The big oil change: a closer look at the Haber-Menaldo analysis. Jørgen J. Andersen BI Norwegian Business School - PDF

This file was downloaded from the institutional repository BI Brage - (Open Access) The big oil change: a closer look at the Haber-Menaldo analysis Jørgen J. Andersen BI Norwegian

Please download to get full document.

View again

of 54
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.

Word Search

Publish on:

Views: 34 | Pages: 54

Extension: PDF | Download: 0

This file was downloaded from the institutional repository BI Brage - (Open Access) The big oil change: a closer look at the Haber-Menaldo analysis Jørgen J. Andersen BI Norwegian Business School Michael L. Ross University of California, Los Angeles This is the authors accepted and refereed manuscript to the article published in Comparative Political Studies, 47(2014)7: DOI: The publisher, Sage, allows the author to retain rights to post the accepted version (version 2) of the article in any repository other than those listed above (i.e. you may not deposit in the repository of another institution or a subject repository) until 12 months after first publication of the article in the journal (Publishers policy May 2014) The Big Oil Change: A closer look at the Haber-Menaldo analysis Jørgen Juel Andersen (BI Norwegian Business School) Michael L. Ross (UCLA) Draft: July 26, 2012 Word count (including footnotes and references but excluding tables and appendix): 9372 Abstract: The claim that oil wealth tends to block democratic transitions has recently been challenged by Haber and Menaldo (2011), who argue that past studies were tainted by endogeneity and omitted variable bias. Using historical data going back to 1800, and models with country and year fixed effects, they conclude there is no evidence of a resource curse. We revisit their data and models, and show they are only correct for the period from 1800 to the 1970s: since about 1980, there has been a pronounced resource curse. We argue that oil wealth only became a hindrance to democratic transitions after the transformative events of the 1970s, which dramatically increased the economic importance of oil, and enabled developing country governments to capture the oil rents that were previously siphoned off by foreign-owned firms. We also explain why the Haber-Menaldo study failed to identify this: partly because the authors draw invalid inferences from their longitudinal analysis of resource-rich states; and partly because they assume that the relationship between oil wealth and democracy has not changed for the last 200 years. Earlier versions of this paper were presented at seminars at the Norwegian University of Science and Technology (NTNU), the Norwegian Business School (BI), Stanford University, and the Annual Meeting of the American Political Science Association. For their helpful comments on earlier drafts of this paper, we thank Silje Aslaksen, Lisa Blaydes, Sambit Bhattacharyya, Bill Clark, Indra de Soysa, Larry Diamond, Steve Haber, Kevin Morrison, Irfan Nooruddin, Nathan Nunn, Paul Poast, Ragnar Torvik, Jay Ulfelder, and David Wiens. Corresponding author: Michael Ross, UCLA Political Science Department, P.O. Box , Los Angeles CA Tel: Fax: Jørgen Juel Andersen is Associate Professor in Public Policy in the Department of Accounting, Auditing and Law, at BI Norwegian Business School in Oslo. Michael L. Ross is Professor of Political Science at the University of California, Los Angeles. 1 Many studies have found that authoritarian countries with more oil wealth are less likely to transition to democracy. Haber and Menaldo (2011) challenge these studies, arguing they are tainted by reverse causality and omitted variable bias. Using new data on natural resource wealth for the years , and statistical models that control for country fixed effects and many other factors, they find no evidence to support the resource curse claim. They conclude that: No matter how we look at the long-run data including just making simple country-by-country graphs we cannot find a systematic tendency that matches the concept of a resource curse (3). 1 The Haber-Menaldo article has had a powerful impact on the resource curse debate, calling into question widely-held beliefs about the politically malignant effects of petroleum wealth. In the first eighteen months after its February 2011 publication, it was cited more than 100 times by other scholars, and featured in influential policy journals (Kenny 2010), World Bank publications (Sinnot, Nash, and de la Torre 2010; Barma et al. 2011), and prominent blogs as evidence that claims about the resource curse are false. 2 An earlier study by Gurses (2009) used a similar fixed-effects model and came to a similar conclusion. 1 Although the term resource curse can refer to many different phenomena, we follow Haber and Menaldo in using this term to refer to the hypothesis that countries with greater oil wealth are less likely to transition from authoritarian to democratic rule. Like Haber and Menaldo, we focus on the effects of oil and natural gas which make up over 90 percent of the world minerals trade rather than nonfuel minerals. 2 The Haber-Menaldo study was featured on the Freakonomics blog on April 4, 2011, and The Monkey Cage blog on August 23, The number of citations is from a Google Scholar search on April 10, Why do these findings differ so much from other recent studies, which conclude that there is a resource curse? 3 We employ the Haber-Menaldo data and models and show that each side of the debate is partly correct: from 1800 to the 1970s the period that dominates the Haber-Menaldo dataset there is no strong evidence of a resource curse. Yet since the late 1970s the period that is the focus of most other studies oil wealth has strongly inhibited democratization. The emergence of a resource curse or more properly, a petroleum curse in the late 1970s is consistent with a closer look at the history of the global oil industry. Although the Haber-Menaldo analysis begins in 1800, no country produced economically significant quantities of oil before Until the late 1960s, most of the rents generated by oil production in non-western countries were captured by a handful of large, vertically-integrated international oil companies sometimes called the Seven Sisters. 5 But in the 1970s, the industry was transformed by a wave of nationalizations and contract revisions that enabled the governments of host countries to seize control of these rents. We refer to this transformation as the big oil change. Theories of the rentier state were formulated, beginning in the mid-1980s, in response to the big oil change; the idea that the oil producers were afflicted by a resource 3 Aslaksen (2010), Ramsay (2011), Tsui (2010), Cuaresma, Oberhofer, and Raschky (2010), Andersen and Aslaksen (forthcoming), Ross (2012). 4 According to the Haber-Menaldo data, in 1918, Mexico became the first country to produce $100 per capita worth of oil. By 2006, 49 countries were producing at least $100 per capita of oil (constant dollars). 5 The seven companies were Standard Oil of New Jersey (later Exxon), Standard Oil of California (later Chevron), Anglo-Iranian Oil Company (later BP), Mobil, Texaco, Gulf, and Royal Dutch Shell. 3 curse began to circulate in the early to mid 1990s. 6 In both literatures, the central concern is what happens to a country s politics and economy when the state accumulates large resource rents a condition that only became widely true in the 1970s. The Haber- Menaldo study, however, combines data from 175 years when governments did not typically capture most of these oil rents (from 1800 to about 1975), with data from about 30 years when they did capture them. The powerful anti-democratic effects of oil since the late 1970s are hence obscured by the weaker relationship between oil and democracy in the period. We use Haber and Menaldo s data and error-correction model to illustrate this change. When we allow for a break in the effect of oil around 1980, we find that oil has strongly inhibited democratization in the post-break period. This result holds in the presence of country and year fixed effects, and under a wide range of conditions: with each of the variables that Haber and Menaldo use to measure resource wealth; when we use a dynamic fixed effects model in place of Haber and Menaldo s error correction model (ECM); and when we use any year between 1982 and 1990 to identify the temporal break. We also show that when oil income is allowed to affect regime types over three, five, or seven years, rather than a single year, these anti-democratic effects become much larger and emerge earlier. Haber and Menaldo claim that their finding of no resource curse is also supported by other evidence particularly their longitudinal analysis of 53 resource- 6 The earliest modern study of the rentier state was published by Mahdavy in 1970, and focuses largely on Iran. Little else was written on the topic before the late 1987, when Luciani and Beblawi edited a seminal volume. The term resource curse dates to a study by Richard Auty in 1993, and was popularized by a 1995 paper by Jeffrey Sachs and Andrew Warner. 4 reliant states. They report that most of these countries eventually became more democratic, even though they had been treated with resource wealth. This leads them to conclude that 45 of these 53 states were either resource blessed or unaffected by resource wealth, while just eight of them might have been resource cursed. We show that Haber and Menaldo s inferences from these data are invalid: it is not possible to determine whether a given treatment (oil wealth) has had an effect by only observing countries that have been treated. To make valid inferences about a treatment, researchers must compare outcomes in the treated population to outcomes in a control population. Haber and Menaldo observe that countries treated with oil wealth have grown slightly more democratic over time, and interpret this as evidence against the resource curse. We show, however, that the oil states examined by Haber and Menaldo (the treated group) made much less progress towards democracy than the non-oil states (the control group). In other words, once we employ the correct counterfactual, we observe that the oil treatment strongly inhibited democratization. 7 Despite our criticisms, there is also much we admire about the Haber Menaldo study. They have offered a smart and spirited challenge to a well-established (but not always rigorous) literature; they have gathered a large quantity of new historical data, and made their data public and transparent; and they have helpfully clarified some of the conditions under which resource wealth is not associated with less democracy. Yet their study obscures the powerful effect that oil revenues have had on authoritarian rule over 7 There is another important reason that Haber and Menaldo fail to identify a resource curse: they decline to test the most credible version of the resource curse hypothesis, which is that when autocratic states collect a lot of oil revenues, they become less likely to transit to democracy (e.g., Smith 2004; Ulfelder 2007; Papaioannou and Siourounis 2008; Ross 2009; Morrison 2009; Andersen and Aslaksen 2013, Clark, Poast, and Weins 2012). 5 the last three decades the period of greatest concern for most researchers and policymakers. Our findings have important implications for the study of both the resource curse, and democratic transitions more broadly. Virtually all of the resource curse literature assumes that the malignant (or benign) effects of petroleum wealth have changed little over time; some studies draw explicit parallels between the influence of oil today and other resources in the past (e.g., Karl 1997). More generally, studies of democracy have increasingly employed historical data to identify the factors associated with democratic transitions (Acemoglu, Johnson, Robinson, and Yared 2009; Boix 2011; Freeman and Quinn 2012). A key assumption in most of this research is that the correlates of democracy are constant over time; if this were true, datasets that cover longer periods of time should give scholars more leverage to identify these correlates. But if causal relationships tend to change over time, the identification of short or medium term causal effects in one era might tell us little about their salience in a different era. Our study shows that the causal effect of at least one important factor petroleum wealth on democratic transitions changed sharply from the 1960s to the 1980s, as the global distribution of petroleum rents shifted from firms to governments. We hope our analysis encourages scholars who use long datasets to study the resource curse, democratization, and other social and political phenomena to become more sensitive to changing historical patterns, including discontinuities in economic and political relationships. The next section of this paper uses simple graphs and cross-tabs to illustrate our argument that oil wealth only began to inhibit democratic transitions after the 1970s. 6 Section two suggests this was probably caused by the wave of nationalizations that swept the oil-producing world in the 1970s. In section three we begin our analysis of the Haber-Menaldo results, focusing on their longitudinal analysis of resource-rich countries; we explain why their inferences from these data are logically flawed. We replicate Haber and Menaldo s core statistical results in section four, and show how they are altered once we account for a temporal break in the effect of oil in the late 1970s or early 1980s. Section five shows that these results are robust. Our paper concludes by reflecting on the implications of the big oil change of the 1970s, and the methodological issues raised by the Haber-Menaldo study. 1. Looking at oil and democracy over time Most studies of the resource curse rely on datasets that begin in 1960 or 1970; Haber and Menaldo construct a dataset that goes back much earlier for some of their variables, to They argue that this unusually long time-series allows them to identify the longrun equilibrium relationship between natural resources and regime type. Yet most of the years in the Haber and Menaldo dataset are uninformative: between 1800 and 1860, no country produced a single barrel of oil; and until the 1940s, only a couple of countries chiefly the US, Venezuela, and Mexico produced economically significant quantities. Figure 1 shows that the number of countries producing significant quantities of oil which we define as $100 per capita (in constant 2007 dollars) over the 207 year period covered by their data. In 2006, there were 49 such countries. Yet no country crossed this modest threshold before 1918; as late as 1949, there were just four significant oil producers. 7 Moreover, using a very long time series even when time-series and crossnational observations are pooled, as in about half of the Haber-Menaldo models has an important drawback: it can open the door to misleading inferences, if the relationship between the independent and dependent variables has changed over time. There is good reason to suspect that the relationship between oil and regime types has indeed changed over time. In Figure 2 we plot the mean polity scores of the oilabundant countries and compare them to all other states. 8 While the polity scores of the oil countries was slightly below the score of the non-oil countries for most of this period, they seem to diverge sometime in the early 1980s. Both sets of countries became more democratic after 1976, but the non-oil countries moved further and faster towards democracy. Of course, this figure may be suspect because the composition of both sets of countries changes over time: more countries are becoming sovereign, increasing the membership of both groups; and countries shift between oil abundant and not oil abundant when they discover or run out of oil, and when global oil prices rise or fall. Figure 3 shows a simple way to circumvent this problem: it compares the polity scores of two constant groups of countries the eleven states that produced at least 100 constant dollars per capita in oil income in 1960, and continued to be significant producers 8 Note we are not looking at country s oil dependence (meaning the ratio of oil exports to GDP), but its oil abundance (meaning the value of a country s oil and gas production divided by its population). As many scholars have pointed out, measures of oil dependence are biased upwards in poor countries, making it a poor measure of resource wealth (Ross 2004, 2009; Dunning 2008). We once again define oil abundant countries as those whose oil income exceeded $100 dollars per capita (in constant 2007 dollars) in a given year. 8 throughout this period, and the remaining 107 states that were also sovereign in Now the divergence between the two lines is sharper, and appears to start in the late 1970s. In the Appendix, we demonstrate that the results do not change if we move our base year from 1960 to 1970 (Figure A1), or if we raise the threshold for identifying oil producers from $100 to $500 dollars (Figure A2). 10 We can also see the temporal break with simple cross-tabulations, by comparing the rate of democratic transitions (still using the Haber Menaldo data) among countries that produce at least 100 dollars in oil income, compared to everyone else (Table 1, top half). Over the whole 203 year period covered by the Haber Menaldo data (row one), there is no statistically significant difference between the oil producers and the non-oil producers. The similarity of these two numbers is the basis of much of the Haber- Menaldo analysis. Since production was minimal in all but a few countries before 1940, we think it is more enlightening to focus on the period between 1940 and 2002 (which is the last year in the Haber-Menaldo data containing their dichotomous measure of regime types). From 1940 to 1980 (row two), oil producers and non-oil producers had democratic transitions at almost exactly the same rate; after 1980 (row three), the non-oil producers became democratic at more than twice the rate of the oil producers. If we move the threshold for identifying oil producers up to $500 (Table 1, bottom half), the patterns are 9 The eleven oil producers are Bahrain, Canada, Gabon, Iran, Iraq, Kuwait, Qatar, Saudi Arabia, Trinidad, United States, and Venezuela. 10 Using 1970 as the base year yields 16 significant oil producers: Algeria, Bahrain, Canada, Gabon, Iran, Iraq, Kuwait, Libya, Oman, Qatar, Russia (Soviet Union), Saudi Arabia, Trinidad, United States, the United Arab Emirates, and Venezuela. Moving the threshold to $500 in 1970 creates a group of 11 countries: Bahrain, Gabon, Iraq, Kuwait, Libya, Oman, Qatar, Saudi Arabia, Trinidad, United Arab Emirates, and Venezuela. 9 even stronger: since 1980, the non-oil producers have been five times more likely to transit than the oil producers. In short, both a visual inspection of the historical data, and simple crosstabulations, suggests that the democracy paths of oil-producers and non-oil producers diverged sometime in the late 1970s or early 1980s. 2. Historical Change in International Oil Markets Why did this occur? Most theories of the resource curse argue that oil helps prolong authoritarian rule because it generates large rents for the government, which the ruler can use to both lower taxes and increase patronage and pork barrel spending (e.g., Mahdavy 1970, Luciani 1987, Ross 2001). But governments have not always been able to capture these rents. Before the 1970s, the global petroleum industry was dominated by a handful of vertically-integrated companies that colluded to maintain control of world supplies (Yergin 1991). In all but a few countries, these firms owned or controlled the local subsidiaries that extracted and exported the host country s oil. Globally, they controlled the shipping and marketing of almost all of the world s petroleum, and used both highlyfavorable contracts and transfer pricing to capture most of the rents for themselves (Hartshorn 1962, Levy 1982). In the 1960s and 1970s, international petroleum markets were transformed by a series of closely-related developments: oil supplies begin to grow tighter, as rising demand outpaced new discoveries; the major oil exporters of the developing world began to col
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks