibetan natives, compared to Han Chi-nese or South American high-altitude natives, also have a remarkable lack of muscularization ofpulmonary arteries and low constriction of blood vessels as a response to de�ciency in oxygenreaching the tissues. (Gupta, Rao, Anad, Banerjee & Boparai 1992, Groves & et al. 1993, Beall& et al. 1998).

6 Data

The protest data were mainly gathered from TibetInfoNet, an independent information serviceon contemporary Tibet, with details supplemented by reports from International Campaign forTibet Organization (ICT 2008).10 They were then cross-checked with information gathered fromthe Department of Information and International Relations, Central Tibetan Administration inDharamsala, India (DIIR 2008). In the few cases where the reports di¤ered in the occurrenceof an incident, the observations were dropped.11 Since most accounts recorded in these sourcesgave only county-level location information, observations were aggregated to the county level.For each county, this paper reports whether a protest occurred or not, as well as how manyincidents occurred. According to the classi�cation introduced in ICT (2008), a protest mayinvolve any act of demonstration that involved one or more groups of people including laypeople,monks, nuns, and students. These demonstrations also lead to one or more outcomes includingdetentions, casualties, and fatalities. Violence, if any, could have been perpetrated by securityforces, protestors, or by both sides. This broad categorization of protests was intended tocapture any evidence of unrest in the Tibetan regions.

In order to assess the extent to which a homogenous Tibetan region historically witnessedChinese in�uence, this paper introduces a novel index of town name classi�cations, and arguesthat these names are indicators of the past history. It uses Tibet Township Map and PlaceName Index Database based on the 2000 Chinese census township level administrative units.12

The index includes Tibetan, Chinese and English township names in Tibet Autonomous Region(TAR), as well as Tibet Autonomous Prefectures in the surrounding provinces of Gansu, Qing-hai, Sichuan and Yunnan.13 The main motivation for creating the index was to record traces ofremaining Tibetan place names in areas that have seen dramatic changes in the composition ofpopulation groups, especially since the government-incentivized migration of Han people intotraditionally Tibetan areas. Recording each town�s existing name in various languages entailed�nding information from various sources, including the o¢ cial toponym lists published by the

10For a full list of sources from which the incidents reports are obtained, see tibetinfonet.com. The sourcescome from Tibetan, Chinese as well as international media.11Out of a total of 160 counties in TAR and surrounding counties, 19 counties had con�icting incident reports.

2 counties in Gansu, 4 counties in Qinghai, 6 in Sichuan and 7 in Xizang were consequently dropped from theanalysis. Dropping these counties does not substantially change the main results presented in this paper.12Tibet Township Map and Place Name Index Database was constructed at Princeton University Library Digital

Map and Geospatial Information Center by Tsering W. Shawa.13The year 2000 was the �rst time the Chinese township level census data became available to the public.

The township is a fourth-level of census geography unit. It includes Jiedao (the urban street location which ispart of a larger urban area), Zhen (township with an urban population), and Xiang (township with no urbanpopulation). In addition, the township classi�cation includes virtual townships such as sheep breeding areas,special farming areas, etc. The township names are written in three scripts: Tibetan, Chinese, and English.Pinyin is the romanization standard for Chinese script, and Wylie is a standard transliteration of Tibetan scriptin scholarly literature in the English world, and the romanization is provided for a general reader to sound outTibetan town names. The Wylie transliteration is created from Tibetan script using Universal Tibetan FontConverter. The program was developed by Tashi Tsering at the China Tibetology Research Center in Beijing.

Chinese government in these areas, the Tibet map in Tibetan language published by Tibetan-Government-Exile and by Amnye Machen Institute, and toponym database from the Tibetanand Himalayan Library. In the few case that there were only Chinese toponyms on o¢ cialrecords but the author recognized the existence of Tibetan counterparts for the towns throughhis own heritage and personal interviews with townspeople, these Tibetan names were recordedas well.

For the purpose of the following empirical analysis, this paper introduces another classi�-cation of these toponyms. Each town is categorized as having only Chinese name (C), bothChinese and Tibetan name (CT), and only Tibetan name (T). Since all towns have their o¢ cialnames written in Chinese characters and in Pinyin, the extent to which the original town namehas been preserved is measured on how phonetically or literally it is similar to the Tibetancounterpart (if the Tibetan name does exist). A town in C-category means that it has a nameinstantly recognizable as Chinese with no counterpart in the Tibetan language; that is, thetown name has no literal or phonetic connection to any Tibetan word. This plausibly impliesstrong historical Chinese in�uence on the town, or a recent foundation of the town under thegovernment initiative to relocate the Han majority. In either case, historical Tibetan presenceis essentially non-existent in the area. A town in CT category on the other hand has twodistinct names, one in Chinese and one in Tibetan. Furthermore, the Chinese name is not aphonetic or literal translation of the Tibetan name, such that each name has its own origin.A town in T category has a Chinese name that is a direct phonetic translation of the town�sTibetan name. In a small number of cases, Chinese town names are literal translations of theTibetan counterparts and are also coded as T. As an example, the town of Yeniugou (Donglungin Tibetan) in Qinghai Province is classi�ed as having only Tibetan town name, despite thepronounced di¤erence in the two town names. Both Yeniugou and Donglung in fact mean thesame thing; they are translated as the wild yak country. Because the protest data are obtainedat the county level, the level of historical Chinese in�uence is also aggregated at the countylevel by the area-weighted fraction of the county that is in either C or CT category.14 Thereare 1960 counties in the 141 counties of the �ve provinces included in the dataset. 50 townswere dropped due to ambiguity of name origins.

The following analysis also controls for a number of geographic variables. Each county�secological surroundings are de�ned by a set of fractions of the land occupied by di¤erent biomesfrom ESRI 2008 Data. These biomes are identi�ed as climatically and geographically di¤erentfrom each other. Temperate Broadleaf and Mixed Forests, and Temperate Conifer Forests aregrouped together as Temperate Forest (Mixed Forests include Coniferous forests). Other typesof biomes include Deserts and Xeric Shrublands, Montane Grasslands and Shrublands, Rockand Ice, and Tropical and Subtropical Moist Broadleaf Forests. The set of ecological variables,together with the mean elevation, provide information on the type of plant structures, climateand vegetation occupying each county. They describe the di¤erent habitat types which likelyin�uenced the initial settlement patterns of people in history. The empirical analysis alsoincludes standard location variables, as well as measures for the least cost path from eachcounty centroid to the four historically major trading towns surrounding Tibet: Chengdu,Xining, Kunming and Lanzhou (see Appendix 2 for details on how the path distance wascalculated).

14Since a GIS map illustrating administrative boundaries of Chinese townships was unavailable, the followingapproach was taken to calculate the area of each township. The land area of each town was obtained by usingThiessen Polygon feature in ArcGIS on township centroids. The features creates a border between two adjacenttowns by �rst calculating the map distance between two towns�centroids, and drawing a line through the pointthat is equidistant from each of the centroids.

The paper introduces a set of variables describing historical presence of di¤erent Buddhistschools. The georeferenced data obtained from Tibet Institut give information on the �ve di¤er-ent sects of Tibetan Buddhism: Bon, Gelug, Kagyu, Nyingma and Shakya.15 For each county,the number of sites related to each Buddhist school is recorded. For more recent policy-drivenfactors that potentially in�uenced variation in the level of protests, the paper uses the totalnumber of Buddhist religious sites o¢ cially recognized by the Chinese government as a mea-sure of the central administration�s policy towards religious tolerance (Han & Paik 2011).16 Thedata come from the Atlas of Religions in China, collected by the National Bureau of Statisticsof China and distributed by the University of Michigan China Data Center. Finally, the paperuses district level contemporaneous variables from China�s National Bureau of Statistic�s His-torical China County Population Census Data with GIS Maps for the year 2000. These includethe urbanization rate, total county population and Han population.

7 Findings

Table 1 presents summary statistics, �rst showing the area-weighted fraction of the averagecounty occupied with towns of Chinese origin. On average 14.6 percent of a county experiencedhistorical Chinese in�uence, according to the town names. There are about 12 towns per county,and in total 1910 towns in 141 counties are included in the analysis. Map 1 shows that thecounties north and east of TAR, those that have been integrated as part of China since the Qingdynasty, also experienced changes in their town names. Most of counties in TAR on the otherhand appear to have had little or no Chinese in�uence in their town names in history. Table 1also shows that about one third of all counties included in the analysis experienced one or moreprotests in 2008. It is apparent from Map 2 that these incidents spread widely across the regions,although the majority of incidents took place east of Tibet. Surprisingly only about 30 percentof all incidents were reported as having direct involvement of monks, and even less percentageof the total number of incidents (5%) were reported as having nuns involved. About 7 percentof the incidents resulted in one or more fatalities, and about 5 percent of the reported incidentshad accounts of violence from the protestors. Overall the number of incidents resulting in one ormore casualties is low, with 13 percent of the total number reporting such outcomes. A highernumber of incidents had reports of detentions and protestors being taken away; the data donot provide details on the aftermath of detentions, and thus the actual number of casualties islikely underrepresented. The highest number of incidents is reported in the provinces of Xizang(Tibet) and Sichuan with 14 incidents each, and is closely followed by Qinghai Province thatsaw 13 incidents of protests. On the other hand, 6 out of 7 counties in TAP Gansu provincewitnessed uprisings, in contrast to those in Yunnan which experienced zero reported protest.

15Tibet-Institut, Rikon-Zürich. 1987. Tibet, ethnisch-kulturhistorische Karte : für den Zeitraum von 1280 bis1965, entspr. der Zeit der chinesischen Yüan-, Ming- und Ch�ing-Dynastien 1:5,000,000. Rikon, Schweiz: TibetInstitut.16Han & Paik (2011) argues that the number of registered Buddhist religious sites in each county represents

the extent to which the county experienced the Chinese government�s reversal policy of repressing religiouspractice in 1980s and 1990s. That is, the number of religious sites re�ects two policy-driven outcomes. First,it represents the relative religious tolerance that the county enjoyed in 1980�s. Second, more registered sitesmeans that the level of monitoring that the county in turn experienced was likely higher as the state policytowards religious practice hardened in 1990�s. The authors �nd that this reversal of religious tolerance policytowards certain counties is signi�cantly correlated with the level of protests in 2008. That is, the counties thatexperienced relative freedom but later were punished by repression were more likely to stage protests, than thosethat constantly were repressed. Since many prominent temples are not o¢ cially recognized, especially if theyhave anti-government agendas and hence are persecuted more, the o¢ cial count does not re�ect the true extentto which certain counties have more Buddhist monasteries than others.

94 out of 141 counties experienced zero incident, and these counties are spread throughout all�ve provinces. The wide spread of protests is evident in the statistics, which also suggest thatthere are signi�cant within-province variations in terms of incident occurrence.

The average county elevation is 4120 meters, indicating that the counties in the analysis aretypically located in the highlands. Furthermore, 38 percent of all the counties in the analysisare located at altitudes higher than 4500 meters. Figures 1 and 2 show in detail the locationspread of townships. In Figure 1, townships with Chinese names have a unimodal distributionthat peaks around 2700 meters. Among these towns, those with both Tibetan and Chinesenames tend to be located in higher altitudes, with the peak of the distribution at around 3800meters compared to 2799 meters for those that only have Chinese names. No town with aChinese name is located above 4600 meters, as shown clearly in Figure 2. While most townswith Chinese heritage are located in the lowlands at around 2700 meters in altitude, towns withTibetan heritage are generally located much higher up in altitude. In fact the highest numberof Tibetan towns is found at 4500 meters, and the distribution of Tibetan towns shows a sharpdecline only past 5200 meters. The spread of these town locations suggests that there was adecreasing level of Chinese presence in Tibetan regions as the elevation increased, and lendssupport to the claim that the 4500 meter-threshold based on genetic di¤erences may be usefulin identifying the exogenous variation in town names.

In Table 2, the dependent variable is a binary indicator for whether a county witnessed aprotest or not. Columns 1-6 measure the average e¤ect of the historical presence measure on theprobability that a county reported at least one protest incident in 2008.17 Under both probitand province �xed-e¤ect OLS speci�cations, the historical presence variable gives statisticallysigni�cant and negative coe¢ cient values. The result in Column 3 for example suggests that a 10percent increase in the county area with historical Chinese in�uence will reduce the likelihoodof experiencing any protest in 2008 by 11 percent. The signi�cance of the heritage variableremains robust to initial controls such as proximity and ecology (Column 1), cost of travel andbene�t of trade (Column 2), as well as both historic cultural factors and recent policy changes(Column 3).

Tibetan counties with Chinese heritage may simply re�ect the fact that these areas tendto be closer to the borders along China, and that they also likely have similar biogeographicalendowments. In such cases, these areas would have naturally had more interaction with theChinese. Residents in highlands also may have di¤erent attitudes towards migrants than thosein the lowlands. For example Scott (2009) suggests that those living in the highlands may bemore hostile towards those living in the valleys and lowlands, since they made a conscious choiceto avoid integration with the central state by moving to mountain slopes. Rough terrains havealso been found to favor militias and insurgent groups in conducting guerilla attacks (Fearon& Laitin 2003). The variables in Column 1 are intended to control for these various geographice¤ects. However as the regression results show, the e¤ect of historical in�uence is robust tomatching towns by location and biogeography. While control variables such as the mean countylongitude and the fraction of land covered by Temperate Forest give signi�cant coe¢ cient valuesunder certain speci�cations, there is no consistent evidence suggesting that temperate regions inthe eastern part of Tibet experienced less protests in 2008. Another plausible argument is thatthe towns along the ancient route connecting Tibet and China in the famous tea-horse trade,and the counties to which these towns belonged to, were the ones that enjoyed amicable relations

17Under the probit speci�cation, two of the biogeography control variables, fraction of county occupied byDesert and Tropical Forest were dropped because a subsample of the data predicted failure perfectly. The twovariables were consequently dropped from the regression results (Davidson & MacKinnon 1993, Pg. 73).

between the two groups in the long run.18. Tibetans traded their horses for tea with the Chinese,creating a network of traders spanning the Tibetan Plateau since the 11th century, as well as atrade route called the Chamagudao ("tea-horse caravan route") that were used until the 1950s.Chengdu for example has been one of the major trading towns, located along the passagewayfrom Sichuan Province�s tea-growing area. The tea-horse trade route connected Chengdu withthe gateway town of Qamdo, on the pathway to Lhasa. Chengdu lies on the southern routethat included other bustling trade towns including Yaan and Kanding. To the south in theprovince of Yunnan, Kunming has been the provincial capital that connected the traders fromYunnan�s tea-growing areas to Tibet through the routes along Hengduan Mountains. Column 2introduces a set of least cost paths from each of the county centroids to the four major tradingcenters; they are intended to control for the e¤ect of historical Chinese in�uence due to certaincounties bene�tting more from trading. The historical in�uence e¤ect however remains robustafter controlling for the trade distance variables.

The historical Chinese in�uence e¤ect may not hold signi�cant after addressing the extentto which di¤erent Buddhist schools historically dominated certain regions and fostered Tibetanidentity. Within Tibetan Buddhism, the Dalai Lama�s Gelug sect has traditionally had thebiggest following, thereby wielding signi�cant political domination over other sects includingKagyu, Nyingma, and Shayka, as well as the non-Buddhist Bon sect. The three major Gelugmonasteries around Lhasa �Drepung, Sera, and Ganden �exerted considerable in�uence onTibet�s political decision making before the 1950s. In contrast, the Bon religion historicallyendured waves of persecution by the dominant Buddhist religion in Tibet, and Bon templestended to be built in areas where Buddhism had less in�uence (Kværne 1995). The pointestimates in Column 3, which controls for the historical existence of various Buddhist schoolsin counties, suggest that the regions with historically dominant Gelug presence have a higherlikelihood of witnessing a protest, while the ones with Bon presence face the opposite e¤ect.These results are however not signi�cant, while the historical in�uence variable remains robustto the inclusion of these controls.

In view of the central administration�s policy towards Tibet since the 1950s, Han & Paik(2011) argues that the level of religious tolerance played an important role in the politicalmobilization outcome in 2008. In the paper, the number of o¢ cially registered Buddhist reli-gious sites in each county represents the extent to which the county experienced the Chinesegovernment�s reversal policy of repressing religious practice in the 1980s and the 1990s. First,the level of registration in each county re�ects religious tolerance that the county enjoyed inthe 1980�s. Second, more registered sites meant that the level of monitoring that the countyin turn experienced was likely higher as the state policy towards religious practice hardenedin the 1990�s. The paper �nds that this reversal of religious tolerance policy towards certaincounties is signi�cantly correlated with the level of protests in 2008. That is, the counties thatexperienced relative freedom but later were punished by repression were more likely to stageprotests than those constantly repressed. Adding these control variable again does not diminishthe statistical signi�cance of the coe¢ cient on the main variable of interest.

Columns 7 to 9 present IV estimates using the 4500 meter threshold; they estimate the

18For example, Yushu township (Kyegudo in Tibetan) in Yushu Prefecture, Qinghai province, traces its nameorigin to the area�s relation to its geographic importance as a major trading mart,. It has been a major townlinking Kham, Amdo, and Central Tibet. It forms a transitional zone between the two, a centuries-old nucleusof trade. The region�s rich grasslands made it an ideal yak, sheep and horse hearding area, and the town bustledwith traders bringing tea from Western China into Tibet. The traders in return took back local wool and hides tomarkets as far away as the North China coast. The region�s current "development with Chinese characteristics"builds on a long legacy of trading and contact within and beyond the Tibetan cultural sphere (Marshall &Cooke 1997).

e¤ect of historical Chinese presence in certain areas due to their lowland environment below4500 meters. The e¤ect is negative and statistically signi�cant under all speci�cations. Fur-thermore, the magnitude of the e¤ect increases relative to the OLS and probit speci�cations.The coe¢ cient value suggests that a 10 percent increase in the area of a county with historicalChinese presence leads to a 26-32 percent decrease in the likelihood of experiencing at least oneprotest. Table 3 uses the same speci�cations to the number of protests. Columns 1 to 3 showthat the historical in�uence e¤ect remains strong and reduces the number of incidents by 60-90percent of what it might have been otherwise. Both the OLS and IV estimation give negativecoe¢ cient values.

Tables 4a and 4b use the same controls as in the previous tables to assess the e¤ect ofhistorical Chinese in�uence on contemporaneous variables. First in Table 4a, the urbanizationrate in 2000 is used as a measure of the level of economic development that regions experiencedin the present. IV estimation results under Columns 5 and 6 suggest that the counties withhistorical Chinese presence in fact tend to be less urbanized. Table 4b shows that countieswith historical Chinese in�uence may have greater Han population, but this result loses itssigni�cance under IV estimation. As a case study of this result, Appendix 1 presents a listof examples from Tibet Township Map and Place Name Index Database. The 22 towns inKangding County, Sichuan Province, show both Chinese and Tibetan origins. Geographicallythe region has had close contact with the Chinese from the east as it lies close to the easternedge of the region of Kham, which has been a gateway between Tibet and China. Each town�sethnic name origin however is not representative of the current level of Tibetan population.For example, Liuba (Lugpa in Tibetan) is a town with total population of 2660 in the year2000. 98 percent of the population is Tibetan, and the Chinese name is a phonetic translationof the Tibetan name. The town name is hence classi�ed as having only Tibetan originatedname (T). Pusharong (Phouharrong in Tibetan), however is a town with a Chinese name thathas no recognizable Tibetan counterpart, despite its dominant Tibetan population (97%). Thetown of Jiju (Kyikyi) with 99 percent Tibetan population provides an example of a townthat continues to have both Tibetan and Chinese names distinct from each other, despite thedominant presence of Tibetan population. These town names therefore appear to indicatepurely historical, not present, Chinese heritage. There are few exception to this pattern; for asmall number of towns that have only recently been established, their name origins do re�ectsome current population trend. One example as discussed above is the town of Bayi in NyingchiCounty, Tibet. Given the recent development of the city and its increase in size, the currentpopulation is representative of a town with both Chinese and Tibetan names. In 2000, abouthalf, or 43 percent of total county population was reported to be Tibetan, while 53 percent ofpopulation was Han Chinese.

Table 5 shows a set of results including both the urbanization and Han population indicatorsin 2000 on the right hand side. The urbanization rate is a measure for income levels, and animportant determinant of civil con�ict (Fearon & Laitin 2003, Miguel et al. 2004). Greater Hanpopulation in a county means a lower concentration of Tibetan population, with a decreasedcapacity of the group to organize political protests (Toft 2002, Toft 2003, Horowitz 1985, Posen1993, Fearon 1998). While the paper recognizes these contemporaneous variables as probablyendogenous to the main historical variable, the �ndings suggest that even controlling for currentsocioeconomic variables, the historical variable appears to have a negative correlation with theprobability of experiencing one or more protests. The �ndings also suggest that the currentpresence of Han population is negatively correlated with the probability of uprising, whileurbanization is positively correlated with the dependent variable.

8 Discussion

After Chinese migration into a region, the integration process determined the level of accep-tance of the Chinese state dominance in the region. But how was the initial Chinese in�uencesustained in the long run to a¤ect political mobilization? One way to explain the Chinese legacyin Tibet, and its impact on the present-day protests, is by specifying persistent institutionalmechanisms. For example, complementary and lasting institutions may have emerged to sup-port Tibetan-Chinese trade. As Jha (2009) explains in his study of medieval trade institutionsin India, trade institutions such as merchant guilds may have emerged, allowing both groups togain by trading complementary goods and promoting leniency towards each other. This theoryof trade institutions in preventing ethnic violence, by allowing the gains from inter-ethnic tradeto be shared between groups, certainly applies to many towns in Tibet. These towns werelikely to be located along China and Tibet�s historic tea-horse trade routes, which existed formillennia up until 1950s when highway roads �nally opened Tibet to the rest of the world.

As town name examples above show, however, there were also towns founded because theywere the sites of military garrisons during the Qing dynasty, with the emperor�s intentionto monitor Tibetan activities. Such places would have been understandably less likely to becenters of amicable relations.19 Even in these cases, however, the outcome of historical Chinesein�uence appears to have been that of decreased political mobilization by the Tibetan residents.The empirical �ndings above furthermore show that proximities to these trade centers do notchange the signi�cance of the heritage e¤ect on political mobilization.

Another institutional explanation is that since the concentration factor is linked to morecollective action and political mobilization (Fearon 1998, Weidmann 2009, Toft 2002, Toft 2003),historically diminished group concentration somehow may have had a lasting impact on thesolidarity of Tibetan descendants in these regions. This argument does not simply imply thatonce the initial Tibetan concentration was diminished, it always remained small, such that thecurrent political mobilization was an outcome of the concentration factor. In fact, based onthe town name index and the 2000 population census data, numerous towns as shown abovehave strong Chinese heritage but predominant Tibetan population. Instead, the sustainedleniency towards the Chinese state in certain regions can be attributed to their lack of TibetanBuddhist monasteries in history. That is, the Chinese heritage may have in�uenced the extentto which Tibetan Buddhism dominated these regions and fostered Tibetan identity. Buddhismhas been the key de�ning feature of Tibetan identity, such that the traditional theocraticnature of the Tibetan political system20 meant that �Buddhist ideology and values dominatedthe population�s world view and the state�s raison d�être� (Goldstein 2007, Pg. 23). As suchareas with traditionally predominant Tibetan population also established monasteries, theseareas then became centers of both Buddhist and political activities. Furthermore, Tibetanmonasteries were the largest landowners with concentrated manorial labors. This politicalsystem, from the point of view of the CCP when they �rst came into Tibet in 1950, was themain obstacle for socialist reforms, and one of the main causes for subsequent con�icts. Forthose in a Tibetan region, more monasteries in a region meant more resources and organizationthat could be drawn upon. Political mobilization in a region has thus mainly been carried outby local monks, who make up a substantial proportion of the entire population. The regionswith higher Chinese concentration on the other hand would likely have had less exposure to

19There are of course exceptions; see the example of Dartsedo below for example, a town that was both a tradecenter and a military garrison.20The Dalai Lama, regarded as the reincarnation of the Bodhisattva Avalokiteshvara, has been both Tibet�s

religious and political leader.

both the cultural and political in�uences of the religious institutions. This paper argues thatthe recent political mobilization may stem from those areas with historically strong Buddhistin�uence, which also have been areas with less Chinese presence.

Buddhism in Tibet had the distinct feature of mass monasticism that encouraged people,particularly males, to join monasteries; this meant that Tibetan families had close familialties to the monasteries. Before 1951, there were about twenty-�ve hundred monasteries inTibet. Surveys showed that there were 97528 monks in Central Tibet and Kham in 1694, and319270 in 1733, about 13 percent of the total population (Goldstein & Rimpoche 1989, Pg.21).An estimate in 1930 reported that between 10 to 20 percent of Tibetan males were monks(Goldstein 2007, Pg. 13). Monasticism in Tibet, therefore, was "not the otherworldly domainof a minute elite but a mass phenomenon" (Goldstein & Rimpoche 1989, Pg.21). Traditionallythe overwhelming majority of monks were placed in monasteries by their parents; becominga monk was lifelong commitment and the monks helped support their families by sharing themoney distribution that monks received throughout the year. In other cases, recruitment wasthe result of a corvee tax obligation: monastic serfs with three sons often had to make one amonk. Monks faced economic problems in the case that they chose to leave their monasteries;they lost whatever rights they might otherwise have had in their family farm when they enteredthe monastery, and also reverted to their original serf status and were liable to for service totheir lord (Goldstein & Rimpoche 1989, Pg.23). According to Goldstein & Rimpoche (1989),while prominent monasteries, such as Dera, Drepung and Ganden monasteries in Lhasa (calleddensa sum, or the "three seats" ) had networks of a¢ liated monasteries throughout the country,there was no abbot for the whole monastery; the overall entity, the monastery, was in realitya combination of semi-autonomous subunits known as tratsang, or colleges. Each monk fromabroad had to enroll in a speci�c residential subunit known as khamtsen, determined by hisregion of origin; khamtsen therefore exhibited considerable "internal linguistic and culturalhomogeneity"- the membership was automatic and mutually exclusive. This tendency for highlylocalized devotion of monks to their regions implied that there was little collaboration acrossthe colleges for establishment of monasteries in Chinese-dominant areas.

In addition to lack of initiatives for monks to spread their Tibetan values to regions out-side their local areas, laypeople had little individual incentives to invest in a relationship withmonasteries outside of the familial ties. New monasteries were expensive to build, and peo-ple already had invested relations with the monasteries in which their families were monks.Within the tightly knit local network of communities and monasteries, the teachings of monksemphasized keeping Tibetan culture and identity, which in turn fostered Tibetan nationalismas the people�s sovereignty came under threat (Kolas 1996). Importantly, the spread of thesevalues was sustained over generations. Furthermore, until 1960 the majority of the country�sland and people were organized into manorial estates, and most Tibetans were aristocrats andserfs bound to the lands. There was essentially no movement of people across the lands, andTibetans�loyalty remained in the monasteries in their homelands. The consequent repression ofTibetan monasteries after 1960 by the Chinese government meant that any spread of Buddhistbeliefs was restricted even further.

Because there are no available data detailing historical records of monastery establishmentthroughout Tibet, it is di¢ cult to verify whether regions dominated by Chinese in�uence in factsustained a smaller number of monasteries relative to other regions. However, the argumentspresented here suggest that under the institutional mechanism, it would have been di¢ cultto establish new monasteries and sustain old ones in the Chinese-in�uenced regions. Withthe manorial system restricting migration, Tibetans�familial ties to local monasteries and thesegregated structure of monks based on the geographic origin, the network of monasteries

promoting Tibetan solidarity likely did not proliferate in the Chinese regions even as Buddhismcontinued to de�ne the political discourse of Tibetan nationalism elsewhere.

9 Conclusion

This paper aims to show that historical in�uence of ethnic integration has long term impacton the current levels of political mobilization. It has presented empirical evidence from Tibet�srecent wave of protests to support this hypothesis, and an historical institutionalism argumentto describe how leniency towards the Chinese was sustained. The long term implication inTibet�s case is that historical Chinese in�uence in a region leads to the same area experiencingless mobilization against the central state today. The lack of Buddhist monasteries in the regionthroughout history may have decreased political mobilization, by diminishing the role that thereligious institutions played in spreading Tibetan culture and identity.

In view of the Chinese in�uence on the 56 ethnic groups living in the country, the sinicizationof Tibet stands out as a unique case. The region�s formidable terrains have essentially blockedmuch of the Chinese settlement from the east since the ancient times and con�ned inter-ethnicrelations mostly to the region�s geographically de�ned borders. The large plateau environmenthowever distinguished the region from other ethnic groups living in high altitudes. Tibetanswere not con�ned to small living areas along steep mountain slopes, and yet still separated fromthe outsiders by natural barriers. The sinicization process however has accelerated beyond anyhistorical precedent since 1949, and the region witnessed rapid industrialization through theconstruction of paved highways and the completion of the Qinghai-Tibetan railway. The Han-Tibetan integration, led by economic development in urban centers attracting Han migration,re�ects the central government�s future direction of ethnic policy in Tibetan areas. Certainplaces, such as Bayi, have been speci�cally designated as new development centers by theBeijing administration, and have seen a rapid growth of Han population. These places tend tobe located in the lowlands and are now reachable with little travelling concerns.

In response to the changing times, Tibetans�resistance towards the Chinese state has man-ifested in grievance and fear over Han Chinese in-migration and settlement in ethnic Tibetanareas. In a �ve-point proposal made in 1987 during his address at the US Congress, the DalaiLama speci�cally pointed out the demand that China should abandon its population trans-fer policy (Shakya 1999, Pg. 415). In 2008, the Dalai Lama once again claimed that Beijingwas planning mass settlement of Han Chinese and Hui Muslims in Tibet to dilute Tibetanculture and identity (Borger 2008). This paper provides evidence to support the long termconsequences of such integration policy. Historically, China-based empires had the pattern ofsettlement of Han Chinese population in peripheral regions to solidify control (Pan 1992). Dur-ing late Qing period, for example, the empire faced growing threats from the Russian Empire,and encouraged the settlement of Han Chinese in Mongolia and Manchuria to boost its controlof these peripheral regions. The continuing settlement of Han Chinese into Inner Mongolia, forexample, led to marginalization of Mongols in their ancestral land today and may explain therelatively paci�ed nature of ethnic Mongol activism (Bulag 2004). Manchuria has become threenortheastern provinces with overwhelmingly Han Chinese population, such that the Manchushave both lost their native language and been completely assimilated. Recent increases of eth-nic con�icts in both Tibet and Xinjiang Province are mainly attributed to the administration�sGreat Development of the West settlement initiative encouraging the Han migration. Whilethe spike in ethnic con�ict may come as unavoidable, natural outcomes in these regions andthus question the e¤ectiveness of economic development as a means to pacifying the West, thesettlement policy may yet be a successful one if one considers its eventual goal to be assimilation

and subsequent integration.

Appendix 2: Least Cost Path ConstructionThe least cost path from each county centroid to one of the trading centers outside TAR is

constructed using ArcGIS Spatial Analyst Tools�Path Distance and Least Cost Path features.The Path Distance �rst calculates, for each potential route from a centroid to a trading center,the total cost of travelling based on a number of factors. This paper includes three factors,including surface distance, vertical factor, and cost distance. The total cost is in the samemeasurement unit as the map distance. That is, the total cost is simply the multiple of themap distance by factors making travel more di¢ cult. In order to understand how the total costof a path travelled is calculated, imagine drawing a 9-by-9 grid, with 9 identically sized squarecells and with the center cell containing the county centroid in its center. The Path Distancefeature calculates, for each distance travelled from the center cell to one of the 8 adjacent cells,the cost of travelling. This cost of travelling a distance equivalent to 1 cell-width (1 kilometerin the map used in this paper) is calculated by the following function:

TotalCost = CostDistance � SurfaceDistance � V erticalFactor

where Cost Distance is equal to 1 if the adjacent cell is ground, and 2 if it contains a majorriver or a lake 21. Surface Distance is the actual ground distance (as opposed to map distance)which takes into account the terrain slope of each cell. It is the distance that must be travelledwhen moving from one cell to another. Surface Distance is calculated by the Pythagoreantheorem, where the base distance is the cell size, and the height is the di¤erence between theelevation of the two cells. Vertical Factor further controls for the di¢ culty in ascending anddescending steep slopes. One can imagine that it is much more di¢ cult to cross a 1-km surfacedistance of a 60-degree mountain slope, instead of the same surface distance of a 10-degreeslope. Similarly, it may be equally di¢ cult to descend a mountain of slippery minus 60-degreemountain slope instead of a minus 10-degree slope. In order to take such factors into account,Vertical Factor allows for di¤erential impact of di¢ culty in travelling based on the slope of eachterrain for each route. While there are many alternatives, this paper uses the Vertical Factordetermined by the secant function of the slope, to re�ect the fact that it becomes in�nitelydi¢ cult to cross a 90-degree or minus 90-degree slope, and the closer the slope is to zero, theeasier it is to travel across.22

In the second step of the least cost path construction, ArcGIS uses the outputs from PathDistance feature as part of inputs in Cost Distance feature, which then picks the path with theleast cost of travel between each county centroid and the trade center destination. Appendix2 Map shows the �nal result for all county centroids and their least cost paths to Xining, oneof the trade centers located in Qinghai Province. Note that each connected line between acounty centroid and Xining is not straight; the rugged lines are meant to re�ect the di¢ cultiesin crossing rivers and rough terrains especially pertinent in Tibet, and better estimate theactual path through which people in each town must have gone through to reach major towns.On average the map distance between a county centroid and Xining is 802 kilometers with astandard deviation of 521 kilometers, while the average total cost distance is 837 kilometerswith a standard deviation of 546 kilometers.21This arbitrary assignment of cost distance values is simply meant to distinguish the level of di¢ culty in

travelling across a river or a lake, as opposed to travelling on ground. Increasing the cost distance value for cellsthat contain water to a greater value does not change the �nal empirical results.22Again, the secant function is only one of the alternatives o¤ered in the feature; changing the option to other

alternatives does not substantially change the main empirical result.

References

Acemoglu, D., Johnson, S. & Robinson, J. (2001). The Colonial Origins of Comparative Devel-opment: An Empirical Investigation, The American Economic Review 91(5)(1369-1401).

Acemoglu, D., Johnson, S. & Robinson, J. (2002). Reversal of fortune: Geography and in-stitutions in the making of the modern world income distribution, Quarterly Journal ofEconomics 117: 1231�1294.

Acemoglu, D., Johnson, S. & Robinson, J. (2005). The Rise of Europe: Atlantic Trade, Institu-tional Change, and Economic Growth, The American Economic Review 93(3)(546-579).

Barnett, R. (2009). The Tibet protests of spring, 2008: con�ict between the nation and thestate, China Perspectives 3: 6�24.

Beall, C. & et al. (1998). Hemoglobin concentration of high-altitude Tibetans and BolivianAymara., Am J Physiol Anthropol. 106: 385�400.

Bigham, A. & et al. (2010). Identifying signatures of natural selection in Tibetan and Andeanpopulations using dense genome scan data, PLoS Genet. 6(9).

Borger, J. (2008). Tibet could be �swamped�by mass chinese settlement after Olympics, saysDalai Lama, The Guardian .

Bulag, U. E. (2004). Inner Mongolia: The dialectics of colonialization and ethnicity building,in M. Rossabi (ed.), Governing China�s Multiethnic Frontiers, University of WashingtonPress, Seattle.

Burenhult, N. & Levinson, S. C. (2008). Language and landscape: a cross-linguistic perspective,Language Sciences 30(2): 135�150.

Davidson, R. & MacKinnon, J. (1993). Estimation and Inference in Econometrics, OxfordUniversity Press, New York.

DIIR (2008). 2008 Uprising in Tibet: chronology and analysis, UN, EU and Human Righst Desk,Dept of Information and International Relations (DIIR), Central Tibetan Administration(CTA), Dharamsala, HP, India.

Fearon, J. D. (1998). Commitment problems and teh spread of ethnic con�ict, in D. Lake &D. Rothchild (eds), The International Spread of Ethnic Con�ict: Fear, Di¤usion, Escala-tion, Princeton University Press, Princeton.

Fearon, J. & Laitin, D. (2003). Ethnicity, Insurgency, and Civil War, American Political ScienceReview 97(1): 75�90.

Goldstein, M. C. (1997). The Snow Lion and the Dragon : China, Tibet, and the Dalai Lama,University of California Press, Berkeley, CA.

Goldstein, M. C. (2007). A History of Modern Tibet. Volume 2: The Calm before the Storm,1951-1955, University of California Press, Berkeley, CA.

Goldstein, M. C., Jiao, B. & Tanzen, L. (2009). On the Cultural Revolution in Tibet : theNyemo Incident of 1969, University of California Press, Berkeley, CA.

Goldstein, M. C. & Rimpoche, G. (1989). A History of Modern Tibet, 1913-1951: The Demiseof the Lamaist State, University of California Press, Berkeley, CA.

Groves, B. & et al. (1993). Minimal hypoxic pulmonary hypertension in normal Tibetans at3,658 m, J Appl Physiol. 74: 312�318.

Gupta, M., Rao, K., Anad, I., Banerjee, A. & Boparai, M. (1992). Lack of smooth muscle in thesmall pulmonary arteries of the native Ladakhi: is the Himalayan highlander adapted?,Am Rev Respir Dis. 145: 1201�1204.

Han, E. & Paik, C. (2011). Reversal toward repression and changing dynamics of ethnic de-mography: Evidence from Tibet. Working Paper.

Herman, R. (1999). The Aloha state: place names and the anti-conquest of Hawai�i, Annals ofthe association of American geographers 89(1): 76�102.

Horowitz, D. L. (1985). Ethnic Groups in Con�ict, second edn, University of California Press,Berkeley.

Horsman, S. (2006). The politics of toponyms in the Pamir Mountains, Area 38(3): 279�291.

ICT (2008). Tibet at a turning point: the spring uprising and China�s new crackdown, TheInternational Campaign for Tibet, Washington, DC.

Jha, S. (2009). Trade, institutions and religious tolerance: Evidence from india, StanfordUniversity Graduate School of Business Research Paper .

Kolas, A. (1996). Tibetan nationalism: The politics of religion, Journal of Peace Research1: 51�66.

Kværne, P. (1995). The Bon Religion of Tibet: The Iconography of A Living Tradition, Serindia,London.

Leon-Velarde, F. & et al. (2005). Consensus statement on chronic and subacute high altitudediseases, High Altitude Medicine and Biology 6(2): 147�158.

Lin, H.-t. (2006). Tibet and Nationalist China�s Frontier: Intrigues and Ethnopolitics, 1928-49,University of British Columbia Press, Vancouver, BC.

MacInnis, M. J. & et al. (2010). Evidence for a genetic basis for altitude illness: 2010 update,High Altitude Medicine and Biology 11(4): 349�369.

Marshall, S. D. & Cooke, S. T. (1997). Tibet outside the TAR: Control, exploitation adnassimilation development with Chinese characteristics, The Alliance for Research in Tibet,Washington, DC.

McKay (2007). Fit for the frontier: European understandings of the Tibetan environment inthe colonial era, New Zealand Journal of Asian Studies 9(1): 118�32.

Michalopoulos, S. (2009). The origins of ethnoloinguistic diversity. Working Paper.

Miguel, E., Satyanath, S. & Sergenti, E. (2004). Economic shocks and civil con�ict: an instru-mental variables approach, Journal of Political Economy 112(4): 725�754.

Montalvo, J. & Reynal-Querol, M. (2005). Ethnic Polarization, Potential Con�ict and CivilWar, American Economic Review 95(3)(796-816).

Moore, L. (1998). Human adaptation to high altitude: regional and life-cycle perspectives, Am.J. Phys. Anthropol. 27: 25�64.

Murdoch, D. R., Pollard, A. J. & Gibbs, J. S. R. (2006). Altitude and expedition medicine, inJ. N. Zuckerman (ed.), Principles and Practice of Travel Medicine, John Wiley and SonsLtd., chapter 15.

Pan, Y. (1992). Early Chinese settlement policies towards the nomads, Asia Major 5(2): 41�75.

Posen, B. R. (1993). The Security Dilemma and Ethnic Con�ict, Survival 35(1): 27�47.

Qi, Y. & et al. (2009). Genetic interaction of Hsp70 familiy genes polymorphisms with high-altitude pulmonary edema among Chinese railway constructors at altitudes exceeding 4000meters, Clinica Chimica Acta 405: 17�22.

Radding, L. & Western, J. (2010). What�s in a name? Linguistics, geography, and toponyms,The Geographical Review 3(100): 394�412.

Scott, J. C. (2009). The art of not being governed, Yale University Press, New Haven, CT.

Shakya, T. (1999). The Dragon in the Land of Snows: A History of Modern Tibet since 1947,Pimlico, London.

Simonson, T. S. & et al. (2010). Genetic evidence for high-altitude adaptation in Tibet, Science2.

Spolaore, E. & Wacziarg, R. (2009). The Di¤usion of Development, The Quarterly Journal ofEconomics pp. 469�529.

Toft, M. D. (2002). Indivisible territory, geographic concentration, and ethnic war, SecurityStudies 12(2): 82�119.

Toft, M. D. (2003). The Geography of Ethnic Violence: Identity, Interests, and the Indivisibilityof Territory, Princeton University Press, Princeton, NJ.

Topgyal, T. (2011). Insecurity dilemma and the Tibetan uprising in 2008, Journal of Contem-porary China 20(183-203): 183�203.

Tyagi, P. & Malik, H. (2008). Suggested operating schedules for suddent induction of unacclima-tized or partially acclimatized aircrew for air operations at high altitude, Ind J AerospaceMed 52(2).

Ward, M. (1990). Mountain medicine and physiology: a short history, The Alpine Journal95(1990/91): 191�92.

Weidmann, N. B. (2009). Geography as motivation and opportunity: Group concentration andethnic con�ict, Journal of Con�ict Resolution 4(526-43): 526�43.

West, J. (1998). A history of high altitude physiology and medicine, Oxford University Press.

West, J. (2006). Completion of the rairoad to Lhasa, High Alt. Med. Biol. 7: 263�264.

Wu, T. (2001). The Qinghai-Tibetan plateau: How high do tibetans live?, High AltitudeMedicine and Biology 2(4): 489�499.

Wu, T. Y. & et al. (2009). Reduced incidence and severity of acute mountain sickness in Qinghai-Tibet railroad construction workers after repeated 7-month exposures despite 5-month lowaltitude periods, High Altitude Medicine and Biology 10(3): 221�234.

Wu, T. & et al. (2005). Tibetans at extreme altitude, Wilderness Environ. Med. 16: 47�54.

Yi, P. & et al. (2011). Genetic variations in Tibetan populations and high-altitude adaptationat the Himalayas, Mol Biol Evol. 28(2): 1075�81.

Appendix 1: Tibet Township Place Name Index

Town Names in Kanding County, Sichuan Province

Tibetan Chinese Pinyin Wylie

Tibetan

Phonetic

Tibetan

Population

Percent

Tibetan

Language

Category

ལའོ་ཁྲིན་ 炉城镇 Lucheng la'okhrin Laokrin 12431 35.04 C

འགུ་ཐང་ 姑咱镇 Guzan guthang Guthang 4545 40.16 T

ར་རྔ་མཁར་ 新都桥镇 Xinduqiao ra rnga mkhar Rangakhar 4905 81.59 C

ཡེས་ལྲིན་ 榆林乡 Yulin yeslin Yeslin 1329 65.63 C

གཡག་རྭ་ 雅拉乡 Yala g.yagrwa Yagra 669 23.33 T

ཧྲི་ཅྲིས་ 时济乡 Shiji hricis Hrichis 1817 74.28 C

མག་འཐེན་ 前溪乡 Qianxi mag'then Magthen 1514 80.15 T

ཧྲིག་ལན་ 舍联乡 Shelian hriglan Higlan 1652 84.85 C

སྨད་འབུང་ 麦崩乡 Maibeng smad'bung Madbung 2177 86.60 T

སན་ཧྲིག་ 三合乡 Sanhe sanhig Sanhig 2308 75.42 C

སྲིད་ཐང་ 金汤乡 Jintang skyidthang Kyithang 1808 54.39 T

ཕག་ཐག་ 捧塔乡 Pengta phagthag Phagthag 2011 77.26 T

ས་བདེ་ 沙德乡 Shade sabde Sadhe 2917 95.92 T

ལུག་པ་ 六巴乡 Liuba lugpa Lugpa 2617 98.38 T

ཕོའུ་ཧ་རོང་ 普沙绒乡 Pusharong pho'uhrarong Phouharrong 2053 96.70 C

སྲིད་སྲིད་ 吉居乡 Jiju skyidskyid Kyikyi 2232 99.47 TC

ཝ་ཚེ་ 瓦泽乡 Waze watshe Watse 3835 97.31 T

འགག་པ་ 呷巴乡 Xiaba gagpa Gagpa 3373 94.85 T

ལྕགས་གད་ 甲根坝乡 Jiagenba lcagsgad Chagad 2263 98.95 T

བོན་པོ་གཤྲིས་ 朋布西乡 Pengbuxi bonpogshis Bonposhis 2646 99.74 T

ལྷ་སྒང་ 塔公乡 Tagong lhasgang Lhagang 7349 95.69 T

ཁོབ་ཡུལ་ 孔玉乡 Kongyu khobyul Khobyul 2980 80.98 T

%

#

##

#

#

#

#

###

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

##

###

##

#

#

#

##

###

# #

##

#

##

##

# #

##

### ##

##

#

#

#

#

#

#

#

## #

#

##

##

##

#

#

#

#

#

#

#

#

#

#

#

##

#

#

#

#

##

#

#

#

#

#

#

#

#

#

#

#

#

###

#

##

#

#

####

#

###

##

#

##

#

#

#

#

#

#

#

#

#

#

## #

#

#

#

# #

#

#

#

#

#

#

#

#

#

#

#

#

4

Appendix 2: Least Cost Path to Xining

Elevation (in meters)High : 8752

Low : 1

% Xining# County Centroid

Water

Figure 1

Figure 2

Map 1: Historical Chinese Presence in Tibet Autonomous Region & Prefectures

Area-Weighted Fraction of County with Chinese Town Names

0.000.01 - 0.050.06 - 0.120.13 - 0.240.25 - 0.420.43 - 0.610.62 - 1.00

Map 2: Reported Protest Incidents in Tibet Autonomous Region and Prefectures, 2008

Number of Reported Incidents

012 - 3

4 - 22Incident(s) ReportedNo Incident ReportedConflicting Reports

Table 1: Summary Statistics

Historical Presence Variables

Obs Mean Std. Dev

Area-weighted % of Towns with Chinese

Names in County 141 0.146 0.264

# of Towns with Chinese Names 352

# of Towns with Chinese & Tibetan Names 86

# of Towns with Tibetan Names 1472

Total # of Towns in Data 1910

Geographic Variables

Obs Mean Std. Dev

County Mean Latitude 141 31.995 2.906

County Mean Longitude 141 95.837 6.162

County Mean Elev. Km 141 4.120 0.750

Altitude Illness Threshhold (4500m) 141 0.380 0.470

County Area in Mm-Sq 141 0.014 0.024

# Towns in County 141 11.894 5.655

Ice as Frac. of County Area 141 0.015 0.039

Shrubs as Frac. of County Area 141 0.780 0.280

Temperate Forest as Frac. of County Area 141 0.185 0.274

Tropical Forest as Frac. of County Area 141 0.001 0.006

Desert as Frac. Of County Area 141 0.010 0.090

Least Cost Path to Chengdu in 1000 Km's 141 0.096 2.465

Least Cost Path to Lanzhou in 1000 Km's 141 0.078 2.274

Least Cost Path to Xining in 1000 Km's 141 0.003 2.142

Least Cost Path to Kunming in 1000 Km's 141 0.450 2.559

Contemporaneous Variables

Obs Mean Std. Dev

Fraction of Pop in Urban Area 141 0.169 0.208

Total County Pop. In Millions 141 0.064 0.073

Han Pop. As Fraction of Total Pop 141 0.167 0.243

Summary Statistics-Continued

Religion Variables

Obs Mean Std. Dev

Total Number of Registered Sites (source: Atlas of

Religions in China) 141 16.745 15.884

Historical Religious Variables (source: Tibet-Institut)

# of Bon Sites 141 0.305 0.654

# of Gelug Sites 141 1.149 1.526

# of Kagyu Sites 141 0.333 0.900

# of Nyingma Sites 141 0.390 0.954

# of Shakya Sites 141 0.305 0.765

Incident Variable

Obs Mean Std. Dev

Protest Recorded 141 0.333 0.473

# of Incidents 141 0.759 2.204

=1 if Casualty Reported 141 0.128 0.335

=1 if Detention Reported 141 0.227 0.420

=1 if Fatality Reported 141 0.064 0.245

=1 if Layperson Involved 141 0.213 0.411

=1 if Monks Involved 141 0.277 0.449

=1 if Nuns Involved 141 0.050 0.218

=1 if Students Involved 141 0.050 0.218

=1 if Violence by Security Reported 141 0.121 0.327

=1 if Violence by Protestors Reported 141 0.050 0.218

Number of Incidents, by Province

Name

Total # of

Counties

# of Counties

with Protest

# Counties with

No Protest

Gansu 7 6 1

Qinghai 39 13 16

Sichuan 26 14 12

Tibet 66 14 52

Yunnan 3 0 3

Total 141 47 94

Table 2: Chinese Heritage & Probability of Protest

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Probit Probit Probit VARIABLES dF/dX dF/dX dF/dX OLS OLS OLS IV-2SLS IV-2SLS IV-2SLS

Area-weighted Historical Presence -1.448*** -1.292*** -1.101** -0.821*** -0.585*** -0.483*** -3.180*** -3.121** -2.546* (0.393) (0.425) (0.442) (0.190) (0.176) (0.184) (1.100) (1.308) (1.305)

County Mean Latitude -0.010 0.049 0.064 -0.018 -0.037 -0.011 0.026 0.048 0.059

(0.035) (0.127) (0.137) (0.036) (0.111) (0.106) (0.054) (0.158) (0.145) County Mean Longitude 0.046*** -0.241** -0.223* 0.027* -0.238*** -0.190*** 0.004 -0.011 -0.035

(0.018) (0.122) (0.128) (0.015) (0.061) (0.068) (0.020) (0.205) (0.170)

County Mean Elev. Km -0.165 -0.400 -0.250 -0.076 -0.156 -0.144 -0.851** -0.919** -0.827* (0.192) (0.267) (0.295) (0.110) (0.139) (0.141) (0.385) (0.427) (0.440)

County Elev. Std. Dev. Km -0.001** -0.001* -0.001 -0.001** -0.000* -0.000 -0.001** -0.001** -0.001*

(0.000) (0.000) (0.000) (0.000) (0.000) (0.000) (0.001) (0.001) (0.001) County Area in Mm-Sq -1.866 -6.006 -8.877 -0.535 -2.939 -2.248 -2.401 -1.543 -1.295

(3.674) (5.466) (6.692) (1.358) (1.887) (2.109) (2.534) (2.615) (2.481)

Temperate Forest as Frac. of County Area

-2.646* -3.436** -2.940* -1.581 -2.276 -2.083 -3.232* -3.356** -3.329*

(1.358) (1.606) (1.787) (1.315) (1.514) (1.364) (1.668) (1.679) (1.685)

Shrubs as Frac. of County Area -2.553* -3.217** -2.870* -1.466 -2.100 -1.922 -3.225* -3.325** -3.229* (1.347) (1.547) (1.716) (1.333) (1.480) (1.329) (1.650) (1.648) (1.651)

Desert as Frac. of County Area -1.473 -1.932 -2.115 -3.035* -3.501* -3.734*

(1.379) (1.639) (1.488) (1.778) (2.020) (1.941) Tropical Forest as Frac. of County

Area

-1.580 -1.813 -2.861 -17.587* -16.964* -15.621*

(3.356) (3.065) (3.403) (9.539) (10.124) (9.202) Least Cost Path to Chengdu in

1000Km

-1.779* -2.045* -0.957 -1.163 -1.663 -2.012

(1.035) (1.113) (0.859) (0.882) (1.590) (1.397) Least Cost Path to Lanzhou in

1000Km

-2.088 -1.353 -2.142** -1.096 1.841 1.968

(1.587) (1.657) (1.079) (1.151) (2.851) (2.427) Least Cost Path to Xining in

1000Km

1.260 0.717 0.522 0.167 -1.062 -1.026

(1.294) (1.403) (0.941) (0.934) (1.374) (1.284) Least Cost Path to Kunming in

1000Km

-0.243 0.259 -0.354 -0.108 0.854 0.830

(0.851) (0.944) (0.782) (0.753) (1.139) (1.083)

# of Bon Temples -0.123 -0.082 -0.007

(0.087) (0.051) (0.075)

# of Gelug Temples 0.071* 0.045 0.056 (0.037) (0.031) (0.036)

# of Kagyu Temples -0.068 -0.050 -0.010

(0.063) (0.041) (0.055) # of Nyingma Temples 0.011 0.015 -0.038

(0.076) (0.037) (0.060)

# of Shakya Temples 0.085 0.046 0.101 (0.094) (0.051) (0.064)

Total Number of Registered

Religious Sites

0.005 0.006* 0.003

(0.004) (0.004) (0.004)

Observations 133 133 133 141 141 141 141 141 141 Pseudo/Adjusted R-Squared 0.287 0.329 0.398 0.124 0.175 0.229

Number of Provinces

t-statistic (excluded instrument)

5 5 5 5

4.71

5

3.68

5

3.14

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 3: Chinese Heritage & Protest Incidents

(1) (2) (3) (4) (5) (6) (7) (8) (9)

VARIABLES Neg. Bin. Incidence

Ratios

Neg. Bin. Incidence Ratios

Neg. Bin. Incidence

Ratios

OLS OLS OLS IV-2SLS IV-2SLS IV-2SLS

Area-weighted Historical

Presence

0.008*** 0.021*** 0.042** -2.238*** -0.980 -0.461 -13.752** -13.728** -9.383

(0.012) (0.031) (0.058) (0.848) (0.650) (0.869) (5.791) (6.344) (6.614) County Mean Latitude 0.885 3.083** 2.624* -0.003 -0.106 -0.364 0.209 0.323 -0.059

(0.112) (1.639) (1.305) (0.108) (0.533) (0.635) (0.226) (0.680) (0.786)

County Mean Longitude 1.106* 0.766* 0.767* 0.048 -1.166* -1.046* -0.065 -0.028 -0.374 (0.060) (0.117) (0.111) (0.039) (0.648) (0.545) (0.081) (0.859) (0.794)

Shrubs as Frac. of County

Area

0.303 0.336 1.878 0.384 -2.732 3.033 -8.198** -8.891** -2.621

(1.487) (2.278) (12.129) (3.289) (2.640) (4.766) (4.135) (3.801) (6.333)

Desert as Frac. of County

Area

0.000 0.000 0.000 0.586 -1.117 2.491 -7.034 -9.001* -4.513

(0.000) (0.000) (0.000) (3.472) (3.523) (5.217) (4.763) (5.347) (7.650)

Temperate Forest as Frac.

of County Area

0.118 0.038 0.417 -1.325 -4.748* 1.802 -9.383* -10.177** -3.587

(0.600) (0.267) (2.784) (2.931) (2.617) (4.767) (4.895) (4.461) (6.829)

Tropical Forest as Frac. of

County Area

0.000 0.000 0.000 -1.956 -4.738 2.721 -80.078* -80.888* -52.478

(0.000) (0.000) (0.000) (9.864) (11.701) (20.884) (43.346) (47.020) (52.468)

County Mean Elev. Km 0.477* 0.462 0.490 -0.519 -0.711 0.048 -4.303** -4.546** -2.905

(0.186) (0.362) (0.392) (0.347) (0.739) (1.077) (1.935) (2.105) (2.713) County Elev. Std. Dev.

Km

0.997** 0.998 0.998 -0.001** -0.000 -0.000 -0.004** -0.005* -0.003

(0.001) (0.002) (0.002) (0.001) (0.001) (0.001) (0.002) (0.002) (0.003) County Area in Mm-Sq 0.000 0.000* 0.000* -2.770 -16.006 -7.761 -11.877 -8.991 -3.638

(0.007) (0.000) (0.000) (4.553) (10.624) (10.594) (12.038) (11.281) (9.483)

Least Cost Path to Chengdu in 1000Km

0.130 0.006* -0.631 -3.534 -4.182 -7.207

(0.420) (0.017) (4.243) (3.708) (7.391) (6.598)

Least Cost Path to Lanzhou in 1000Km

0.014 5.286 -12.411 -6.967 7.609 6.288

(0.073) (26.840) (10.535) (5.991) (13.057) (12.594)

Least Cost Path to Xining

in 1000Km

16,021.862* 66.179 3.290 -2.111 -4.673 -7.269

(79,471.213) (320.562) (3.462) (4.470) (6.257) (6.393)

Least Cost Path to Kunming in 1000Km

0.003 0.116 -4.127 0.249 1.946 4.309

(0.010) (0.399) (3.014) (2.897) (3.897) (3.990)

# of Bon Temples 0.544* -0.165 0.156 (0.180) (0.191) (0.321)

# of Gelug Temples 1.295** 0.410 0.456

(0.137) (0.282) (0.303) # of Kagyu Temples 0.888 -0.022 0.154

(0.184) (0.213) (0.311)

# of Nyingma Temples 1.281 0.851 0.620 (0.207) (0.622) (0.714)

# of Shakya Temples 0.893 -0.641 -0.403 (0.186) (0.617) (0.677)

Total Number of

Registered Religious Sites

1.017 0.001 -0.014

(0.012) (0.014) (0.021)

Observations 141 141 141 141 141 141 141 141 141 Number of Provinces 5 5 5 5 5 5

Pseudo/Adjusted

R-Squared t-statistic (excluded

instrument)

-0.0375

0.00194 0.142

4.71

3.68

3.14

Standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Table 4a: Effect of Chinese Heritage on Current Urbanization

(1) (2) (3) (4) (5) (6) VARIABLES OLS OLS OLS IV-2SLS IV-2SLS IV-2SLS

Area-weighted Historical Presence 0.125 0.080 0.048 -0.591 -0.728* -0.923*

(0.106) (0.112) (0.111) (0.372) (0.436) (0.537) County Mean Latitude -0.019 -0.043 -0.050 -0.005 -0.016 -0.017

(0.015) (0.045) (0.049) (0.020) (0.053) (0.058) County Mean Longitude -0.016** -0.010 -0.012 -0.023*** 0.062 0.061

(0.006) (0.042) (0.047) (0.008) (0.072) (0.075)

Shrubs as Frac. of County Area -0.008 -0.167 -0.219 -0.543 -0.558 -0.834 (0.520) (0.540) (0.516) (0.556) (0.599) (0.651)

Desert as Frac. of County Area 0.682 0.412 0.393 0.207 -0.088 -0.369 (0.560) (0.628) (0.622) (0.576) (0.712) (0.774)

Temperate Forest as Frac. of County Area -0.078 -0.195 -0.248 -0.580 -0.539 -0.835

(0.507) (0.545) (0.531) (0.558) (0.610) (0.680) Tropical Forest as Frac. of County Area -4.791*** -5.665** -6.023*** -9.654*** -10.495*** -12.029***

(1.697) (2.221) (2.120) (3.238) (3.915) (4.553) County Mean Elev. Km -0.204** -0.225** -0.243** -0.439*** -0.468*** -0.564***

(0.081) (0.109) (0.112) (0.136) (0.162) (0.207)

County Elev. Std. Dev. Km -0.000 -0.000 -0.000 -0.000 -0.000** -0.001** (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

County Area in Mm-Sq -0.559 0.247 0.482 -1.126 0.692 0.931 (0.910) (1.219) (1.359) (1.178) (1.368) (1.646)

Least Cost Path to Chengdu in 1000Km -0.229 -0.247 -0.454 -0.647

(0.317) (0.355) (0.475) (0.579) Least Cost Path to Lanzhou in 1000Km 1.447** 1.408** 2.717*** 2.850***

(0.577) (0.642) (0.972) (1.071) Least Cost Path to Xining in 1000Km -1.469*** -1.501*** -1.974*** -2.063***

(0.460) (0.470) (0.598) (0.630)

Least Cost Path to Kunming in 1000Km 0.237 0.258 0.622* 0.700* (0.239) (0.261) (0.362) (0.414)

# of Bon Temples 0.005 0.040 (0.018) (0.029)

# of Gelug Temples -0.002 0.003

(0.010) (0.014) # of Kagyu Temples 0.029 0.048

(0.019) (0.029) # of Nyingma Temples -0.022 -0.047

(0.018) (0.037)

# of Shakya Temples -0.003 0.023 (0.018) (0.032)

Total Number of Registered Religious Sites -0.001 -0.002 (0.001) (0.002)

Observations 141 141 141 141 141 141 Number of Provinces 5 5 5 5 5 5

Pseudo/Adjusted R-Squared 0.202 0.250 0.230

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Table 4b: Effect of Chinese Heritage on Current Han Population

(1) (2) (3) (4) (5) (6) VARIABLES OLS OLS OLS IV-2SLS IV-2SLS IV-2SLS

Area-weighted Historical Presence 0.564*** 0.500*** 0.489*** 0.401* 0.312 0.243

(0.096) (0.096) (0.101) (0.212) (0.241) (0.269) County Mean Latitude 0.021*** 0.022 0.016 0.024*** 0.028 0.024

(0.008) (0.030) (0.032) (0.009) (0.033) (0.037) County Mean Longitude -0.006** 0.035 0.028 -0.008** 0.051 0.046

(0.003) (0.025) (0.028) (0.003) (0.038) (0.039)

Shrubs as Frac. of County Area 0.164 0.129 0.126 0.042 0.038 -0.030 (0.194) (0.184) (0.201) (0.217) (0.187) (0.223)

Desert as Frac. of County Area 0.661** 0.506 0.525 0.554** 0.390 0.332 (0.277) (0.306) (0.325) (0.258) (0.293) (0.330)

Temperate Forest as Frac. of County Area 0.159 0.142 0.145 0.045 0.063 -0.003

(0.197) (0.188) (0.215) (0.219) (0.187) (0.237) Tropical Forest as Frac. of County Area -0.957 -1.492 -1.550 -2.061 -2.613 -3.072

(2.489) (2.549) (2.602) (2.783) (2.833) (3.022) County Mean Elev. Km -0.066 -0.080 -0.079 -0.120* -0.136* -0.161*

(0.050) (0.062) (0.065) (0.072) (0.076) (0.090)

County Elev. Std. Dev. Km 0.000 0.000 0.000 0.000 0.000 -0.000 (0.000) (0.000) (0.000) (0.000) (0.000) (0.000)

County Area in Mm-Sq -0.716 -0.001 -0.038 -0.844 0.103 0.076 (0.550) (0.669) (0.710) (0.609) (0.713) (0.769)

Least Cost Path to Chengdu in 1000Km -0.092 -0.101 -0.144 -0.202

(0.274) (0.300) (0.247) (0.277) Least Cost Path to Lanzhou in 1000Km 1.203*** 1.094** 1.498** 1.459**

(0.444) (0.479) (0.614) (0.650) Least Cost Path to Xining in 1000Km -0.861** -0.852** -0.978** -0.994**

(0.348) (0.366) (0.411) (0.418)

Least Cost Path to Kunming in 1000Km 0.178 0.191 0.267 0.303 (0.161) (0.190) (0.212) (0.236)

# of Bon Temples 0.003 0.012 (0.011) (0.014)

# of Gelug Temples -0.001 0.001

(0.008) (0.008) # of Kagyu Temples 0.003 0.008

(0.009) (0.011) # of Nyingma Temples -0.005 -0.011

(0.012) (0.017)

# of Shakya Temples -0.006 0.000 (0.010) (0.013)

Total Number of Registered Religious Sites -0.001 -0.001 (0.001) (0.001)

Observations 141 141 141 141 141 141 Number of Provinces 5 5 5 5 5 5

Pseudo/Adjusted R-Squared 0.646 0.670 0.658

Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1

Table 5: Chinese Heritage & Endogenous Variables

(1) (2) (3) (4) Probit Probit

VARIABLES dF/dX dF/dX OLS OLS

Area-weighted Historical Presence -1.166** -0.938* -0.450 -0.372 (0.505) (0.527) (0.292) (0.272)

County Mean Latitude 0.061 0.065 0.003 0.024 (0.134) (0.133) (0.106) (0.107)

County Mean Longitude -0.162 -0.152 -0.181*** -0.171**

(0.132) (0.126) (0.066) (0.067) Shrubs as Frac. of County Area -2.591 -2.669 -1.998 -1.903

(1.729) (1.660) (1.327) (1.364) Desert as Frac. of County Area -1.997 -2.024

(1.490) (1.513)

Temperate Forest as Frac. of County Area -2.506 -2.538 -2.122 -2.012 (1.814) (1.742) (1.365) (1.400)

Tropical Forest as Frac. of County Area -3.629 -1.915 (3.157) (3.221)

County Mean Elev. Km -0.184 -0.119 -0.177 -0.111

(0.332) (0.318) (0.137) (0.137) County Elev. Std. Dev. Km -0.001 -0.001 -0.000 -0.000

(0.001) (0.001) (0.000) (0.000) County Area in Mm-Sq -7.914 -7.443 -2.309 -2.483

(6.526) (6.356) (2.093) (2.167)

# of Bon Temples -0.112 -0.117 -0.077 -0.078 (0.084) (0.086) (0.051) (0.050)

# of Gelug Temples 0.075** 0.075** 0.041 0.042 (0.037) (0.037) (0.031) (0.031)

# of Kagyu Temples -0.058 -0.059 -0.044 -0.053

(0.061) (0.061) (0.041) (0.042) # of Nyingma Temples 0.006 -0.001 0.014 0.020

(0.072) (0.071) (0.037) (0.037) # of Shakya Temples 0.072 0.076 0.046 0.046

(0.091) (0.092) (0.051) (0.051)

Total Number of Registered Religious Sites 0.003 0.004 0.005 0.005 (0.004) (0.004) (0.004) (0.004)

Least Cost Path to Chengdu in 1000Km -2.190** -2.051* -1.374 -1.320 (1.087) (1.093) (0.913) (0.879)

Least Cost Path to Lanzhou in 1000Km -0.547 -0.771 -0.549 -0.791

(1.803) (1.861) (1.154) (1.163) Least Cost Path to Xining in 1000Km 0.282 0.448 -0.155 0.176

(1.463) (1.486) (0.980) (0.999) Least Cost Path to Kunming in 1000Km 0.750 0.728 0.008 -0.037

(0.966) (0.966) (0.768) (0.757)

Han Pop. As Frac. of Total Pop. -0.842 -1.279* -0.411 -0.623** (0.667) (0.760) (0.250) (0.282)

Total County Pop. In Millions 2.179 2.180 0.904 0.950 (1.411) (1.439) (0.724) (0.646)

Frac. of Pop. in Urban Area 0.517 0.340*

(0.363) (0.176)

Observations 133 133 141 141 Pseudo/Adjusted R-Squared 0.415 0.427 0.233 0.243

Number of Provinces 5 5

Standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1