Sie sind auf Seite 1von 38

Good Policy or Good Luck in Latin America?

A View from the Stock Market*


Juan Jos Cruces

Javier Garca-Cicco

Universidad Torcuato Di Tella

Universidad Catlica Argentina Central Bank of Chile

October, 2011

Abstract
We propose a methodology to disentangle good policy from good luck using stock market price data. In particular, we decompose average returns into Global, Regional and National contributions. We interpret the Global contribution as worldwide luck, the Regional contribution as a mixture of common merit and neighborhood luck of the countries in the sample, and the National contribution as reflecting the differential merit of presidents (i.e. good policy). We apply this approach to seven Latin American countries from 1976 until 2010. Our results indicate that, in assessing the good performance exhibited over the last decade, in most cases good luck was relatively more important than good policies. We also explore if our measures of good policy and good luck are correlated with presidential approval rates, finding that in some cases global factors can be relevant in explaining the support of domestic policy by the population. JEL Codes: H11, E63, G15

We would like to thank Guido Sandleris and Juan Pablo Nicolini for comments, UErnesto Risso (Editorial de Comunicaciones Pblicas), Alejandro Izquierdo for providing quarterly terms of trade data, Urbi Garay for providing Venezuelan market exchange rate data, Ricardo Hermelo (TNS Gallup Argentina), Jos Luis Machinea, Lucas Klobovs (Ipsos), Roby Muntoni and Michael Orzano (S&P Index Services), Alejandra Fernandez (MSCI). Pablo Gonzalez Ginestet and Matas Vieyra provided excellent research assistance. The views and conclusions presented in this paper are exclusively those of the authors and do not necessarily reflect the position of the Central Bank of Chile or its Board members. Email: juan.cruces@utdt.edu Email: javier_garcia@uca.edu.ar, jgarcia-cicco@bcentral.cl

I hold it to be true that Fortune is the arbiter of one-half of our actions, but that she still leaves us to direct the other half, or perhaps a little less. Niccol Machiavelli, The Prince (1515)

1. Introduction
The last quarter of the twentieth century was a turbulent time for most countries in Latin America (LA); experiencing financial booms and busts, currency crisis, periods of chronic (or even hyper-) inflation, defaults on debt obligations, among other recurrent and painful economic problems. In most of these episodes, bad policy choices were at the root of the problem (e.g. government over-borrowing, fiscal and monetary inconsistencies, excessive regulation of financial markets, highly distortionary trade policies, etc.). In sharp contrast, the last decade has been more tranquil and prosperous, with economies experiencing high rates of growth and with an increasing role of LA countries in the world economy. 1 Given the role that bad policy decisions played during the periods of economic distress, one might be tempted to argue that the recent flourishing period can be attributed to good policy. However, it is also true that international conditions have recently been more favorable for the region, with important and persistent increases in terms of trade and more widely available international capital. This complicates the task of distinguishing between good policy and good luck. It is noteworthy however, that administrations with sharply contrasting economic policy views have all witnessed higher growth and economic wellbeing than during the previous decade. While each administration claims that the contemporaneous success speaks to the suitability of their own policies one wonders to what extent these recent developments are truly the result of national policies or of a favorable aggregate shock that affects the region as a whole, making the effects of local policies of second order importance. In this paper, we present a methodology that exploits stock market data to identify the role that favorable external conditions have played in the evolution of LA economies over the last 25 years. In particular, we decompose the average return of stock market indices during the mandate of a given president in three components: Global, Regional, and National. We will interpret the first one as a measure of worldwide luck; the second one will capture a mixture common luck for the regions and common regional merit; while the last one will reflect the differential contribution of each president (i.e. good policy). We apply our approach to the universe of LA countries with relatively liquid public equity markets: Argentina, Brazil, Chile, Colombia, Mexico, Peru, and Venezuela. Disentangling good policy from good luck is a difficult task, and complications arise from two different fronts. First, we have to decide which variable is appropriate to assess how policy may affect the performance of the economy. Second, conditional on the chosen unit of measure, we have to build an appropriate counterfactual of economic performance that would allow us to asses the importance of good luck. In terms of the first issue, we argue that stock prices are (at least conceptually) particularly well suited to asses the impact of economic policies due to their forward-looking nature. Stock prices depend on the expectation about future economic developments, and the implementation of a policy that is expected to improve the economic
1

Albeit with only a relatively minor downturn associated with the global financial crisis and posterior recession triggered in 2008 by the collapse of Lehman Brothers. Also Argentina suffered a particularly severe crisis associated with the end of convertibility around the turn of the century.

situation in the future (either near or more distant) should be reflected in these prices. Other variables that are relatively more backward looking (notably, GDP) will not necessarily capture these expected effects, making them a possibly biased measure of economic performance. In other words, our approach will be judging administrations by the perceived value generated by their policies at the time of their disclosure to the market, which is consistent with the view that policies should be evaluated relative to the situation in which they were implemented, and not necessarily based on an ex-post analysis. In Section 2 we will discuss in more detail the merits of stock prices as a measure to capture the effect of policy relative to other alternatives, where we will also comment on some possible limitations. In term of the second task, our main methodological contributions is to compare the national portfolio index with the evolution of a portfolio that is evaluated at world prices but that has the industrial composition of the domestic index (which we label the twin index). We argue that this is an appropriate counterfactual of economic performance given that industrial composition of stock market indices in LA is markedly different than in developed economies and, moreover, they are important differences even between LA countries. Importantly, this twin index will be country specific which would allow us to have a counterfactual for the global conditions that is different for each country. To motivate this choice, we present an example based on the recent experience of Argentina and Brazil. Figure 1 displays the evolution of the Argentine and Brazilian stock market indices (the lines labeled P_Argentina and P_Brazil), measured in terms of the corresponding domestic units of consumption. 2 We can see how the stock market flourished in these countries between 2003 and 2007, both experiencing important downturns during the global financial crisis, recovering strongly thereafter. A first natural approach to assess the relative performance with respect to the rest of the world is to compare this recent evolution with a global index (P_World). The excess return in both countries was quite significant, which might lead us to say that good luck played a limited role. However, as we argued, comparing domestic and global indices might be misleading, for the industrial composition of each portfolio could be significantly different across the two countries. Therefore, we also report in each figure the evolution of the twin index for each of these countries (P_Twin_Arg and P_Twin_Brazil). Comparing the domestic index with the corresponding twin index, in the case of Argentina these two lines are closer to each other, particularly after 2007, which puts more importance on the good-luck hypothesis. On the contrary, the excess return of the Brazilian index relative to its twin index is relatively more significant, which might be an indication that the role of domestic factors (good policy among them) was more relevant for Brazil. Our methodology will seek to decompose these excess returns relative to the country-mimicking global portfolio, controlling also for the evolution of terms of trade. After obtaining our decomposition for the Global, Regional, and National contributions, we also investigate whether these variables are correlated with measures of presidential approval based on public opinion survey. While in principle it might we desirable that peoples opinion about economic policy the government depends on whether good policies are being implemented or not, it is not clear a priory that this is the results we will find given that dissecting good policy
2

The definition of the series is presented in Section 2.

from good luck is a complicated task. In fact, the results show that external factors are many times more relevant that the national contribution to explain the evolution of approval rates. The rest of the paper is organized as follows. Section 2 discusses the advantages of using stock market data, indicating some possible limitations as well, and puts our work in the context of the related literature. Section 3 describes the data and the econometric methodology. Section 4 contains the results of our decomposition, comparing frits countries and then presidents within each country. We then present the results of the regressions explaining presidential approval rates. Finally, Section 5 concludes.

2. Conceptual Framework and Related Literature


In this section we first discuss the advantages of looking at stock prices to assess the role of economic polices. We then present a simple conceptual framework to understand how different polices may impact the national stock price index, as well as to how it may be affected by global and regional conditions. In third place we place our work in the related literature, and we finish with a discussion of some possible limitations of our approach.

2.1. The Benefits of Using Stock Market Data


In assessing the impact of government policies, the ideal object of study would probably be the contribution of each administration to societys economic wellbeing. Unfortunately, this is not a feasible object and therefore one has to look for an appropriate proxy. The general choice in the literature (as we discuss below) is some macroeconomic aggregate, generally the growth rate of real gross domestic product (GDP). In contrast, we propose to focus on changes in the value of a diversified basket of publicly traded local equities. Thus, we will be measuring the soundness of policy from the standpoint of a local buy-and-hold investor in publicly listed equities. Given this difference with the usual practice, it is important to explain the advantages of this metric. While of course economic growth is a relevant variable to analyze, we think that to look at GDP to try to disentangle good policy from good luck has several important drawbacks. First, changes in GDP will probably be silent the intertemporal effects of policy. For instance, we can think of a president that implemented polices that allowed to spur growth during its mandate, but those policies may be detrimental for long run growth (e.g. to deplete the stock of non-renewable resources). It may also be the case that a policy that is implemented by a given president has its full effect on the next presidents term. A related example is the discovery of the new capital stock (for instance, the natural gas in Camisea, Peru, or the discovery of new oil fields Brazils Atlantic shore). In all these case, looking at GDP will probably assign the contribution to the wrong period. Another example is countercyclical policy; for instance, fiscal rules to save the proceeding from exports of commodities. While these types of policies are generally considered as desirable from a long run perspective, a country that runs countercyclical policy in a booming period will likely experience a smaller increase in GDP that another that implements pro-cyclical policies. Another issue related with business cycles fluctuation is that governments taking office at the trough face of the cycle will probably witness a GDP increase pretty much regardless of the type of policies implemented. 3

These examples highlight the fact that GDP is (relatively) a backward-looking variable, and therefore can give us a biased measure to asses the role of good policy. Stock market prices, on the contrary, are relatively more forward looking. The key assumption for this claim is informational efficiency: stock prices will impound all information about expected future conditions. Stock prices will attribute a positive contribution to the official in office when improvement in future conditions first becomes likely, regardless of whether they impact economic activity in the present or in the future. Moreover, using stock prices is easier to control for changes external conditions and for business cycle effects, and can then be used to sort out world and regional shocks from idiosyncratic ones. Thus, in this context, a measure based on asset prices should mitigate the problems of a measure based on GDP. 3

2.2. Conceptual Framework


Before turning to the empirical analysis it will be instructive to lay down a simple conceptual framework to highlight the determinants of this index and how can it be affected by government policies. Our object of interest is the value of the stock market index in a given country. Given the open economy nature of these economies, it is convenient to first break down the domestic index (in a generic initial period 0 in country i, P0i ) into the value of firms in the tradable sector ( P0i ,T ) and in the non tradable sector ( P0i , NT ). The domestic index will be a weighted average of these two sectoral indices, i.e.
P0i = 0 P0i ,T + (1 0 )P0i , NT .

(1)

Consider representative firm in sector j (j=T, NT). The value of that firm expressed in US dollars will be given by the discounted sum of expected future profits conditional on the information available at time 0, this is
P
i, j 0

(
t =0

E 0 ti , j 1 + r0f + r0World + r0LATIN + r0i

( )

(2)

The (after-tax) profits in each period t in the numerator in equation (1) is given by,

ti , j = [ ptj qtj C j ( t , q tj , ptT , p tNT )](1 t ),

(3)

where Cj(.) represents the cost function in sector j. In this expression, ptT and ptNT denote the prices of traded and non-traded goods respectively, both expressed in dollars. The price of tradable goods clearly has a positive impact on the income side of firm profits in the tradable sector, and can potentially increase costs in either sector if firms use some traded goods as intermediate inputs. In a similar way, an increase in the price of non-tradables will raise profits in
Of course, as any proxy, ours will be an imperfect measure of our ideal object, and we will discuss some of the limitations below.
3

the non-tradable sector and can potentially increase costs if these goods are used as inputs. The cost function is also increasing in the quantity produced ( qtj ), while the variable t denotes productivity, which tends to reduce costs, and we think of it having both a global and a domestic component, t = tWorld + ti . Finally, given that what is priced is the after-tax profit, the variable

t denotes the tax rate on corporate profits, so that (1 t ) is the fraction of corporate profits that

the owner gets to keep. The denominator in (2) represents the rate at which future profits are discounted. 4 In particular we highlight that the relevant discount rate for domestic cash flows to equity has several components: a) the risk free rate ( r0f ); b) the world market risk premium ( r0World ) representing the compensation required by investors for perceived global risk; c) the Latin American risk premium differential ( r0LATIN ); and d) the domestic risk premium differential ( r0i ) for country i. Having established this basic framework, we can use it to understand different ways that government policies can affect the stock market, and also to see how it will be affected by foreign shocks. In terms of policy, we can distinguish alternatives that may affect the value of firms directly, and others that may have an impact indirectly, probably through general equilibrium channels. In terms of the former, a first obvious way for the government to increase the value of firms in the stock market is to decrease profit taxes. We can interpret (1 t ) more broadly as measuring the respect for private property rights by the government and the quality of the corporate governance environment. Another possibility is for the government to implement policies that increase productivity. This can in principle be done in various ways (e.g. when the government subsidizes research and development activities in the private sector or through government agencies dedicated to these tasks). Another, albeit less direct, possibility to increase productivity is for the government to invest in infrastructure (e.g. more and better roads will likely increase the productivity of transportation companies, etc.). Some policy alternatives have a direct impact on one sector but maybe not on the other. For instance, trade protection polices will likely increase the domestic value of industrial firms in the tradable sector. Also, an example that may increase the value of firms in the non-traded sector is when the government implements a plan to renew the infrastructure (e.g. build new roads) using private firms as contractors. Government policies may also affect the value of domestic firms through more indirect channels. For instance, policies aimed at reducing fiscal deficits and government debt problems may have a positive impact on the stock market if these policies are able to lower the domestic component
4

To ease the notation, equation (1) assumes that the relevant discount rate in future periods is just the compound of the actual rate, or that the rates in the denominator are long term rates as the notation generally used by market practitioners. Of course, a more detailed notation should allow for discount rates to vary over the forecasting horizon. However, given that our goal here is to lay down a simple conceptual framework, this simplification is enough to convey our message.

of the discount rate. In addition, some policy actions can have general equilibrium effects on the exchange rate, favoring one sector over the other, in which case the final effect on the aggregate stock market index will depend on the sectorial composition of the index ( 0 ) . For instance, it is likely for an increase in government expenditure to lead to a real exchange rate appreciation (especially if the change is perceived as persistent). Such an appreciation will be reflection of the increase in the relative price of non-tradables in terms of tradables ( ptNT / ptT ), which will raise the value of firms in the former sector and reduce it in the latter. The effect on the national index will then depend on whether it is composed primarily by traded or non-traded goods firms. 5 Finally, notice that what matters to determine the sectorial impact of a policy is the effect on the real exchange rate. This in particular implies, in terms of the effect of monetary policy in the stock market, that while a monetary-induced change in the nominal exchange rate can have a direct impact on the stock market, the final effect will depend on whether this nominal change is perceived as implying a real depreciation or a real appreciation in the future. We can also use the framework to analyze how foreign factors may affect the national index. First, factors that tend to lower both world and Latin-American interest rates will tend to spur the domestic stock market. The former can be generated, for instance, by an expansionary monetary stance in developed countries, while an example of the latter can be a wave of capital inflows to emerging countries and/or an increase in investor risk appetite for these markets. Another channel through which factors may have a domestic impact is through the price of tradables. Clearly, and increase in these world prices will likely increase the value of firms in that produce tradable goods. Moreover, if this change is perceived as persistent, the increase in tradable prices might generate a real appreciation through its wealth effect; in which case firms is the non-traded sector should be benefited as well. 6 Finally, given that productivity has a world component, technological improvements in the rest of the world that can be reproduced domestically can also affect the domestic index.

2.3. Related Literature


This paper is related to different strands of the literature. In the first place, there are numerous studies that attempt to distinguishing between good policy and good luck focus on the countrys performance in terms GDP growth. One of the seminal papers in this literature is the work by Easterly et al. (1993), documenting how changes in growth rates tend to be more volatile that changes in institutions of a given country, while fluctuations in terms of trade seem to be more important to explain the variance of GDP growth. In this way, according the authors interpretation, good policy (i.e. institutional change) cannot be the main source of long-run growth, while good luck appears as a better explanation. Many studies have followed this work, trying to disentangle the relative role of good luck in explaining GDP fluctuations. Some
Some of the examples of polices highlighted in the preceding paragraphs can also have general equilibrium effects (for instance, it is likely for a policy that raises aggregate productivity to generate a real appreciation, due to the income effect that such a technological improvement will generate). 6 Notice that it might be the case that an increase in the price of some tradables and not in other may generate a detrimental effect on some tradable goods firms. Consider an increase in the price of an exportable commodity. While firms producing this commodity will benefit for this change, the general equilibrium effect (the real appreciations) can generate Dutch disease-style effects that can be detrimental to other exportable sectors (e.g. industrial goods). Therefore, as in the other cases previously mentioned generating non-neutral effects, the final change in the national index will depend on the weight of those negatively-affected firms in the aggregate index.
5

examples are Easterly et al. (1993), della Paolera et al. (2003), Jadresic and Zahler (2000), and Cecchetti et al. (2011). 7 Given that good luck is generally due to external factors, this paper is also related with those studies aiming at identifying the relevance of foreign variables in explaining business cycles fluctuations in small open economies. Some of these studies use time series models (e.g. vector auto-regressions) to identify foreign shocks. An example using LA data is Hoffmaister and Roldos (1997), estimating that terms-of-trade and foreign interest rate shocks can explain near 30 percent of output fluctuations. A more recent example is the work by Boschi and Girardi (2011) estimating that foreign factors explain between 20 and 40 percent of output fluctuations, depending on which LA one is looking at. They also decompose the foreign factors into regional and global components, finding that the former explain almost 2/3 of the contribution of foreign shocks. In this same line, other studies use dynamic and stochastic general equilibrium models to asses the role of foreign factors. For instance, Mendoza (1995) argues that more 50 percent of output fluctuations in emerging countries are due to terms of trade shocks. Neumeyer and Perri (2005) and Uribe and Yue (2006) analyze the importance of world interest rates shock, concluding that near 20 percent of output variance is due to world interest rate shocks. Our work is also related with the literature analyzing capital flows to and from emerging countries. An early example of this line of research is the work by Calvo, Leiderman and Reinhart (1993), while a more recent review can be found in Reinhart and Reinhart (2008). This literature identifies pulling and pushing factors to explain capital flows. The former are related with improvements in the receiving country that make foreign investment more attractive, while the later are used to describe events in the developed word that make investments in more industrialized countries less attractive and thus push capital away from those countries to more emerging economies. One of the stylized facts described in this literature is that a significant part of the inflow part of the capital cycle is explained by external conditions that make emerging countries relatively more attractive (i.e. pushing). At the other end, when capital flows away from emerging countries, how much capital leaves a country compared to others is mostly determined by domestic factors. Finally, it is important to mention some studies that have analyzed the impact on the stock market of particular policy announcements, generally using event-studies tools. For instance, Henry (2000) analyzes the impact of stock market liberalization policies, finding that these announcements produce abnormal returns of more than three percent per month in real dollar terms. Andritzky et al (2007) study the impact of different macroeconomic news for bond prices in emerging countries, finding that policy announcement have little effects of those prices, while they tend to reduce volatility. A final recent example is the case of the work by Moser and Rose (2011) who analyze the impact of regional-trade-agreements announcements in stock market indices.

2.4. Possible Limitations


There is a related literature that tries to explain the phenomenon denominated the great moderation, trying to disentangle the role of good policy and good luck in explaining the reduction in volatility of aggregate variables that occurred since the 1980s, mainly focusing on the U.S. (e.g. Gali and Gambetti, 2007).
7

While we have highlighted the merits of looking at stock prices to disentangle good policy from good luck, it should be noticed that our approach has some shortcomings as well. One of the concerns is that financial markets are not as deep in LA countries as in other developed economies, which may have several consequences for the analysis. First, firms in the stock market generally represent a small percentage of the total value of the stock of capital in the economy (although financial markets tend to develop over time). This is obviously a limitation but it is less clear how it should bias our results. Second, many times these markets are relatively illiquid compared to international standards. This would likely be more problematic (i.e. inefficient pricing) for intraday analysis, but is probably not that important given that we will be working with month-end prices over 30 years. Third, LA governments many times have imposed restrictions for international transactions (capital controls), which in principle it is not clear how they would affect the performance of the stock market. 8 Another source of potential problems with our measures is that the anticipation of changes in administration may attribute credit to the wrong president. Finally, one can think of example in which GDP might give a better proxy that stock prices. For instance, if households are liquidityconstrained, a high GDP growth may help families to invest in human capital and so seize long term opportunities that would not materialize if it did not invest. In this sense, a GDP-based measure would better capture aggregate wealth created than a measure based on stock prices. In conclusion, our approach (as any proxy of an ideal but unfeasible object) has its own limitations. However, we consider that given that ours is an unexplored methodology and that, as we highlighted, can overcome many of the drawbacks of other alternatives, it is worth to explore what stock market data has to say about these issues.

3. Data and Methodology9


The key variable of interest is the value-weighted national stock market return of each country. For the purpose of this project, we think that the sensible unit in which to measure of returns is a basket of local consumption goods. Our object of interest is the value of the capital stock as expressed in terms of the amount of consumption goods that local residents could purchase with it. 10 Our benchmark specification relies on the Standard and Poors/IFC Global and Investable indices, which are value-weighted. We use the Global indices since the start of the sample until they became discontinued in October 2008 and the Investable indices from then until December

One argument can be that restriction on international financial transactions will decrease the value of a stock, for instance, because the international demand for domestic stocks is eliminated. However, during some episodes of severe capital controls, the stock market may serve as a device to take money out of the country (see, for instance, the analysis of Auguste et al, 2002, for an analysis of the Argentine case during 2001-2002), rising stock prices. 9 The Data Appendix contains a detailed list of sources. 10 To the extent that there may be distributional effects within a country, it should be emphasized that when we refer to country conditions we mean from the point of view of investors. Unless the distributional aspects of policy are dominant, the effects of policy for investors and non-investors should be positively correlated. We refer to net, as opposed to gross, effects because some policies could benefit some corporations while harming others (e.g. international trade liberalization).

2010. 11 The Global indices, which are available since 1975, comprise a larger number of stocks, but some of these stocks are only available for purchase by local investors due to constraints on foreign investment. They seem the appropriate object of interest since we are measuring value creation in terms of local consumption goods, which is tantamount to assuming that the marginal investor is a local one. The Investable indices are available since 1993 and include fewer stocks since they exclude from the Global indices those shares that cannot be purchased by foreign investors. Given that some countries suffered prolonged periods of high (or even occasional episodes of hyper) inflation during the sample period, we started with the S&P/IFC index returns expressed in US dollars which we found to be more reliable than those expressed directly in local currency. We then converted these into US dollars, and then to domestic currency index values in nominal terms. To this end we used the free market exchange rates that resulted from cross checking among four sources: the S&P/IFC Emerging Markets Data Base itself, the series in the World Currency Yearbook, Economatica and International Financial Statistics of the IMF. 12 Through this process we corrected a few errors in the original local currency returns series of the S&P/IFC indices (e.g. as the Brazilian return in July 1994 when a new currency was introduced). We next converted these nominal price series into real price series by dividing the local price at the end of a given month by one plus the average consumer price index variation during the following three months as reported by the national statistical agencies. 13 With a price series expressed in units of domestic consumption basket, we next computed the monthly log returns thereof, Rti . In this paper, we seek to gauge the change in value of the capital stock during a given administration. We therefore use log returns because in OLS estimation the coefficient estimates on dummy variables will be the arithmetic averages of the residual during a specific part of the sample. The residual in turn, will be measured in the same units as the dependent variable. Arithmetic averages of returns are well known to overestimate the actual cumulative returns obtained by stock holders in the past. By contrast, if the return series is expressed in logarithmic units, its arithmetic average turns out to the geometric average of the original returns series. In other words, the OLS coefficient of a dummy variable will correctly report the cumulative return obtained by equity holders during a given part of the sample and controlling for other determinants of total returns. Besides providing equity returns, S&P/IFC provide the names of the actual stocks in each index, their industries, and the fraction of the index that they represent at the end of each month. The industry classification is done at eight digits following the Global Industry Classification Standard (GICS). In some cases in the early years, this shows that only a handful of stocks make up the national index. To seek a minimum level of diversification, we incorporate each country in the sample in the month in which it first appears with more than 10 stocks in the Global index. This date of entry into the sample is December 1975 for Chile and Mexico, November 1981 for Brazil, December 1984 for Colombia and Venezuela, January 1985 for Argentina, and December

Venezuela is an exception since S&P ceased computing equity returns indices for this country in March 2007. Economatica (http://www.economatica.com) is a leading provider of Latin American stock market data. 13 This admittedly crude measure of expected inflation allows avoiding important jumps in prices in high-inflation or hyperinflation contexts in which the consumer price index takes a few months to show the hike of some prices which are affected by government decisions (e.g. public utility prices or the price of gasoline).
12

11

1992 for Peru. The last observation for all countries is December 2010, with the exception of Venezuela that ends in March 2003. As it is standard in asset pricing papers, we need to control for global conditions as reflected in the global market portfolio. A change in such conditions could affect the value of the capital stock in a given country for reasons unrelated to regional or local shocks. In performing such a comparison, it is important to bear in mind that the industrial structure represented in emerging stock markets is quite volatile both across countries and within countries over time. 14 In order to control for global conditions while taking into account the change in industrial structure, for each country-month in the sample, we compute the return on a specific global portfolio that has the same industrial structure as that same country at the end of the previous month (i.e. the twin index). More specifically, we use the industry weights vector of each national S&P/IFC index at the end of month t-1 to multiply the vector of global returns by i industry during month t, and so obtain the return on twin index, which we denote Rtw . Global returns of each industry are taken from the U.S. dollar-denominated Datastream Global Equity Indices. The industrial classification structure is based on the Industry Classification Benchmark (ICB) at four digits i.e. a slightly coarser categorization than that provided by S&P/IFC for the national indices. In order to compute real returns of these global indices we used the US consumer price index. We matched GICS and ICB codes using the description of each industry in the respective definitions. As shown in Figure 1 (already discussed in the Introduction), this turns out to matter substantially to interpret the source of stock market movements in the countries in the sample. Given this two series of returns, the dependent variable in our estimations is the monthly difference between the local return and the twin index return for that month, which we call i i Z ti = Rti Rtw . Table 1 displays the mean and the standard deviation of Rti , Rtw and Z ti for each country in the whole sample. [Table 1] At this point it is worth highlighting that in our benchmark specification the local return is measured in terms of local consumption units while the return on the twin index is measured in terms of U.S. consumption units. We choose to do this mainly because we want to avoid declaring that a given administration generated much welfare during a certain amount of time just because it allowed the value of the local currency to appreciate. However, it is not clear in principle, which rout one has to take. To see this, notice that the return in a given country i measured in terms of consumption units of country i, which in logs we can call Rti ,i , differs from
14

Emerging countries are noteworthy for having a much less diversified production structure than developed ones. Moreover, the depth of financial market is much more limited than that of developed markets and this is especially reflected in the listing of publicly traded equity securities. For example, Figure A.1 in the appendix shows that the telecommunications sector was 24% of the Brazilian market in 1995 but only 2.5% in 2010. By contrast, in Mexico, this sector was 20% in 1995 and still 34% in 2010. As another example, the financial sector went from being 40% of Brazilian market cap in 1985 to 12% in 1995 and back to 22% in 2010. In the Mexican index, the importance of the financial sector is relatively much more stable with 5% of market cap in 1985, 10% in 1995 and 9% in 2010. The contrast is even more striking in the other five countries in the sample.

10

the return in that same country measured in consumption of the US, Rti ,US , by the bilateral real depreciation rate, rerti ,US ; this is, Rti ,i Rti ,US = rerti ,US . 15 Suppose then a period in which the rate of return in US consumption units of the twin portfolio of country i equals the return in i country is consumption units of the local portfolio, i.e. Rti ,i = Rtw . In this case, our measure of excess return Z ti will equal zero. If, instead, we measure the domestic portfolio in US, the excess return will equal the real appreciation, i.e. Rti ,US Rtw = Rti ,US Rti ,i = rerti ,US . Thus, in this case a real appreciation of the domestic currency would be taken to be a positive National contribution. Alternatively, imagine a period in which the rate of return in US consumption units of twin portfolio of country i equals the rate of return in US consumption units of the local i portfolio, i.e. Rti ,US = Rtw . If we were measuring country is return in US consumption units, there would be no local contribution, but if we were measuring the country is returns in local consumption units (as in our benchmark case) we will declare a positive local contribution when there is a real devaluation. Therefore, given these differences that may arise, we will report the results with the benchmark specification (local returns in local consumption units and twin portfolios in U.S. consumption units) and then include in the robustness section the results using U.S. consumption units to deflate local returns.
i

Another important source of (empirically measurable) exogenous variation in stock returns that we control for are the countrys terms of trade (TOT). In particular, we include as a control variable the first difference of the TOT. It should be noticed, however, that we face a data limitation with the frequency of the TOT series, that for some countries it is only quarterly available. Therefore, we will present two different sets of results; one based on monthly regressions without controlling for TOT, and another one using quarterly data that includes TOT. 16 There could be many other potential sources of exogenous variation in stock returns that are unrelated to current or expected future government policy measures. One way to control for the shocks that are common to the region and not reflected in the returns of the global portfolio is to put a time effect in the empirical model, and we use a yearly specification for this. The coefficient on each of these dummy variables will capture the average common variation in excess returns for that year. This could reflect changes in investor appetite for emerging market risk, capital flows into or away from Latin America, or the average effect of similar policy measures implemented by the governments of the seven countries in the region. Our estimation strategy is implemented in two parts. In a first stage, we include a dummy variable for each country and run the following regression,
Z ti = Rti Rtw = +
i

2010

=1976

I (t ) + i I i (t ) + TOTt i + ti
i =1

Notice that if relative PPP were to hold, measuring returns in local or foreign consumption goods will be the same. While it is probably not reasonable to assume that PPP will hold in the short run, given that we will be interested in the average return during a presidential mandate of around four years, relative PPP might not be a bad approximation. 16 The quarterly observation of a monthly variable was constructed by averaging the three monthly observations in the particular quarter.

15

11

where t denotes the time period and i refers to the country, I (t ) is a dummy variable that takes the value of one if the month t belongs to the year , and therefore is a fixed effect per year.

I i (t ) is a dummy variable that takes the value of one when the observation corresponds to the country i; thus i is a fixed effect for said country. 17 We also assume that the incidence of TOT in the dependent variable is the same for all countries, an assumption that we relax in the robustness section.
The equation is estimated by OLS, which yields consistent estimates under the assumption of i small-open economies (i.e. Rtw and TOTt i are strictly exogenous), and compute robust standard errors. 18 This type of model has been frequently used in the finance literature to identify industry effects in major equity markets, as well as country and regional effects. A seminal example is the work of Heston and Rouwenhorst (1994), while some more recent cases are Baca, Garbe, and Weiss (2000), and Brooks and Del Negro (2004). In a second step, we define dummy variables for the presidents who were in office during the sample. The dummy for each president equals one in the month in which s/he took office, and it remains on until the last full month in office. We took the swearing-in dates from the official web sites of the different presidencies. 19 The model to be estimated is therefore,
Z =R R
i t i t wi t

= +

I (t ) +

=1976 i =1 j i =1

2010

Ji

i ji

I j (t ) + TOTt i + ti .
i

I j (t ) is a dummy variable that takes the value of one when the president j on country i was in office during the month t; thus ij i is a fixed effect for said administration. 20 Importantly, we
i

constrain the coefficients , and to take the values that were estimated in the equation of the first step; for the regression in the first stage is the natural candidate to estimate the regional contribution (i.e. the coefficient for the time dummy. Using the coefficients from these two regressions, we decompose the return in the domestic stock market in a given period as follows. First, we identify the Global contribution as the constant plus the return on a country-mimicking global portfolio plus the change in TOT in that period multiplied by its estimated coefficient. The Regional contribution is the year effect identified for the whole region. Finally, the National contribution is the effect of the particular i president. In this way, suppose that the president was in office during the months t = t0 ,..., T i , the
17

For both

and

i we impose that the sum of the coefficients should be equal to zero. In this way, the base i.

category for interpreting the fixed effects coefficients will we either the average year for the average country in the case of
18

, or the average country in the case of

It is important to remember that the starting dates differ across countries, and therefore this is an unbalanced panel. 19 Table A.1 and the Appendix shows the list of presidents for each country, as well as the dates determining the beginning and the end of their mandate. 20 Here we also impose that the sum of the coefficients should be equal to zero.

12

Global, Regional and National contributions for country i under that administration are, respectively, +
i =t0

Ti

Rtw (t ) + i T i t0
i

TOTt i i T i t i , =t 0 0
Ti

i =t 0

Ti

i 0

, and ij i .

4. Results
We present the results in four parts. The first two subsections are related with the Global, Regional and National decomposition: the first one compares countries in the whole sample without making a difference for the particular administration (the first stage), while the second one focuses on presidents. In these subsections, results are presented in scatter plots with the following format. 21 On the horizontal axes we measure the sum of the Global and Regional contributions for the given unit of comparison (country or minister), and in the vertical axes we record the sum of the Global, Regional and National contribution. In this way, the vertical distance between a given data point and the 45-degree line will measure the differential return that can be attributed to the particular unit of comparison. In addition, with the exception of figure 4.a., each graph will contain two data points for each unit of comparison, one that corresponds to monthly regression that omits TOT (red dots) and the other one is produced with the quarterly regression that takes into account TOT (blue triangles). The third subsection discusses a battery of robustness check that we have analyzed. The final subsection takes the estimated decomposition for presidents and tries to asses whether these different components are correlated with alternative measures of approval rates of the governing administration.

4.1. Comparing Countries


Using the results from the first stage regressions (i.e. that including one dummy for each country in the sample) we begin by characterizing the estimated contributions of external factors. Figure 2 plots the evolution of Regional contribution and the simple average (across countries) of the Global contribution, and the sum of both.22 We can see from the figure that the Regional contribution has been generally much more volatile than the Global contribution. Moreover, the extreme troughs of the Regional contribution can be identified with crises hitting the region as a whole: the debt crisis in 1982, the Tequila in 1994 and the Asian crisis and the Russian default in 1997. Moreover, we can see how this decomposition assigns the world shock suffered in 2008 to global factors and not the Regional contribution, which is reasonable given that (for the first time in many years) the recession was due to problems in the developed world. [Figure 2] Having described the evolution of the Global and Regional contributions, we turn to the comparison of each countries performance, without making a distinction for each president, which is reported in Figure 3. Focusing first on the horizontal axes, we can see that the country
21 22

The results of the estimated coefficients were included in the Appendix (Tables A.2) to save space. These are estimated from the monthly regression that do not include the terms of trade. The Global contribution that we estimate is actually a monthly series, but we include in this figure only the return per year just to simplify the exposition.

13

that had the best luck on average during the sample was Colombia, followed by Argentina. At the other end, Venezuela was the country facing the worst external conditions. [Figure 3] In terms of the National contribution, the vertical distance between each point and the 45-degree line shows that the countries with better national contributions are Peru and Chile. Colombia also had a positive National contribution, and also Brazil although much closer to the 45-degree line. In contrast, Argentina, Mexico and Venezuela had a negative National contribution on average, the latter being the one with worst performance. Finally, the figure also shows that taking into account the TOT does not change the results significantly. In particular, it seems that controlling for TOT matters relatively more for the inference about the Global and Regional Components than for the measure of the national component. Overall, controlling for TOT assigns worst external conditions but a higher contribution of the National component for Colombia, Peru, Mexico and Venezuela, better Global and Regional conditions but a smaller National component for Brazil and Chile, and worst external and domestic conditions for Argentina.

4.2. Comparing Presidents


We next turn to the analysis of the performance of different presidential administrations. 23 Before turning to the description of the results for each country, it is instructive to have first an overall picture. The first graph in Figure 4 identifies the points corresponding to each president for all the countries in our sample, for the case in which TOT are not included. In line with the country comparison, we can see here that average president experienced positive external conditions. If we group presidents by decades, we can see how almost all presidents during the 80s and the 2000s experience positive external conditions (Global plus Regional), while most of the negative values are for presidents in the 90s. In terms of the National contribution, there is no clear clustering of observations. The only thing to notice is that the dispersion of the National contribution is smaller in the 2000s than in the previous decades. [Figure 4] After this general picture, we turn to analyze presidents performance in each country. Starting with Argentina, all the presidents have faced favorable external conditions, with the exception of De la Rua, which was mainly due to a negative Global contribution. The president facing better external conditions was Nestor Kirchner, followed by Alfonsin and Menems first mandate,24 while Duhalde, 25 Menems second mandate and Fernandez de Kirchner face intermediate values of luck. In terms of the National contribution, Duhalde, Alfonsin and Menems second mandate had positive values. Fernandez de Kirchner and Nestor Kirchner had negative but mild National
In general, controlling for TOT does not change the results significantly. Therefore, while we report these results in the graphs, we will generally omit to comment on them; unless there is a case in which a significant difference is estimated. 24 It is important to keep in mind that in the case of Alfonsin our sample starts in January 1985, impling that we are missing the performance in the first two years of his mandate. 25 As we will discuss in the robustness section, Duhalde is one of the few cases in which the results are significantly changed if we measure the domestic return in terms of US consumption units.
23

14

contributions, while Menems first mandate and particularly De la Rua display important negative National contributions. In terms of Brazil, while six of the eight presidents experienced good external conditions the other two faced either null or negative Global and Regional contributions. In terms of the National contribution, Franco is actually the president in the whole sample with a larger value (more than 50%), and Figuieredo and Cardosos second term had positive value. Both terms by Lula and Cardosos first terms had almost null National contributions, while Sarney and Collor exhibit significantly negative values. Turning to the case of Chile, al presidents faced positive external conditions, Aylwin being the one with best luck, followed by Piera and Pinochets first mandate. At the other enbd, Frei Ruiz-Tagle faced the worst external conditions, followed by Bachelet and Pinochets second mandate. The National contribution is in positive values for both of Pinochets mandates and for Bachelet. Piera and Frei had almost zero national contributions, while Lagos and Aylwin had negative records. In the case of Colombia, with the exception of Samper, all presidents experienced good external conditions. In fact, Betancur is the highest value for Global plus Regional contributions in our sample (two thirds of which is due to Global factors). In terms of the National contribution, Barco, Uribes first mandate and Gaviria had positive values, Uribes second mandate had and almost null contribution, while Bentacur, Pastrana and Samper display negative values. In terms of Mexican presidents, all but one (Zedillo) experienced positive Global and Regional contributions. As we can see, the inference about Echeverria is the one in the sample that displays the largest difference when TOT are considered: without TOT, the external conditions are in mid-ranges and the National contribution is negative, while if we control for TOT good luck appears to be a much relevant factor and the National contribution is almost nil. Dela Madrid appears as having a significant National contribution, while Fox, Calderon and Salinas had almost zero National contributions. Zedillo and Portillo are in the negative range. For Peru, all the presidents in the sample faced positive external conditions, expect for Fujimoris second term that has values near zero. In terms of the National contribution, Alan Garcia and Fujimoris first term had positive values, while the second term of Fujimori, Paniagua and Toledo exhibit negative contributions. Finally, in Venezuela only Caldera Rodriguez faced negative external conditions. However, not surprisingly given the results in the country comparison, all presidents displayed negative values for their National contribution.

4.3. Robustness
In this subsection we check the robustness of our results in the benchmark estimation to a number of alternative specifications. 26 A first exercise consists in expressing the domestic returns in terms of U.S. consumption units. As we commented before, it is not clear what is the
For the sake of space, we have included all the figures of these alternative exercises in a separate appendix that is available from the authors upon request.
26

15

appropriate way to proceed (in the benchmark specification, we might be spuriously favoring a president that implemented a large real depreciation, while if we measure everything in terms of U.S. consumption units someone that implemented a large appreciation will look good in our metric). In terms of the countries comparison, results do not change significantly, which is consistent with our previous claim that if relative PPP were to hold (which is not unreasonable while averaging over more than 30 years) no differences should be observed. In terms of presidents, most of them do not change significantly, so we just comment on those that change form one specification to the other. For Argentina, the biggest change is for Duhalde, that in the benchmark was on the upper-right quadrant and significantly above the 45-degree line and now is on the lower-left quadrant below the 45-degree line (closer to De a Rua). 27 In addition, all other presidents seem to have experienced larger Regional contributions, and the National contribution for Fernandez de Kirchner is now marginally above the 45-degree line. In Brazil, Figueiredos Regional contribution is smaller in this specification, placing him in the negative range of the horizontal axes. The same pattern appears for Cardosos second term, while Lulass first term displays a positive national contribution. For Chile, Pinochets second mandate has a significantly worst Regional component, placing him in values close to zero in the horizontal axis. In addition, Piera appears as having a better Regional contribution and above the 45-degree line. In the case of Colombia, Uribes second term appears above the 45-degree line, while Pastrana displays now a negative combination of Global and Regional conditions. For Mexico, Echeverria is now in the lower-right quadrant, while Dela Madrid faced a milder Regional component. In Peru, the only difference is that Paniagua appears with a positive National contribution. Finally, for Venezuela, the significant change is that Chavezs first term appears in the negative values of the horizontal axes, and closer to de 45-degree line. Returning to the returns measured in domestic consumption units, a first check was to include a dummy for hyperinflation episodes, 28 for it is likely (at least conceptually) for the stock market to experience a boom during such an episode. However, the results do not seem to change significantly controlling for hyperinflations. Another exercise is relates to the TOT coefficient; which in our benchmark case is restricted to be the same for all countries. Allowing for this coefficient to vary across countries does not change significantly the results, with some exemptions. In Brazil, Franco and both of Lulas terms display a more favorable sum of Global and Regional contributions. For Chile, Piera appears above the 45-degree line. Finally, for Venezuela, Chavez displays a value in the horizontal axis closer to zero. A final exercise consists in eliminating from each presidential dummy the last six months of its administration. 29 The goal of this alternative is to control for the possibility that at the end of the
This was expected given that during Duhaldes presidency the convertibility plan was abandoned with a devaluation of close to 200% in nominal terms. Moreover, Duhaldes periods was relatively short (2 years) which tends to exacerbate the difference. 28 Defined according to Cagans classical definition "beginning in the month the rise in prices exceeds 50 percent and ending in the month before the monthly rise in prices drops below that amount and stays below for at least a year." The resulting episodes are: Argentina 5-1989 to 3-1990, Brazil 12-1989 to 3-1990 and 6-1994, and Peru 91988 and 7-1990 to 8-1990. 29 We constructed a new dummy for each country with these dates that we have taken away from each president to preserve the country fixed effect specification.
27

16

mandate of a president the stock market might move according to the expectation about what would happen with the next president. 30 For Argentina, De la Rua appears with a better but sill negative National contribution, while Duhalde is closer to the 45 degree line. Alfonsin turns from a positive negative contribution to negative one, and Nestor Kirchner appears closer to the 45degree line. In Brazil, Collor, Francom Cardoso and Lulas second term appears with a better National contribution. In Chile, Pinochets first mandate has a larger National contribution, while the value for Aylwin is even more negative. A similar difference appears in Colombia (Uribes first mandate more positive and Bentacur more negative). For Mexico, none of the presidents change significantly, while in Peru only Fujimoris second term appears now above the 45degree line. Finally, in Venezuela both Perez Rodriguez and Lusinchi move at value of National contributions close to zero.

4.4. Good Luck, Good Policy and Approval Rates


Given our decomposition of the domestic return into Global, Regional and National, we would like to have a metric that allows us to say to what extent is this decomposition relevant (in other words, why should be care about this decomposition?). We tackle this issue by assessing whether these different components are correlated with measures of presidential approval rates. We collected data from different opinion surveys for Argentina, Brazil, Chile, Colombia and Peru. 31 Using these, we created a quarterly series of approval rates for each country. 32 For each of these, we ran three regressions. Each of them shares two regresores: a constant and a dummy variable in the quarter the president takes office. 33 In the first regression, we also include the total return of the previous quarter (this columns is labeled with (a) in the table). In the second and third regressions, instead of using the total return we use the Global contribution, the Regional contribution, and the difference between the total return and these two, all of them for the previous quarter. The difference between these two sets of regression (labeled, respectively, (b) and (c) in the table) is that the former one uses the results based on the model without TOT and the latter uses the decomposition obtained controlling for TOT. The goal of this exercise is to see if, in the cases in which the total return is a significant predictor of approval rates, our decomposition can explain whether that correlation is due to Global, Regional or National factors. It is important to notice that, a priori, it is not clear what type of results we should expect. If our measure of good policy is related to the public perception about the policies implemented by the government, then we should expect the National contribution to be highly correlated with approval rates. Indeed, this would probably be the desirable results (i.e. people preferring presidents that implement policies that are perceived as good). However, as it is clear from this work, separating good luck from good policy is quite difficult, so if people cannot disentangle
However, we do not attribute these last six month to the next president because, to be fair, no policies had actually been implemented. Moreover, some times in our sample presidents have to leave office early (or to call elections sooner) because the country is in a crisis, so it would not be fair either to attribute the last months to the next president. 31 For Argentina and Chile we have two sources available. The data appendix describes the sources and which question in the survey was used. We also have surveys available for Mexico and Venezuela, but with just a few observations and therefore we do not report the results based on these sources. 32 Some of the series have monthly data and others are quarterly. For those that we have monthly data, we just took the simple average of the observations in the particular quarter. 33 We included this dummy because approval rates might jumps when a new president starts its mandate.
30

17

both of them they may approve a government just because they confuse good luck for good policy. Moreover, in terms of incentives from the political sector, generally the governing party has incentives to convince the public that a good economic development is due to good policy, in the same way that they will try to blame bad luck every time the country experiences a poor aggregate performance. On the contrary, parties outside the government will likely try to convey the opposite message. Another motivation for this analysis is based on the literature that analyzes the impact of economic conditions in the results of elections. A seminal contribution in this line of research is Fair (1978, 1996), while more recent examples are Lewis-Beck and Stegmaier (2000) and Brender and Drazen (2008). The key message of this line of research is that current economic conditions affect the probability of winning an election, highlighting for example that the most important economic variable is the growth rate of output per capital in the year of the election. However, in this literature the distinction between good policy and good luck is not analyzed, but it is interesting to see if the voting behavior simply depends on the outcome or if it determined by a more thorough exercise where people try to determine if the incumbent government have implemented good polices or not, regardless of the outcome. The results are presented in Table 2. In each table we first report the mean and standard deviation of the particular approval series that we use, then we show the regression results, and finally we report the estimated effect of a one (annualized) standard deviation of each of the return measures on the approval rate. [Table 2] Starting with the first available survey for Argentina, we can see that the total return ha a significantly and positive correlation with the approval rate. In particular, if the (annualized) return increases by one standard deviation, the approval rate will significantly increase by almost 9 points. This is a big figure given that the average of this approval rate is close to 40. Once we look at the decompositions, no matter whether we control for TOT or not, the Global contribution and, to a less extent, the Regional contribution display positive and significant correlations with the approval rate, while the National contribution is insignificant. To quantify this result, an annualized shock of one standard deviation in the Global component increases the approval rate by 11 points, while this figure is around 6 point for the Regional contribution. Moreover, the results based on the other approval rate that we have available for Argentina shows similar results. For the case of Brazil, neither the total return nor the decomposition has significant effect to predict approval rates. For Chile, the first survey shows that the total return has a positive and significant correlation with the approval rate (the effect of a one-standard-deviation shock is more than 6 points), while in terms of the decomposition only the Regional Contribution seems have a significant effect (the effect of a one-standard- deviation shock is around 6 points). The other survey we have available for Chile shows similar results, with a slightly larger effect for a shock to the regional contribution (more than 7 points). Using the survey we have available for Colombia, we can also see that the total return is positively correlated with the approval rate (the effect of a one-standard-deviation shock is close to 12 points), while in the decomposition what seems to matter is the Global contribution (the 18

effect of a one-standard- deviation shock is around 11 points). In the case of Peru, the total return is not significantly correlated with the approval rate; while this is the only case in which the National contribution is significant (the effect of a one-standard-deviation shock is close to 7 points). Finally, we have also ran a battery of robustness exercises for these approval regressions, that are also available in the separate Appendix, all of them giving qualitatively similar results. In particular, we first used the returns lagged two or three quarters, to consider the case that economic developments may take longer to affect approval rates. Second, we consider the same exercises but using the return expressed in terms of US consumption units. Finally, with the concern that the measures of presidential approval might be persistent, we redo the original exercise but grouping the observations by semesters. However, results appear to be robust to these changes as well.

5. Conclusions
We presented a methodology that allows decomposing the performance of stock market returns in seven LA countries into Global, Regional and National components. We argued that this is an appropriate approach to separate good policy from good luck given the forward-looking nature of assets prices. This decomposition allowed comparing the different administrations that were in place in the last 25 years, evaluating not only their differential contribution but also describing whether they were favored by international conditions or not. In particular, it seems that the good performance of LA economies in the last 5-10 years was in many cases more related with good luck than with good policy. In fact, our model explains an average return (across years and across countries) of 15.4% during the last decade; which can be decomposed as 6.7% due to the Global contribution, 8.6% due to the Regional contribution, and 0.1% due to the National contribution. While any such decomposition is itself arbitrary to a certain extent, we later ask if it can be useful in explaining peoples opinion about presidential administrations. The estimation highlights that in many cases good luck appears to be relevant in explaining approval rates. We think that these results have relevant implications to understand the relationship between economic outcomes and political opinions and voting. In the same way that economic crises may lead to the resignation of economic ministers and even presidents (which has not been uncommon in LA history), good economic outcomes is many times used as a campaign slogan; either for a party seeking another term, or as an argument of a party that was in power some prosperous bygone times. If we think that good policies are country-invariant, and that opposed policies are all accompanied by good outcomes during a given time period, then it cannot be the case that they result from the policies themselves. It is clear that good outcomes do not necessarily result from good policies. Still in some cases it seems to be the case that popular approval rates results from outputs not inputs. Therefore, we consider that an important line of future research should be aimed at better understanding the mechanics that link people perception of economic outcomes and the political preferences.

19

References
Andritzky, J., G Bannister, and N. Tamirisa, 2007, The impact of macroeconomic announcements on emerging market bonds, Emerging Markets Review, vol. 8(1), pages 20-37, March. Auguste, S., K. Dominguez, H. Kamil, and L. Tesar (2002), Cross-Border Trading as a Mechanism for Capital Flight: ADRs and the Argentine Crisis, NBER Working Paper, No. 9343. Baca, S., B. Garbe, and R. Weiss (2000), The Rise of Sector Effects in Major Equity Markets, Financial Analyst Journal, Vol. 56, No. 5, Sep. Oct, 34-40. Brender, A. and A. Drazen, 2008, How Do Budget Deficits and Economic Growth Affect Reelection Prospects? Evidence from a Large Panel of Countries, American Economic Review, 98(5): 220320. Boschi, M. and A. Girardi, 2011, The contribution of domestic, regional and international factors to Latin America's business cycle, Economic Modelling, vol. 28(3), pages 1235-1246, May. Brooks. R., and M. Del Negro (2004), International Stock Returns and Market Integration: A Regional Perspective, Working Papers Series, Federal Reserve Bank of Atlanta, 02-20b. Calvo, G, L. Leiderman and C. Reinhart, 1993, Capital Inflows and Real Exchange Rate Appreciation in Latin America: The Role of External Factors, IMF Staff Papers, vol. 40(1), pages 108-151, March. Cecchetti, S., M. King, and J. Yetman (2011), Weathering the Financial Crisis: Good Policy or Good Luck? mimeo, Bank for International Settlements. della Paolera, G., M. Bozzoli, and C. Irigoin (2003) Passing the Buck: Monetary and Fiscal Policies in Argentina: 1852-2000, in della Paolera and Taylor (eds.), A New Economic History of Argentina, Cambridge University Press. Easterly, W., M. Kremer, L. Pritchett and L. Summers (1993) Good policy or good luck?, Journal of Monetary Economics, Volume 32, Issue 3, December, 459-483. Easterly, W., N. Loayza, and P. Montiel (1997), Has Latin America's post-reform growth been disappointing?, Journal of International Economics, vol. 43(3-4), 287-311. Fair, R. (1978), The Effect of Economic Events on Votes for President, Review of Economics and Statistics, vol. 60, pp. 159-73. Fair, R. (1996), Econometrics and Presidential Elections, Journal of Economic Perspectives, 10 (3), pp. 89-102. 20

Frankel, J. (2010), The Natural Resource Curse: A Survey, NBER Working Paper, No.15836. Gal J. and L. Gambetti (2009), On the Sources of the Great Moderation, American Economic Journal: Macroeconomics, vol. 1(1), pages 26-5. Henry , P., 2000, Stock Market Liberalization, Economic Reform, and Emerging Market Equity Prices, The Journal of Finance, Vol. 55, No. 2, pp. 529-564. Heston, S. and G. Rouwenhorst (1994), Does Industrial Structure Explain the Benefits of International Diversification? Journal of Financial Economics, 36(1), 3-27. Hoffmaister, A.W. and Roldos, J.F., 1997. Are business cycles different in Asia and Latin America? IMF Working Paper, 97/9. Jadresic, E. and R. Zahler (2000), Chile's Rapid Growth in the 1990s-Good Policies, Good Luck, or Political Change? IMF Working Paper, No. 00/153. Lewis-Beck. M. and M. Stegmaier, 2000, Economic Determinants of Electoral Outcomes, Annual Review of Political Science, 3: 183-219. Mendoza, E., 1995, The Terms of Trade, the Real Exchange Rate and Economic Fluctuations, International Economic Review, 36(1), February, 101-137. Moser, C. and A. K. Rose, 2011, Who Benefits from Regional Trade Agreements? The View from the Stock Market, NBER Working Paper, No. 17415, September. Neumeyer, P. and F. Perri, 2005, Business Cycles in Emerging Economies: The Role of Interest Rates, Journal of Monetary Economics, 52, March, 345-380. Reinhart, C. and V. Reinhart, 2008, Capital Flow Bonanzas: An Encompassing View of the Past and Present, NBER Working Paper, No. 14321, September. Uribe, M. and V. Yue, 2006, Country Spreads and Emerging Countries: Who Drives Whom? Journal of International Economics, 69, June, 6-36.

A. Data Sources
Data Appendix for 7 countries - Argentina, Brazil, Chile, Colombia, Mexico, Peru and Venezuela. Global Return of each industry: U.S. dollar-denominated total and global monthly Return Index (RI) for Datastream Global Equity Indices from 01/01/1976 to 01/01/2011. The stocks are classified by industry and sector types for example financials is an industry within which a number of sectors are included such as banks, life assurance and real estate. The classification

21

structure is based on the Industry Classification Benchmark (ICB). Datastream Global Equity Indices break down into six levels. We took account RI at level 4. Equity Returns: U.S. dollar-denominated monthly S&P/IFCG Index Return for S&P Emerging Market Indices from January 1976 to October 2008 and from November 2008 to December 2010, we use S&P/IFCI Index Return for S&P Emerging Market Indices. CPI: Argentina: Instituto Nacional de Estadsticas y Censos (INDEC) up to September 2006 and Buenos Aires City from October 2006; Brasil: Fundacao Instituto de Pesquisas Economicas (FIPE). Chile: Instituto Nacional de Estadsticas (INE). Colombia: Banco de la Repblica de Colombia. Mxico: Banco de Mxico. Per: Instituto Nacional de Estadstica e Informtica (INEI). Venezuela: Banco Central de Venezuela. U.S.A: Economic Research, Federal Reserve Bank of St. Louis. Exchange Rate: S&P/IFC Emerging Market Data Base. We corrected an error in the original series of Brazil in which did not account the new currency introduced in 1994. Risk Free Rate: Year Treasury Constant Maturity Rate up to12/31/1981 and 3-Month Treasury Constant Maturity Rate from 01/29/1982. Both sources are Economic Research, Federal Reserve Bank of St. Louis. Terms of Trade: Argentina: quarterly data from Instituto Nacional de Estadsticas y Censos (INDEC) (base 1993=100); Brazil: Instituto de Pesquisa Economica Aplicada (IPEA) monthly data (base average 2006=100); Chile: http://www.bcentral.cl/eng/studies/workingpapers/pdf/dtbc98.pdf monthly data (base January-1977=100) up to 1995 and from 1996 quarterly data from Banco Central de Chile (base 2003=100); Colombia: monthly data from Banco de la Repblica de Colombia; Mexico: monthly data from Banco de Mexico (base 1980=100); Peru: monthly data from Banco Central de Reserva del Per (base average 1994=100); and Venezuela: quarterly data built (base 1990Q1=100) by Alejandro Izquierdo (IADB). Presidents and Ministers Argentina Presidents: http://www.casarosada.gov.ar Brasil. Presidents:http://www.presidencia.gov.br; Chile. Presidents:http://www.memoriachilena.cl/historia/presidentesdechile.asp, http://es.wikipedia.org/wiki/Presidente_de_Chile. Colombia. Presidents: http://es.wikipedia.org/wiki/Anexo:Presidentes_de_Colombia. Mexico. Presidents: http://www.bicentenario.gob.mx/index.php?option=com_content&view=article&id=637 http://es.wikipedia.org/wiki/Anexo:Presidentes_de_M%C3%A9xico . Peru. Presidents:http://es.wikipedia.org/wiki/Anexo:Presidentes_del_Per%C3%BA.

22

Venezuela. Presidents: http://es.wikipedia.org/wiki/Presidente_de_Venezuela Approval Measures Argentina - Gallup: Question: Taking into account the current economic situation, do you approve or disapprove the government performance in terms of its economic management? Series: Percentage responding yes. Argentina - IPSOS: Question: How do you evaluate the performance of the national government in terms of the economic policy? Do think is very good, good, bad or really bad? Series: Percentage responding very good or good. Brazil Datafolha: Question: In your opinion, the current presidents performance is optimal, good, regular, bad or abysmal? Series: Percentage responding optimal or good. Chile CEP: Question: Do you approve or disapprove the way the president is managing the government? Series: Percentage responding approve. Chile CERC: Question: Do you approve or disapprove the performance of the current government? Series: Percentage responding approve. Colombia Gallup: Question: Do you approve or disapprove the way the president is performing his presidential duties? Series: Percentage responding approve. Peru IPSOS: Question: Do you approve or disapprove the performance of the current president? Series: Percentage responding approve.

23

B. Tables
Table 1. Descriptive Statistics of Annual Returns and Excess Returns by Country since index inception until December 2010. (A) National Index (Real)
Mean Argentina Brazil Chile Colombia Mexico Peru Venezuela 13% 15% 18% 19% 11% 18% 1% St.Dev 63.5% 50.1% 28.1% 26.7% 34.3% 28.9% 41.7%

(A) Twin Index (real)


Mean 9% 9% 7% 11% 7% 10% 6% St.Dev 18.2% 17.4% 15.0% 16.4% 14.8% 22.1% 16.4%

(A) - (B)
Mean 4% 5% 11% 9% 4% 9% -2% St.Dev 61.9% 48.1% 29.1% 28.5% 31.4% 24.7% 42.8%

24

Table 2.a. Determinants of Approval Rates.


Argentina - Gallup (a) (b) (c) Argentina - IPSOS (a) (b) (c) Brazil - Datafolha (a) (b) (c) Chile - CEP (b)

(a) (c) Descriptive Stats.Approval Mean 39.5 29.8 38.1 57.1 St.Dev. 17.4 18.5 18.6 8.2 Regresion Results Total Return 75.61** 80.20 -30.55 97.82** Global 206.7*** 222.5*** 252.2*** 256.0*** -1.542 13.64 60.35 58.30 Regional 108.8* 114.5* 169.8* 178.1* -43.48 -41.74 129.7** 110.6* National 0.159 0.651 -64.74 -55.57 -49.49 -69.30 26.89 -10.64 New President -0.983 -1.568 -2.415 2.754 1.163 0.148 -2.747 -3.618 -4.157 -3.780 -3.939 -3.710 Constant 39.22*** 37.53*** 37.55*** 31.99*** 29.77*** 29.91*** 38.84*** 38.91*** 38.93*** 48.57*** 48.39*** 48.75*** Observations 82 82 82 50 50 50 69 69 69 45 45 45 Adjusted R-squared 0.0428 0.0653 0.0772 0.0147 0.0789 0.0805 -0.0181 -0.0474 -0.0393 0.0554 0.00726 -0.0006 Effect of 1 S.D. Total Return 8.8 9.3 -3.7 6.6 Global 11.1 12.0 13.6 13.8 -0.1 0.8 3.2 3.4 Regional 5.7 6.0 8.9 9.3 -2.3 -2.2 6.8 5.8 National 0.0 0.0 -4.5 -3.8 -4.0 -5.4 1.2 -0.5 *** p<0.01, ** p<0.05, * p<0.1

25

Table 2.b. Determinants of Approval Rates, Cont.


Chile - CERC (b) Colombia - Gallup (a) (b) (c) Peru - IPSOS (b)

(a) (c) (a) (c) Descriptive Stats.Approval Mean 50.1 59.1 43.2 St.Dev. 11.0 21.9 19.6 Regresion Results Total Return 52.47 133.9** -8.569 Global 49.68 42.26 209.3* 222.8* -15.96 -20.96 Regional 141.5** 142.8** 154.3 145.6 -103.2 -112.1 National -46.43 -43.85 95.00 97.48 123.2* 124.4* New President 1.271 2.271 1.715 7.901 10.33 10.15 17.36** 11.72 11.90 Constant 56.97*** 56.34*** 56.50*** 57.31*** 55.50*** 55.49*** 40.41*** 40.67*** 40.63*** Observations 54 54 54 58 56 56 71 71 71 Adjusted R-squared 0.0121 0.141 0.133 0.0410 0.0200 0.0226 0.0154 0.0356 0.0371 Effect of 1 S.D. Total Return 3.5 11.8 -0.7 Global 2.6 2.5 10.7 11.2 -1.0 -1.4 Regional 7.4 7.5 8.1 7.6 -4.6 -4.9 National -2.1 -2.0 7.2 7.4 7.2 7.2 *** p<0.01, ** p<0.05, * p<0.1

26

C. Figures
Figure 1.a

Argentina's Performance: 2003-2010


6
Base: May 2003=1

5
Real Price in Local Consumption Units

0 2003

2004

2005

2006

2007
P_World

2008

2009
P_Twin_Arg

2010

P_Argentina

Figure 1.b.

Brazil's Performance: 2003-2010


6
Base: Jan-2003=1

Real Price in Local Consumption Units

0 2003 2004
P_Brazil

2005

2006

2007
P_World

2008

2009
P_Twin_Brazil

2010

27

Figure 2.

Returns per year due to Global and Regional Contribution


100% 80% 60% 40% 20% 0% -20% -40% -60% -80% 1976 1980 1984 1988 1992 1996 2000 2004 2008

Regional

Average Global

Regional + Average Global

Figure 3

Global, Regional and National Contribution. 1976-2010 (in % points per year)
25% Peru Colombia

Global + Regional + National

20%

Chile Brazil

15% Argentina 10% Mexico

5%

Venezuela

0% -5% -5% 0% 5% 10% 15% 20% 25%

Global + Regional

28

Figure 4.a.

Global, Regional and National Contributions. All Presidents together over decades (in % points per year)
80%

Global + Regional + National

60%

40%

20%

0% -20% -10% -20% 0% 10% 20% 30% 40% 50% 60% 70%

-40%
80s

Global + Regional
90s 00s

Figure 4.b.

Global, Regional and National Contributions. Argentina, 1985-2010 (in % points per year)
50% 5_Duhalde 1_Alfonsin 6_Kirchner_Nestor 3_Menem_II 2_Menem_I 10%
7_F_de_Kirchner_Cristina

Global + Regional + National

30%

-20%

-10% -10%

0%

10%

20%

30%

40%

50%

-30% 4_De_la_Rua -50%

Global + Regional

29

Figure 4.c.

Global, Regional and National Contribution Brazil, 1981-2010 (in % points per year)
90%

5_Franco

Global + Regional + National

70%

50%

2_Figueiredo
30%

8_Lula_I 7_Cardoso_II

10%

6_Cardoso_I
-20% -10% -10% 0%

9_Lula_II
10%

3_Sarney
20% 30%

4_Collor

40%

50%

-30%

Global + Regional

Figure 4.d.

Global, Regional and National Contribution. Chile, 1976-2010 (in % points per year)
80%

Global + Regional + National

60%

1_Pinochet_I
40%

7_Pinera 3_Aylwin

20%

2_Pinochet_II 6_Bachelet 5_Lagos

0% -10% 0%4_Frei_Ruiz 10% 20% 30% 40% 50%

-20%

Global + Regional
30

Figure 4.e.

Global, Regional and National Contribution. Colombia, 1985-2010 (in % points per year)
70%

Global + Regional + National

50%

3_Gaviria 2_Barco 6_Uribe_I

1_Bentacur

30%

7_Uribe_II
10%

-20%

-10% -10%

0%

10%

20%

30%

40%

50%

60%

70%

80%

5_Pastrana

4_Samper
-30%

Global + Regional
Figure 4.f.

Global, Regional and National Components Mexico, 1976-2010 (in % points per year)
40%

3_Dela_Madrid

Global + Regional + National

4_Salinas 6_Fox
20%

7_Calderon 1_Echeverria
0% -10% 0% 10% 20% 30% 40%

-20%

5_Zedillo 2_Portillo
-20%

Global + Regional

31

Figure 4.g.

Global, Regional and National Contribution Peru, 1993-2010 (in % points per year)
40%

1_Fujimori_I

Global + Regional + National

30%

5_Alan_Garcia_II

4_Toledo

20%

10%

3_Paniagua
0% -20% -10% -10% 0% 10% 20% 30% 40%

2_Fujimori_II

-20%

Global + Regional
Figure 4.h.

Global, Regional and National Contribution Venezuela, 1985-2003 (in % points per year)
40%

Global + Regional + National

1_Lusinchi
20%

2_Perez_Rodriguez_II

0% -10% 0%

5_Chavez_I 10%

20%

30%

40%

3_Velazquez 4_Caldera_Rodriguez
-20%

Global + Regional

32

D. Appendix Tables and Figures


Figure A.1.a. Industrial Composition of the Domestic Stock Market: Argentina, Brazil, Chile and World.
Argentina
1990 1995 2000 2005 2010 1990 1995

Brazil
2000 2005 2010 1990 1995

Chile
2000 2005 2010 1990 1995

World
2000 2005 2005 2010 2010

100%

80%

60%

40%

20%

0% Energy Health Care Materials Financials Industrial Info. Technology Consumer Discret. Telecomm. Consumer Staples Utilities

Figure A.1.b. Industrial Composition of the Domestic Stock Market: Colombia, Mexico, Peru and World.
Colombia
1990 1995 2000 2005 2010 1990 1995

Mexico
2000 2005 2010 1990 1995

Peru
2000 2005 2010 1990 1995

World
2000

100%

80%

60%

40%

20%

0% Energy Health Care Materials Financials Industrial Info. Technology Consumer Discret. Telecomm. Consumer Staples Utilities

33

Figure A.1.b. Industrial Composition of the Domestic Stock Market: Venezuela and World.
Venezuela 1990 100% 1995 2000 2005 2010 1990 1995 World 2000 2005 2010

80%

60%

40%

20%

0% Energy Health Care Materials Financials Industrial Info. Technology Consumer Discret. Telecomm. Consumer Staples Utilities

34

Table A.1. List of Presidents


Mandate Begins Ends 10/12/1982 7/7/1989 8/7/1989 8/6/1995 8/7/1995 12/09/1999 12/10/1999 12/31/2001 01/01/2002 05/24/2003 05/25/2003 12/09/2007 12/10/2007 Present 03/15/1979 03/14/1985 03/15/1985 03/14/1990 03/15/1990 12/28/1992 12/29/1992 12/31/1994 01/01/1995 12/31/1998 01/01/1999 12/31/2002 01/01/2003 12/31/2006 01/01/2007 12/31/2010 01/01/1976 03/10/1981 03/11/1981 03/10/1990 03/11/1990 03/10/1994 03/11/1994 03/10/2000 03/11/2000 03/10/2006 03/11/2006 03/10/2010 03/11/2010 Present Mandate Begins Ends 07/08/1982 08/06/1986 08/07/1986 08/06/1990 08/07/1990 08/06/1994 08/07/1994 08/06/1998 08/07/1998 08/06/2002 08/07/2002 08/06/2006 08/07/2006 08/06/2010 01/01/1976 11/30/1976 12/01/1976 11/30/1982 12/01/1982 11/30/1988 12/01/1988 11/30/1994 12/01/1994 11/30/2000 12/01/2000 11/30/2006 12/01/2006 Present 07/28/1990 07/27/1995 07/28/1995 11/21/2000 11/22/2000 07/27/2001 07/28/2001 07/27/2006 07/28/2006 Present 02/02/1984 02/01/1989 02/02/1989 05/20/1993 06/05/1993 02/01/1994 02/02/1994 02/01/1999 02/02/1999 Present

President
Alfonsn, Ral Ricardo Menem, Carlos Sal Menem, Carlos Sal De La Rua, Fernando Duhalde, Eduardo Alberto Kirchner, Nstor Carlos Fernndez de Kirchner, Cristina Figueiredo, Joo B. Sarney, Jos Collor, Fernando Franco, Itamar Cardoso, Fernando Henrique Cardoso, Fernando Henrique Lula Da Silva, Luiz Inacio Lula Da Silva, Luiz Inacio Pinochet Ugarte, Augusto Pinochet Ugarte, Augusto Aylwin Azcar, Patricio Frei Ruiz-Tagle, Eduardo Lagos Escobar, Ricardo Bachelet Jeria, Michelle Piera Echenique, Sebastian

Country
Argentina Argentina Argentina Argentina Argentina Argentina Argentina Brazil Brazil Brazil Brazil Brazil Brazil Brazil Brazil Chile Chile Chile Chile Chile Chile Chile

President
Bentacur, Belisrio Barco, Virgilio Gaviria, Csar Samper, Ernesto Pastrana, Andrs Uribe Velez, lvaro Uribe Velez, lvaro Echeverria Alvarez, Luis Portillo, Jos Lopez De la Madrid, Miguel Salinas de Gortari, Carlos Zedillo Ponce de Len, Ernesto Fox Qeusada, Vicente Calderon Hinojosa, Felipe Fujimori, Alberto Fujimori, Alberto Paniagua Corazao, Valentn Toledo Manrique, Alejandro Garca Perez, Alan Lusinchi, Jaime Perez Rodriguez, Carlos Andrs Velazquez, Ramn Jos Caldera Rodriguez, Rafael Chavez Fras, Hugo Rafael

Country
Colombia Colombia Colombia Colombia Colombia Colombia Colombia Mxico Mxico Mxico Mxico Mxico Mxico Mxico Per Per Per Per Per Venezuela Venezuela Venezuela Venezuela Venezuela

35

Table A.2.a. Estimation Results: Cross Country Comparison: 1976-2010


Model With Country Dummies VARIABLES without TOT with tot _1976 0.00605 0.0163 (0.0210) (0.0205) _1977 0.0388* 0.0408** (0.0204) (0.0204) _1978 0.0224 0.0216 (0.0239) (0.0240) _1979 0.0399* 0.0369* (0.0204) (0.0207) _1980 -0.00973 -0.00896 (0.0192) (0.0190) _1981 -0.0634*** -0.0616*** (0.0160) (0.0161) _1982 -0.0486** -0.0474** (0.0220) (0.0219) _1983 0.00863 0.00837 (0.0211) (0.0213) _1984 -0.00505 -0.00448 (0.0220) (0.0221) _1985 0.00917 0.00249 (0.0206) (0.0128) _1986 0.000672 0.00891 (0.0158) (0.0162) _1987 -0.0267 -0.0270 (0.0197) (0.0195) _1988 -0.00437 -0.00586 (0.0162) (0.0162) _1989 -0.0130 -0.0118 (0.0269) (0.0268) _1990 0.0329 0.0314 (0.0267) (0.0265) _1991 0.0608*** 0.0619*** (0.0183) (0.0183) _1992 -0.00866 -0.00897 (0.0141) (0.0141) _1993 0.00195 0.00322 (0.00994) (0.00994) _1994 0.000633 -0.00181 (0.0114) (0.0116) _1995 -0.0229** -0.0225** (0.00963) (0.00963) _1996 -0.000315 -0.00151 (0.00714) (0.00707) _1997 0.00438 0.00513 (0.00762) (0.00760) _1998 -0.0522*** -0.0499*** (0.0135) (0.0135) _1999 0.00786 0.00415 (0.00902) (0.00895) Model With President Dummies without TOT with tot 0.00700 0.00881 (0.0387) (0.0409) ar_2_menem_i -0.0138 -0.0137 (0.0280) (0.0279) ar_3_menem_ii 0.00695 0.00777 (0.0106) (0.0106) ar_4_de_la_rua -0.0238 -0.0344 (0.0233) (0.0235) ar_5_duhalde 0.0178 0.0191 (0.0320) (0.0321) ar_6_kirchner_nestor -0.00582 -0.00478 (0.00796) (0.00797) ar_7_kirchner_cristina -0.00245 -0.00224 (0.00994) (0.0101) br_2_figueiredo 0.0241 0.0240 (0.0234) (0.0234) br_3_sarney -0.0203 -0.0209 (0.0267) (0.0267) br_4_collor -0.0311 -0.0310 (0.0401) (0.0399) br_5_franco 0.0438* 0.0424* (0.0244) (0.0247) br_6_cardoso_i 0.00109 0.000805 (0.0138) (0.0138) br_7_cardoso_ii 0.00642 0.00358 (0.00855) (0.00885) br_9_lula_ii -0.000284 -0.000231 (0.00727) (0.00744) ch_1_pinochet_i 0.0193 0.0259 (0.0161) (0.0164) ch_2_pinochet_ii 0.00599 0.00549 (0.00907) (0.00910) ch_3_aylwin -0.00529 -0.00454 (0.0117) (0.0119) ch_4_frei_ruiz -0.000181 -0.000327 (0.00754) (0.00757) ch_5_lagos -0.00547 -0.00885* (0.00501) (0.00524) ch_6_bachelet 0.00722 0.00732 (0.00884) (0.00908) ch_7_pinera 0.00131 -0.00187 (0.0198) (0.0197) co_1_bentacur -0.0150 -0.0171 (0.0142) (0.0136) co_2_barco 0.0182 0.0202* (0.0111) (0.0110) co_3_gaviria 0.00891 0.00873 (0.0154) (0.0155) VARIABLES ar_1_alfonsin

36

_2000 _2001 _2002 _2003 _2004 _2005 _2006 _2007 _2008 _2009 _2010 argen brazi chile colom mexic peru venez delta_tot Constant

-0.0078602 (.0078663) 0.00498 (0.00887) 0.0132 (0.00913) 0.00421 (0.00558) 0.00104 (0.00661) 0.00882 (0.00675) -0.000638 (0.00645) -0.0114* (0.00599) 0.00214 (0.00957) 0.00203 (0.00807) 0.00425 (0.00646) -0.0024704 (.0088448) 0.000793 (0.00672) 0.00383 (0.00406) 0.00226 (0.00472) -0.00182 (0.00473) 0.00434 (0.00459) -0.00694 (0.00777)

0.00513* (0.00299)

-0.0088314 (.0078274) 0.00723 (0.00901) 0.0105 (0.00924) 0.00287 (0.00559) 0.000316 (0.00660) 0.00744 (0.00678) -0.00280 (0.00656) -0.0122** (0.00599) 0.00486 (0.00971) -0.00104 (0.00842) 0.00226 (0.00653) -0.0027912 (.0084207) 0.000576 (0.00672) 0.00373 (0.00407) 0.00245 (0.00470) -0.00180 (0.00469) 0.00455 (0.00460) -0.00670 (0.00807) 0.0796* (0.0475) 0.00521* (0.00296)

co_4_samper co_5_pastrana co_6_uribe_i co_7_uribe_ii me_1_echeverria me_2_portillo me_3_dela_madrid me_4_salinas me_5_zedillo me_6_fox me_7_calderon pe_1_fujimori_i pe_2_fujimori_ii pe_3_paniagua pe_4_toledo pe_5_alan_garcia_ii ve_1_lusinchi ve_2_perez_rodriguez_ii ve_3_velazquez ve_4_caldera_rodriguez ve_5_chavez_i otros_presidente_co otros_presidente_ve delta_tot

-0.0136 (0.0105) -0.00813 (0.0130) 0.0117 (0.0101) -0.00150 (0.00994) -0.00746 (0.0171) -0.0158 (0.00993) 0.00989 (0.0181) -0.000831 (0.00910) -0.00486 (0.00900) 0.00185 (0.00543) -0.00102 (0.00581) 0.0197 (0.0185) -0.00435 (0.00836) 0.000285 (0.0260) -0.00150 (0.00698) 0.0102 (0.0102) -0.00640 (0.0132) -0.0109 (0.0218) -0.0269 (0.0274) -0.00990 (0.0166) -0.000788 (0.0135) -0.000125 (0.0235) 0.0175*** (0.00254)

Observations 2,238 2,206 Observations R-squared 0.035 0.040 R-squared Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1

2,238 0.011

-0.0130 (0.0105) -0.0126 (0.0133) 0.0129 (0.0101) -0.000657 (0.0101) 0.000754 (0.0171) -0.0153 (0.00997) 0.0120 (0.0179) -0.000743 (0.00914) -0.00716 (0.00907) 0.00251 (0.00543) 9.34e-05 (0.00580) 0.0201 (0.0185) -0.00600 (0.00857) -0.00503 (0.0268) -0.00187 (0.00704) 0.0112 (0.0103) -0.000246 (0.0174) -0.0104 (0.0216) -0.0195 (0.0276) -0.00747 (0.0167) -0.0116 (0.0131) -9.09e-05 (0.0245) 0.0185*** (0.00296) 0.0849* (0.0471) 2,206 0.016

37

Das könnte Ihnen auch gefallen