EXTERNAL SHOCKS, DUTCH DISEASE AND INFORMALITY IN BOLIVIA

Rodrigo Gonzales Zuazo*1 CIESS ECONOMÉTRICA UNIVERSIDAD MAYOR DE SAN ANDRÉS September, 2016

Abstract 2014 marked the conclusion of the period of high commodity prices, a fact that has had a significant macroeconomic impact in Latin American countries. Most of the work focused on this situation has been concerned with explaining the role of the terms of trade in the business cycle. On the other hand, there are those who argue that one of the negative consequences of the external boom has been Dutch Disease. However, in many cases these latter authors have not considered important transmission channels in countries lacking a developed industrial sector and with high levels of informality in their labour markets, as in the case of Bolivia. The contribution of this paper focuses precisely on a better understanding of Dutch Disease in Bolivia in this context. JEL classification: E26, F16, F30, J40. Key Words: External shocks, Terms of trade, Dutch Disease, Informal labour market, Bolivia.

* This work was supported by funds from the Swiss Programme for Research on Global Issues for Development (r4d programme) under the thematic research module "Employment in the context of sustainable development" and the research project “Trade and labor market outcomes in developing countries”. The Swiss Program for Research on Global Issues for Development is being implemented jointly by the Swiss Agency for Development and Cooperation (SDC) and the Swiss National Science Foundation (SNSF). The views expressed here are the author’s and do not necessarily reflect those of the SDC or those of SNSF. All errors are the author’s responsibility. For any questions or suggestions, please contact the author [email protected].

1.- Introduction The end of the commodity price boom is a fact which has been causing serious challenges for Latin American economies, particularly in southern countries. The growth and macroeconomic stability achieved in recent years is now history, and the forecasts are not very encouraging given the international economic context 2. Policymakers are facing complicated scenarios with little space for economic policy, putting at risk many achievements in economic and social matters. Throughout the period of the external boom, Latin American economies have experienced strong inflows of foreign exchange. This has drawn the attention of economists to the presence of Dutch Disease (DD) phenomenon in the region and in specific economies. If this “evil” were to occur, the possible transition scenarios for various export and non-tradable sectors in terms of production, employment and remuneration should be very important issues for consideration by the authorities. This work makes two main contributions. The first is to present evidence supporting the presence of DD in Bolivia and countries with similar characteristics during the external boom period. The second is to analyse the development of DD according to the specific characteristics of each economy, focusing on the case of Bolivia. The first point is made using empirical evidence and recent works on Bolivia and other countries in the region. For the second, we provide a rigorous explanation of the dynamics of DD in Bolivia, in which we identify two fundamental characteristics: low industrial development and high levels of labour market informality3. We rely on the construction of a Structural Vector Autoregressive model (SVAR), which allows us to appreciate terms of trade shocks in small countries in the global markets of exported and imported goods and services, i.e., price-taking countries. The results of the empirical model support the DD transmission channels identified during the boom period. The structure of the document is as follows: in section 2, we develop the concept of Dutch Disease and explain why this phenomenon can be considered as an "evil". Section 3 is devoted to reviewing work supporting the presence of DD in several Some simulation exercises using a G-VAR (global VAR) before the new world order are featured in Powell (2016). Two possible scenarios are analysed: on the one hand, a weak growth of the developed economies and China, and on the other hand, possible scenarios of regional growth before the normalization of United States interest rates. 3 Morales et al. (2016) make a similar characterization of the Bolivian economy. 2

countries during the external boom period. In section 4, we explain the economic context of Bolivia during the boom and in section 5, we discuss the development of DD in Bolivia. In section 6, the SVAR model is presented and, finally, the conclusions. 2. Dutch Disease The term Dutch Disease (DD) was established by British magazine The Economist (1977) to explain the phenomenon of gradual deindustrialization that occurred in Holland in the sixties. The discovery of large gas reserves in the North Sea dynamized the sector, driving an export boom; this raised the country's income and the greater flow of foreign exchange caused the appreciation of its currency, the guilder. Several sectors were harmed by the real appreciated exchange rate. The high remunerations to labour and the attractive profits presented by the gas sector caused a migration of the factors of production to the buoyant sector; while one sector grew, the others became depressed. The phenomenon is conventionally associated with an export boom in nonrenewable or renewable natural resources, or both simultaneously, which alters the distribution of factors and the price of currencies. However, based on Edwards and Van Wijnbergen (1989) and Adam (2005), this "evil" could be caused by any factor that causes a sudden increase in the flow of foreign exchange, such as a rise in export prices, capital inflows, external transfers (remittances) or external aid (donations). From this perspective, DD is an extension of the resource curse4 argued by Sachs and Warner (1995): the paradox implies that countries with abundant natural resources tend to grow more slowly than other countries. The pioneering works on the modelling of this phenomenon were those of Corden and Neary (1982) and Neary and Van Wijnbergen (1986). The former developed a variant on Salter's (1959) dependent economy model, which divides aggregate output into two sectors: tradable and non-tradable goods. They introduced a positive shock into a tradable sector, inducing an export boom in the sector. Labour

In fact, as described in Barja et al. (2014), the literature in this field can be categorized into two areas of research: i) The first denominated by Frankel (2010) as the "Natural Resource Curse" that implies the negative consequences of the curse of natural resources, where EH, would only be one of the possible results, and ii) The second so called "Resource Rich Developing Country" (RRDC) are papers that argue a positive relationship between natural resource endowment and economic growth. 4

and capital move to the booming sector from traditional tradable production and services, generating deindustrialization in the country. Later, the formalization of Neary and Van Wijnbergen (1986) identified two main propagation channels in the development of DD using a general equilibrium model: the resource mobility effect and the expenditure effect. The resource mobility effect refers to the movement of factors of production from the export sectors not experiencing a boom to the production of natural resources used by the booming export sector. This is due to an increase in marginal salaries in the boom sector, that is, incentives are offered to labour and capital to migrate to the boom sector, causing a drop in supply to other sectors. This is understood as direct deindustrialization. On the other hand, the expenditure effect refers to the impact of higher foreign exchange flows on the increase in domestic absorption, leading to an appreciation of the real exchange rate. The greater flow of foreign exchange increases private government spending, the latter of which benefits from a higher income from direct revenues (taxes and royalties to the booming sector) and a higher volume of transactions in the economy. According to Frankel (2010), given a flexible exchange rate regime, a larger amount of foreign currency will appreciate the nominal exchange rate. If the exchange rate regime is fixed, the increased supply of money will generate pressure on domestic prices. In both regimes, the result will be the appreciation of the real exchange rate. In addition, the increase in domestic absorption also puts pressure on the price of non-tradables, and since the prices of other exportable goods (which are not part of the boom) are fixed in world markets, the real exchange rate is also appreciated in this manner. Marginal wages in non-tradable sectors rise, generating migration from nonbooming export sectors to non-tradable goods and services. The manufacturing sector is a victim of this. This decline in supply to the manufacturing sector as a result of increased spending is known as indirect deindustrialization. In addition, according to Adler and Magud (2013), governments tend to relax during the boom period as they overcome liquidity constraints and contract more debt. This increase in spending during the expansion phase of activity could generate procyclical destabilizing policies5. The procyclical management of fiscal policy in Latin American countries has once again become a topic of debate. Some works such as Frankel et al. (2013) and Vegh and Vuletin (2016) state that several Latin American countries have graduated from fiscal procyclicality. However, Adler and 5

Another undesirable effect of DD on the financial system could be the excessive expansion of domestic credit, which has the tendency to generate consumption bubbles and increase pressure on the price of non-tradable goods and services, appreciating the real exchange rate. An important question when studying the adverse effects of DD is: why might it be bad for a country to go through DD? Sala-i-Martín and Subramanian (2003) argue that the contraction of non-booming export sectors can be considered as an “evil” only if these sectors are shown to have "special" characteristics, for example, if the development of manufacturers in those sectors could generate positive externalities of the "learning-by-doing" variety. Regarding the harmful effects of DD, the literature also presents papers that have studied the link between real exchange rate misalignment (with respect to its fundamentals) and economic growth. Magud and Sosa (2010) argue that the presence of DD cannot be diagnosed from variations in the real exchange rate, but deviations from its fundamentals (since appreciations can be of equilibrium). The authors extensively review the literature and conclude that while much of the work supports the effects of DD being present in several countries6, there is no evidence that this "evil" impedes economic growth. 3. Evidence of Dutch Disease in Latin America The methodologies for studying the effects of DD vary between theoretical and empirical models. In general terms, there is evidence supporting the presence of DD in several developing countries, and particularly Latin America, during the external boom. Harding and Venables (2016) find robust evidence for DD in 41 countries, including Bolivia. Through a gravity model, they show crowding out effects in the production of export products through the export of natural resources when they are in boom. For developing countries, Lartey et al. (2012) estimate a GMM dynamic panel data model for 179 countries, and find that remittances have had an impact of resource mobility, as well as on spending. In addition, they conclude that these effects are greater in countries with a fixed exchange rate regime. Magud (2013) argue that this response has not been an institutionalized effort as an explicit economic policy objective; rather, it was unintentional thanks to the extraordinary resources of the boom. An evaluation for the Bolivian case can be found in Gonzales and Molina (2017). 6 Magud and Sosa (2010) argue that in most papers conclude that DD exists. They conclude that volatility and real exchange rate misalignment tend to reduce economic growth. In particular, overvaluations would have a negative effect by slowing growth, but there is no evidence of undervaluation.

Rajan and Subramanian (2011) find that flows of external aid in several developing countries have consistently generated competitiveness losses because they have appreciated the real exchange rate. Using the Kalman filter, Ismail (2010) finds evidence of DD in oil exporting countries when the price shock is permanent. More specifically for Latin America, Wong and Pretesky (2014) use panel data for ten countries in the region (including Bolivia) and estimating by fixed effects finds that net flows (exports, foreign direct investments and donations ) have led to a decline in the country's most important manufacturing industrial exports, through the overvaluation of the real exchange rate. Raich (2014) also presents evidence of the effects of DD in Latin America, with an emphasis on Argentina. Economy-specific works such as those by Sierra and Manrrique (2014), who estimate a panel data model for 63 Colombian industrial sectors using the GMM approach, find significant effects between industrial value added and the real exchange rate for 21 sectors, which are negative for 18 of these sectors. Also for Colombia, Goda and Torrez García (2015) estimate three ARDL models to verify that the external boom in mining and oil had a negative impact on the tradable - non-tradable goods production ratio, as well as on the share occupied by manufacturing in the country's total exports. In addition, these effects were deepened by the external funds received during the boom. For the case of Chile, García-Cicco and Kawamura (2015) calibrate a dynamic general stochastic equilibrium model (DSGE) for a small and open economy, where they evaluate three policy options to counteract the unwanted effects of increased income from the export of raw materials7: fiscal rules for Government spending, capital controls and taxes for domestic agents. The results are not definitive regarding the preferable policy, since the associated welfare of the agents, along with the question of whether or not these are Ricardian, will determine the optimal resource reallocation policy to counteract effects of DD. Using Bayesian methods and data for El Salvador, Acosta et al. (2009) estimate a DSGE and find that the effects of DD are generated, since higher remittances decrease the supply of labour and increase the consumption of non-tradables. An increase in the price of non-tradables appreciates the real exchange rate and leads to the relocation of workers in the tradable sector. These results are compared with This increase in income generates unwanted effects since its model incorporates Ricardian and nonRicardian agents on the one hand, and financial market frictions and "learning-by-doing" externalities on the other. 7

those of a Bayesian VAR, which points in the same direction. In its model, the effect of resource mobility is the main carrier of DD. Thus, there is considerable evidence supporting the presence of DD in both developing and Latin American countries during the boom period. Our intention so far has been to show how the extraordinary flow of foreign exchange in countries with similar characteristics to Bolivia, in terms of the effect of shocks on these countries as raw material exporters, has generated the DD phenomenon. Now, we focus on studying the case of Bolivia. 4. The Bolivian Economic Context During the external boom (2004-2014), Bolivia experienced good and historic macroeconomic indicators. High rates of economic growth (5% on average), controlled inflation, along with fiscal and trade surpluses. Good terms of trade could be seen in the growth of oil8, mineral and soy bean9 prices, which were and still are the country's main export products10. Figure 1 (from the figures appendix) shows the evolution of the price and volume indices of traditional exports (hydrocarbons and minerals), which have been increasing (with the exception of 2008-2009). Figure 2 illustrates the evolution of the export price index for minerals, hydrocarbons and non-traditional goods. The revaluation of the real exchange rate in the context of the external boom is shown in figure 311. Figure 4 shows the evolution of the value of exports and imports, where we see that the large increase in exports was accompanied by increasing imports, which has been particularly high in recent years. Figure 5, supports the higher concentration of exports on the one hand, and the greater diversification of imports on the other12. Figure 6 illustrates the evolution of consumer price indices, differentiated into tradable and non-tradable goods. We observe that the evolution of non-tradeable goods prices has been higher. Bolivia exports natural gas to Brazil and Argentina through the state company YPFB. However, the sale price fixed in the contracts is calculated by averaging barrels of oil traded on different world markets, so it is indexed in the same, but not a contemporary way. For more details on the price calculation methodology, consult Aguilar and Valdivia (2011). 9 Increasing trends, excepting the 2008-2009 period (international financial crisis) and 2014 (end of the boom). 10 According to INE's foreign trade statistics for 2014, these three export items represented 91% of the total exported value (FOB value in dollars). 11 The terms appreciation or revaluation mean that the exchange rate falls. However, since the nominal exchange rate in Bolivia has not changed since November 2011, we can state that the exchange rate regime is de facto fixed. 12 The entropy is the concentration index; while it increases there is more diversification. For its calculation, see appendix A. 8

Figure 7 shows the informality rate13 and unemployment rate. It can be seen that the informality rate increased as a result of the international financial crisis and the commodity "super-cycle", resulting in a drastic decrease in unemployment. Figure 8 shows the decreasing trends in the production ratios of manufactured and agricultural products in relation to mining and hydrocarbon production. Figure 9 also shows a declining ratio of manufacturing and agricultural production, but in relation to non-tradable goods and services production. All these are signs of DD in Bolivia. In terms of employment, figure 10 shows the decreasing percentage of the population employed in the agricultural sector and manufacturing industry. Figure 11, on the other hand, shows an increase in the percentage of the population employed in the non-tradable sector, as well as less important sectors of goods and services in the mining and quarrying sector. Finally, figure 13 illustrates a worrying situation, this being an unprecedented decrease in the quantities of textile exports in Bolivia. This is a clear example of DD in this sector. 5. Dutch Disease in Bolivia 5. 1. Evidence The works of Cerruti and Mansilla (2008) and Cerezo (2014) have rejected the presence of DD in Bolivia, based on the absence of real exchange rate misalignment with respect to its fundamentals14. In particular, Cerezo (2014) estimates cointegration equations in an attempt to test the four symptoms of DD identified by Oomes and Kalcheva (2007), namely: i) appreciation of the real exchange rate, ii) deindustrialization of the manufacturing sector, iii) price increases in the non-tradable sector and iv) increase of real wages in this sector. It concludes that there is no conclusive evidence for the period between 1993 - 2010. Furthermore, Baldivieso (2013) finds evidence of DD, demonstrated by a fall in national clothing production. On the other hand, Barja and Zavaleta (2016) and Barja et al. (2014) construct a computable general equilibrium model (CGE) based on the UDAPE social For the calculation of the informality rate, we made estimates with household surveys from the National Institute of Statistics (INE). See appendix A. 14 The papers agree that although there have been some years of misalignment, these have been due to external crisis factors and not caused by the external boom. 13

accounting matrix of 2006, which shows that the price boom in the hydrocarbon and mining sectors generates the effects of DD. This is because the real exchange rate is appreciated and production and employment falls in non-booming export sectors. The authors also demonstrate the possibility of counteracting these effects by institutionalizing a stabilization fund created in times when resources are abundant. In both studies, the boom in the hydrocarbon sector generates considerable opportunities for economic growth, which supports the literature (RRDC). Finally, there are studies that have considered informal labour markets as part of the DD manifestation process. This point forms the central contribution of this work. For Bolivia, we can mention the works of Andersen and Faris (2001), Lay et al. (2006), Mevius and Albarracín (2008), Barja et al. (2014) and Morales et al. (2016). Additional references that following the lines of the present work are Ballesteros (2016) and Arguello et al. (2015), both for the case of Colombia. 5. 2. The Dynamics of Dutch Disease in Bolivia In the previous section, we present empirical evidence supporting evidence for DD in Bolivia, and we will now explain in detail the transmission channels and dynamics of this phenomenon in the country during the external boom period. The triggers for the external boom were the high prices of the main export products15, which had a positive impact on three main aggregates: i) ii) iii)

The national income, due to the increase in exports, The supply of money through an increase in international reserves and Tax revenues through higher royalties and taxes, mainly in the hydrocarbons sector16. In addition, a higher level of aggregate activity in the fiscal sector increased the tax base, providing another channel through which tax revenue increased.

The highest national income: i) ii)

Increased the demand for goods and services with a particular emphasis on durable goods, such as construction. Increased imports, which became increasingly diverse and numerous. This fact is explained by the low levels of development in the national industry, that is, the national supply could not respond to the greater and

As explained in the previous section, for Bolivia these are mainly gas, minerals and soy. It is important to note that in 2006, hydrocarbons contracts were renegotiated between the Bolivian State and companies in charge of the exploitation, modifying greater percentages of oil revenue in the State’s favour, a fact popularly known as the "nationalization of hydrocarbons". 15 16

iii)

more diverse demand from families and companies (see the composition of imports in the following section) Increased the money demand.

The increase in the money supply: i) ii)

Caused interest rates to decline and, given the country's low level of stock market depth, real estate to become a very attractive investment activity. Did not translate directly into inflationary problems due to17: a) The increased demand for money. b) The countercyclical monetary policy applied by the Central Bank of Bolivia (BCB), which remained active thanks to the de-dollarization economic policy. c) A de facto fixed exchange rate (since November 2011) that brought low inflation with it.

The increase in tax revenues generated: i) ii) iii)

An increase in public investment, with special emphasis on the construction of new infrastructure. An increase in expenditure on services (non-tradable goods) by the public administration. A considerable increase in public employment.

All of the abovementioned factors, together with the fact that there were no significant advances in labour productivity18, led to the real revaluation of the exchange rate and the consequent loss of competitiveness19. Employment increased, mainly in the non-tradable goods and services sector (trade, construction, medium-sized and cooperative mining and services in general) and, in sectors with a large informal sector, employment was also increasing.

During the boom period, inflation remained in single-digit figures, except for the years 2007 and 2011, when rates of 11.73% and 11.85% were recorded. 18 See Canavire-Bacarreza and Rios-Avila (2015). 19 Stein et al. (2016) propose an adjustment to how the real effective exchange rate index (REER) is conventionally calculated, which has been based on weighting the share of trade occupied by each country. Their calculation of the real effective adjusted exchange rate (AREER) considers: i) The extent to which countries export to the same markets; ii) The extent to which countries have similar export baskets. For the case of Bolivia, they find that during the period from June 2014 to October 2015, the real appreciation according to the ARRER was more than 25%. 17

In sum, the dynamics of DD in Bolivia can be better understood when we take into account the high levels of labour market informality and the lag of the industrial sectors. 6. Structural Vector Autoregressive Model (SVAR) 6. 1. Model Formulation In this section, we explain the construction of the Structural Vectors Auto Regressive model (SVAR). The model will allow us to appreciate how a shock to the stochastic process of the terms of trade (TT)20 might or might not generate the movements of certain key variables in the dynamics of Dutch Disease in Bolivia, according to the transmission channels we have previously identified. Lanteri (2016) estimates a SVAR to test the presence of DD in Argentina, imposing long-term restrictions à la Blanchard-Quah. We start from the following SVAR specification based on Amisano and Giannini (1997) and Lütkepohl (2005), imposing short-term constraints on matrix B21 and 22 Α(𝐼𝐾 − Α1 𝐿 − Α𝐿2 − ⋯ − Α𝑃 𝐿𝑃 )𝑌𝑡 = Α−1 𝑢𝑡 = Β𝑒𝑡

(1)

We define the vector 𝑌𝑡 of 5 x 1, with the following variables: 𝑡𝑜𝑡𝑡 𝑟𝑒𝑟𝑡 𝑌𝑡 = 𝑒𝑛𝑡𝑡 𝑖𝑛𝑓𝑡 [ 𝑢𝑟𝑡 ] Where:

Unlike some studies that relate certain variables to the price of gas and/or minerals to test the presence of Dutch Disease in Bolivia, we use the terms of trade as a "trigger", as does Lanteri (2016), since it is the key relative price of international trade. 21 The development presented below follows the B model in both references. Also, note that by simplifying the exposure, the constant vector has been suppressed, but as will be seen later, no such vector is specified in the estimation of the SVAR model. 22 Where 𝑌 is a vector of variables with a dimension of K x 1. The matrices Α, Α , Α , … , Α and Β 𝑡 1 2 𝑃 are matrices of with 𝐾 x 𝐾 dimension. The coefficient matrix Α contemporaneously relates the system variables contained in the vector 𝑌𝑡 . The matrices Α1 , Α2 , … , Α𝑃 contain the coefficients of the 𝐾 lags specified in the system. 𝑢𝑡 is the empirical error vector in reduced form, with a dimension of 𝐾 x 1, with 𝑢𝑡 ~𝑁(0, Σ) and also 𝐸(𝑢𝑡 𝑢𝑠, ) = 0 for all 𝑠 ≠ 𝑡. On the other hand 𝑒𝑡 is the orthogonalized structural error vector, with a dimension of 𝐾 x 1 and therefore is 𝑒~𝑁(0, 𝐼𝐾 ), with 𝐸(𝑒𝑡 𝑒𝑠, ) = 0, for all 𝑠 ≠ 𝑡. Structural errors are related to reduced form errors through the coefficient matrix Β. In this way, we can accurately identify the system by imposing constraints on matrices Α and Β (which must be non-singular), which in turn will allow us to analyse the response of the system variables to a disturbance in the vector 𝑒𝑡 . 20

𝑡𝑜𝑡𝑡 , 𝑟𝑒𝑟𝑡 , 𝑒𝑛𝑡𝑡 , 𝑖𝑛𝑓𝑡 , 𝑢𝑟𝑡 are the terms of trade index, the real effective exchange rate index (REER), the import entropy23, the informal employment rate (informal rate) and the unemployment rate, respectively. All variables are expressed in logarithmic differences in relation to their trends. Using the cyclic component will allow us to observe how these variables react to their trends. The calculations and the treatment of all of the variables are explained in more detail in appendix A. In particular, through the estimation of the SVAR model, we are interested in understanding how the variables of the vector 𝑌𝑡 react to external shocks, represented by the TT. Usually in this type of specification for small and open economies, it is assumed that the country is a small player in the world goods markets on which it exports and imports, and therefore does not have the capacity to alter these prices unilaterally, and so it takes TT as given (Agénor and Montiel, 2015). In this way, we propose a univariate AR (1) process to describe the evolution of TT 24: ∗ 𝑡𝑜𝑡𝑡 = 𝑎11 𝑡𝑜𝑡𝑡−1 + 𝑏11 𝑒𝑡

(2)

Next, following Sims (1980), we define the Β matrix as a lower triangular one, and so the covariance matrix of 𝑢𝑡 will be a LU factorization result of a Cholesky decomposition, Σ = ΒΒ′ . Our specification is completed with the assumption made by Uribe and SchmittGrohé (2016) about the exogenous process of TT to explain external shocks in ∗ developing countries, which implies that: 𝑎1𝑗 = 0, ∀ 𝑗 = 2, … , 5. Since our interest lies in TT shocks, the order of the variables is now irrelevant and the SVAR is identified and can be estimated. 6. 2. Model Estimation and Results

The import entropy is a concentration indicator. For its calculation, see appendix A. This specification on the stochastic process of the terms of trade is supported by the LR likelihood ratio test, in which the unrestricted model defines that the terms of trade also depend on the lagged values of the other four variables of the vector 𝑌𝑡 , This null hypothesis is rejected with a Chi-square statistic value of 17.82 and four degrees of freedom. 23 24

Quarterly data were used for the period 2000-201425. This is an acceptable period over which to observe the effects of the export price boom and, in general, to appreciate the economic impacts of favourable terms of trade (TT) in the Bolivian economy26. The estimation of the SVAR equations using OLS allows us to find the univariate process parameters that follow the TT: 𝑡𝑜𝑡𝑡 = 0,85 𝑡𝑜𝑡𝑡−1 + 0,04 𝑒𝑡 𝑅 2 = 0,73

(4) (5)

∗ The stability of the autoregressive process is guaranteed by 𝑎11 < 1. Structural innovation can also be seen in the TT process, which is given by the coefficient 0.04.

These results are not very different from those found by Schmitt-Grohé and Uribe (2015). The authors estimate the same exogenous process for TT, with annual data for the period between 1980 and 2011 in 38 developing countries, including Bolivia27. ∗ They find values of 0.52 and 0.08, for 𝑎11 and 𝑏11 respectively, with a setting of 0.29, lower than that found in this work, which is 0.73. Using these results, we can empirically evaluate how the system variables respond to a structural TT shock. We do this with the impulse-response functions (IRF) presented in appendix B. For the IRFs, a transient shock with a structural standard deviation of 0.04 has been introduced to the TT equation. The confidence intervals were estimated at 90%, using the t-statistics generated by 2000 bootstrap replicates. It was chosen to work with a lag in the system, since the criteria of Akaike, Schwarz and Hannan Quinn suggested between one and two lags, it was decided to preserve parsimony. Note also that the vertical axis measures the percentage deviation in relation to the trend of the variables28. Figures 1B and 2B show the responses of the real effective exchange rate index and the entropy of imports respectively. We can see that the real exchange rate is appreciated, deviating about 2% from its trend and with significant effects for more than three years (12 periods)29. This is a characteristic symptom of Dutch Disease,

See appendix A for further details on the treatment of the data used. Although mineral prices have fallen since the middle of 2013, oil prices have declined since August 2014, given the lag in the formula for calculating the sale price of natural gas to Brazil and Argentina, the signs of aggregate deceleration are not highly significant during 2014. 27 See table 1 in Schmitt-Grohé and Uribe (2015). 28 This is because the variables are in logarithmic differences with respect to their tendencies. 29 In the FIR, the significance of the transient effect of the shock is evidenced by the confidence intervals, when these do not include the zero. 25 26

and tells us that when the country goes through favourable external shocks, reflected in the TT, it loses competitiveness30. It is also possible to observe that a positive TT shock leads to the diversification of imports. This last fact is explained by two factors: on the one hand, the classic income effect would be operating31 and on the other, given the country’s low degree of industrial development, this translates into the inability of domestic production to meet the agents’ growing demand for new goods as their level of income rises32. The effects on import entropy are also significant for more than three years. Next, we present the response of the rate of informality and unemployment to the same structural TT shock. These graphs are shown in figures 3B and 4B of appendix B, respectively. It can be observed that the informality rate diverges approximately 1.5% from its trend, due to a deviation of 4% in the TT trend. During the external boom, an increase in informality was observed, and this is a central transmission channel in the propagation of Dutch Disease in Bolivia, as explained above33. On the other hand, there are more than proportional responses to the unemployment rate. Faced with the 4% TT shock, the unemployment rate diverges about 5% from its trend, except that, unlike the other variables analysed, this response shows significant effects only over two years (eight periods). The Granger causality test is presented in appendix C. For all that has been argued so far, we should expect that the Granger causality exists from the terms of trade towards the other four SVAR variables. Here, it is worth reflecting on the determinants of the real exchange rate. There are several works for Bolivia on this topic34, most of which have used the BEER (Behavioural Equilibrium Exchange Rate) methodology. In simple terms, this is used In the Bolivian context, the Central Bank of Bolivia (BCB) has decided to keep the nominal exchange rate fixed since November 2011, so the real appreciation in recent years was due to differences between international and domestic prices. Given the devaluations of several countries in the region in the face of capital outflows, the management of the exchange rate in Bolivia has been a topic of debate in the current conjuncture (for example, see the press opinions of Morales (2015) and Bonadona (2016), and more formally see the work of Barja et al. (2014)). 31 For example, see Colque (2011). 32 This is further reinforced by the fact that the bulk of imports is concentrated in capital and intermediate goods, with a share of more than 70% of the total in June 2016. 33 Refer to section 5.2.- Dynamics of Dutch Disease in Bolivia. 34 For example, Aguilar (2002), Mendieta (2007), Cerruti and Mansilla (2008), Bello et al. (2010), Cerezo et al. (2010), Colque (2011), Cerezo and Salazar (2012) and Cerezo (2014), among others. 30

to estimate a reduced equation for the real exchange rate, which is usually a cointegration equation or a VEC, where the coefficient of the real exchange rate is normalized. Conventionally, the variables selected are the terms of trade, government expenditure (or budget balance), capital flows, trade liberalization and some measure of factor productivity (using the Harrod-Balassa-Samuelson effect). However, the results of the work are varied, but in general TT is a determinant of the real exchange rate; however, there are other variables of greater importance or greater impact, of which commercial opening stands out as the main one. That said, regarding the results of the Granger block causality test set out in appendix C, it should not be surprising that in table 1C, we do not reject the null hypothesis that TT does not cause in Granger's sense to the real exchange rate. On the other hand, Tables 2C, 3C and 4C reject the non-causality in Granger's sense of TT towards import entropy, the informal and the unemployment rate, respectively. Additionally, in appendix D, we present the autocorrelation LM test, the Jarque-Bera normality test on the residences, together with the eigenvalues of the accompanying matrix to verify the stability conditions of SVAR. We can observe that in neither of the equations, individually or jointly, is it possible to reject the hypothesis of the residue’s normality. Secondly, the stability condition is satisfied, since all the modules of the eigenvalues of the companion matrix fall within the unit circle. 7. Conclusions The purpose of this work has been to deepen the understanding of the transmission channels and internal dynamics, typical of a developing country like Bolivia, against external shocks reflected in the terms of trade (TT). Dutch Disease has been extensively dealt with in the literature, and there is evidence to support the presence of this phenomenon in Bolivia, as well as in Latin America. However, most of the work has not widely considered important characteristics, such as a poorly developed industrial sector and high levels labour market informality. Therefore, it is fundamental to adapt the manifestation of DD to the specific characteristics of each country, in order to understand it in its correct dimensions.

The empirical evidence shows that during the period in which there were favourable terms of trade, the real exchange rate has been strongly revaluated. It has also been observed that the manufacturing sector has lost some of its share of total exports. This was accompanied by a drop in production, mainly in textiles, for which negative growth rates were recorded during the boom period. At the same time, there has been a declining share of agricultural employment, and there is evidence that employment has increased in non-tradable sectors of the economy, which may be due to the migration of labour to these sectors (mainly trade and construction). It is difficult to reach a definitive conclusion regarding the presence of Dutch Disease in Bolivia. This work has attempted to sustain the idea that there are certainly indications and evidence that would support the symptoms of this phenomenon, and if it were to present itself, it would be operating through informal labour markets driven by growing and diverse imports. A Structural Vector Autoregressive model (SVAR) has been estimated to analyse the impact of TT shocks on the real exchange rate, the concentration of imports, the rate of informality and unemployment. The results show that the most significant impact has been recorded in labour markets. A deviation of 4% in the trend of TT, would cause the unemployment rate to be diverted almost 5% from its trend. An important question is why it is bad to go through DD. The response provided by Barja et al. (2014) is that, as long as the effects of the boom in the hydrocarbons sector exceed those of mining, there are interesting growth opportunities for aggregate activity. This is even more the case if a stabilization fund is used, financed with extraordinary resources, according to the Resource Rich Developing Country (RRDC) literature on natural resources. We say that favourable TT shocks leave serious considerations for economic policy. Now that the boom is over, the authorities will have to think about direct measures to boost the competitiveness of domestic industry, diversify exports and thus mitigate the impacts on informal employment (“special sector”), generated by the greater and more diverse demand for imports. Currently, the informal employment rate is estimated at more than 60% of total employment in Bolivia. Paying attention to this variable is important because of the direct implications it has on the welfare of many individuals.

References Acosta, P. A., Lartey, E. K., & Mandelman, F. S. (2007). Remittances and the Dutch Disease. Federal Reserve Bank of Atlanta, Working Papers. https://www.frbatlanta.org/-/media/Documents/filelegacydocs/wp0708.pdf. Adler, G., & Magud, N. E. (2013). Four Decades of Terms-of-trade Booms: SavingInvestment Patterns and a New Metric of Income Windfall. IMF Working Paper, Western Hemisphere Department. https://www.imf.org/external/pubs/ft/wp/2013/wp13103.pdf. Agénor, P.-R., & Montiel, P. J. (2015). Development Macroeconomics. Fourth Ed. Princeton University Press. United Kingdom Aguilar, M. A. (2002). Estimación Del Tipo De Cambio Real De Equilibrio Para Bolivia. Revista de Análisis Económico - Banco Central de Bolivia. https://www.bcb.gob.bo/webdocs/publicacionesbcb/revista_analisis/ra_vol0601/capit ulo2final.pdf, 41 - 71. Aguilar, R., & Valdivia, D. (2011). Precios de Exportación de Gas Natural para Bolivia: Modelación y pooling de pronósticos. 2º Bolivian Conference of Development Economics. Amisano, G., & Giannini, C. (1997). Topics in Structural VAR Econometrics. Second, Revised and Enlarged Edition. Berlin Heidelberg: Springer. Andersen, L. E., & Faris, R. (2002). Gas Natural y Distribución de Ingresos en Bolivia. Documentos de Trabajo, Proyecto Andino de Competitividad. http://www.cid.harvard.edu/archive/andes/documents/workingpapers/environmenta lregscompetitiveness/natresourcedependence/gasnatural_distribucioningresos_boliv ia.pdf. Arguello, R., Jimenez, D., Torrez, E., & Gasca, M. (2015). Dutch Disease, Informality, and Employment Intensity in Colombia. Presented at the 18th Annual Conference on Global Economic Analysis, Melbourne, Australia. https://www.gtap.agecon.purdue.edu/resources/download/7656.pdf. Baldivieso, B. (2013). Un caso de enfermedad holandesa: El boom de los hidrocarburos y sus efectos sobre la industria de prendas de vestir en Bolivia entre 2005 y 2011. Universidad Del Salvador, Facultad De Ciencias Económicas, Trabajo Final De Grado. http://www.usal.edu.ar/archivos/di/baldivieso_freitas_bernardo.pdf. Ballesteros, C. (2016). Mining and Energy Boom, Dutch Disease and Informality in Colombia: a DSGE Approach. Universidad EAFIT, Documentos de Trabajo,

Economía y Finanzas, Centro de Investigación Económicas y Financieras. https://repository.eafit.edu.co/handle/10784/9018#.V8XtDE3DDX4. Barja, G., & Zavaleta, D. (2016). Disminución de precios de commodities en un ambiente de ‘enfermedad holandesa’ y ‘bendición/maldición de los recursos naturales’. Latin American Journal of Economic Development, No 25, ISSN: 2074 4706. https://www.researchgate.net/publication/304382533_Disminucion_de_precios_de_ commodities_en_un_ambiente_de_'enfermedad_holandesa'_y_'bendicionmaldicion _de_los_recursos_naturales', 7 - 40. Barja, G., Fernández, B., & Zavaleta, D. (2014). Diminishing Commodity Prices and Capital Flight in a Dutch Disease and Resurce Curse Enviroment: The Case of Bolivia. Papers LACEA 2015. http://lacer.lacea.org/handle/123456789/53011. Bello, O., Heresi, R., & Pineda, R. E. (2010). El tipo de cambio real de equilibrio: un estudio para 17 países de América Latina. CEPAL, Serie Macroeconomía del Desarrollo, No 82. http://repositorio.cepal.org/bitstream/handle/11362/5467/S0900212_es.pdf;jsessioni d=B11E3F0E749C5E2A64F01420DEEAA47E?sequence=1. Bonadona, A. (May 12, 2016). Ni se les ocurra devaluar. Página Siete. http://www.paginasiete.bo/opinion/alberto-bonadonacossio/2016/3/12/ocurra-devaluar-89524.html. Canavire-Bacarreza, G., & Rios-Avila, F. (2015). On the Determinants of Changes in Wage Inequality in Bolivia. Levy Economics Institute, Working Paper 835. http://www.levyinstitute.org/pubs/wp_835.pdf Cerezo A., S. (2014). Testing the Hypothesis of Dutch Disease in the Bolivian Economy. Latin American Journal of Economic Development, No 21, ISSN: 2074 4706. http://www.scielo.org.bo/scielo.php?pid=S207447062014000100004&script=sci_ab stract&tlng=en, 93 - 116. Cerezo, S., & Salazar, D. (2012). Tipo de cambio real en Bolivia: equilibrio y desalineamientos. Latin American Journal of Economic Development, No 18, ISSN: 2074 - 4706. http://www.scielo.org.bo/pdf/rlde/n18/n18_a02.pdf, 9 - 32. Cerezo, S., Humérez, J., & Cossio, J. (2010). El Desempeño del Régimen Cambiario Boliviano en el Periodo Post Estabilización. Documento de trabajo, BCB. http://www.cemla.org/red/papers2010/red-xv-bolivia02.pdf.

Cerruti, E., & Mansilla, M. (2008). Bolivia: The Hydrocarbons Boom and the Risk of Dutch Disease. IMF Working Paper. https://www.imf.org/external/pubs/cat/longres.aspx?sk=22021.0. Colque, R. (2011). Estimación del Tipo de Cambio Real de Equilibrio: Determinantes Fundamentales y Desalineamientos Evidencia Empírica para Bolivia: 1990 2010. Tesis de Grado Magister en Economía. Pontificia Universidad Católica de Chile. http://economia.uc.cl/wp-content/uploads/2015/07/tesis_rcolque.pdf. Cuba, P., & Gonzales, L. (2016). El Embrujo de los Términos de Intercambio en Bolivia. El Faro: un mundo de ideas. http://www.faroeconomics.org/2016/04/elembrujo-de-los-terminos-de.html. de Mevius, F.-X., & Albarracin, I. (2008). Bolivia and the Dutch Disease: What are the Risks and How to Avoid Them? Instituto de Investigaciones Socio Económicas IISEC, Documento de Trabajo No. 09/08. http://www.iisec.ucb.edu.bo/papers/2006-2010/iisec-dt-2008-09-en.htm. Edwards, S., & Van Wijnbergen, S. (1989). Disequilibrium and Structural Adjustment. In H. Chenery, & T. Srinivasan, Handbook of Development Economics, 1481 - 1533. Elsevier. Frankel, J. A. (2012). The Natural Resource Curse: A Survey of Diagnoses and Some Prescriptions. Harvard Kennedy School, Faculty Research Working Paper Series. https://dash.harvard.edu/bitstream/handle/1/8694932/RWP12014_Frankel.pdf?sequence=1. Frankel, J. A., Vegh, C., & Vuletin, G. (2013). On graduation from fiscal procyclicality. Journal of Development Economics, Elsevier, Volume 100, Issue 1, January 13, Pages 32-47. García-Cicco, J., & Kawamura, E. (2015). Dealing with the Dutch Disease: Fiscal Rules and Macro-Prudential Policies. Inter-American Development Bank, Department of Research and Chief Economist. https://publications.iadb.org/bitstream/handle/11319/7087/Dealing_with_Dutch_D isease_Fiscal_Rules_and_Macro_Prud%20ential_Policies.pdf?sequence=1. Goda, T., & Torres García, A. (2015). Flujos de Capital, Recursos Naturales y Enfermedad Holandesa: el Caso Colombiano. Ensayos sobre Política Económica, Elsevier, Volumen 33, Número 78, 10 de diciembre. http://www.banrep.gov.co/es/espe78-4, 197–206. Gonzales, R., & Molina, J. M. (Forthcoming - 2017). On the Graduation From Fiscal Procicality: The Case of Bolivia. Latin American Journal of Economic Development.

Harding, T., & Venables, A. J. (2016). The Implications of Natural Resource Exports for Non-Resource Trade. IMF Economic Review, Palgrave Macmillan, Vol. 64(2), 268-302. INE. (2010). Indicadores de Comercio Exterior de Bolivia 2006 - 2009. La Paz, Bolivia: http://www.ine.gob.bo/pdf/IndicadoresCOMEX/IndicadoresCOMEX06-09.pdf. Ismail, K. (2010). The Structural Manifestation of the `Dutch Disease’: The Case of Oil Exporting Countries. IMF Working Paper. https://www.imf.org/external/pubs/cat/longres.aspx?sk=23801.0. Lanteri, L. N. (2016). Efectos de la enfermedad holandesa (‘Dutch disease’) Alguna evidencia para Argentina. Revista de Economía del Rosario. Vol. 18(2) JulioDiciembre-2015. http://revistas.urosario.edu.co/index.php/economia/article/view/4944, 187-209. Lartey, E. K., Mandelman, F. S., & Acosta, P. A. (2012). Remittances, Exchange Rate Regimes and the Dutch Disease: A Panel Data Analysis. Review of International Economics, Wiley Blackwell, Vol. 20(2), 377-395. Lay, J., Thiele, R., & Wiebelt, M. (2006). Resource Booms, Inequality, and Poverty: The Case of Gas in Bolivia. Kiel Working Paper No. 1287. http://citeseerx.ist.psu.edu/viewdoc/download;jsessionid=FC97C73D2CAEA6D77 BEDD1E22D81F450?doi=10.1.1.163.8375&rep=rep1&type=pdf. Lütkepohl, H. (2005). New Introduction to Multiple Time Series Analysis. Berlin Heidelberg: Springer. Magud, N., & Sosa, S. (2010). When and Why Worry About Real Exchange Rate Appreciation? The Missing Link between Dutch Disease and Growth. IMF Working Papers. https://www.imf.org/external/pubs/cat/longres.aspx?sk=24395.0. Mendieta, P. H. (2007). El Equilibrio de la Competitividad Cambiaria Boliviana: Un Enfoque Empírico. Documento de Trabajo para Presentación en el Encuentro de Economistas de Bolivia. https://www.bcb.gob.bo/webdocs/EEB/TCR%20PM%20%20mendieta.pdf. Morales, J. A. (July 27, 2015). Política cambiaria para adultos. Página Siete. http://www.paginasiete.bo/opinion/2015/7/27/politica-cambiaria-paraadultos-64448.html. Morales, R. (2012). El Desarrollo Visto Desde el Sur. La Paz, Bolivia: INESAD. Morales, R., Alarcón, S., & Gonzales, R. (2016). Dutch Disease and the Labor Market in Bolivia. http://r4d.africantransformation.org/.

Neary, P., & Van Wijnbergen, S. (1986). Natural Resources and the Macroeconomy. Centre for Economic Policy Research. Blackwell, Oxford, UK, 352 pp. Oomes, N., & Kalcheva, K. (2007). Diagnosing Dutch Disease: Does Russia Have the Symptoms? IMF Working Paper, Middle East and Central Asia Department. https://www.imf.org/external/pubs/ft/wp/2007/wp07102.pdf. Powell, A. (2016). Time to Act Latin American and the Caribbean Facing Strong Challenges. Washington, DC. https://publications.iadb.org/bitstream/handle/11319/7533/Time-to-ActLatin-America-and-the-Caribbean-Facing-StrongChallenges.pdf;sequence=1: IDB. Raich, A. S. (2014). Tipo de Cambio Real y Competitividad: Enfermedad Holandesa en Latinoamérica. Universidad Nacional de Cuyo - Facultad de Ciencies Económicas, Trabajo de Investigación. Prof. Tutor: Alejandro Trapé. Rajan, R. G., & Subramanian, A. (2011). Aid, Dutch Disease, and Manufacturing Growth. Journal of Development Economics, Elsevier, Vol. 94(1), January, 106118. Sala-i-Martin, X., & Subramanian, A. (2003). Addressing the Natural Resource Curse: An Illustration from Nigeria. NBER Working Papers. http://www.nber.org/papers/w9804. Salter, W. E. (1959). Internal and External Balance: The Role of Price and Expenditure Effects. The Economic Record, 1959, Vol. 35, Issue 71, 226-238. Schmitt-Grohé, S., & Uribe, M. (2015). How Important are the Terms of Trade Shocks? NBER Working Paper 21253. http://www.nber.org/papers/w21253. Sierra, L. P., & Manrique L., K. (2014). A first approach to the impact of the real exchange rate on industrial sectors in Colombia. CEPAL Review 114. http://repositorio.cepal.org/bitstream/handle/11362/37812/RVI114Sierra_en.pdf?se quence=1, 120 - 134. Sims,

C. A. (1980). Macroeconomics and Reality. Econometrica 48. http://www.ekonometria.wne.uw.edu.pl/uploads/Main/macroeconomics_and_realit y.pdf, 1–48.

Stein, E., Fernández, A., & Rosenow, S. (2016). Has Latin America Experienced Real Depreciation? Insights from a Competition- and Similarity-Adjusted Real Effective Exchange Rate. Washington, DC: Inter-American Development Bank. Unpublished. The Economist. (November 26, 1977). The Dutch Disease. 82-86.

Uribe, M., & Schmitt-Grohé, S. (2016). Open Economy Macroeconomics. In preparation for Princeton University Press. http://www.columbia.edu/~mu2166/book/usg.pdf. Vegh, C., & Vuletin, G. (2013). The Road to Redemption: Policy Response to Crises in Latin America. IMF 14 TH Jacques Plak Annual Research Conference. https://www.imf.org/external/np/res/seminars/2013/arc/pdf/vegh.pdf. Vegh, C., & Vuletin, G. (2016). To be countercyclical or not? That is the question for Latin America. VOX CERP´s Policy Portal. http://voxeu.org/article/becountercyclical-or-not-question-latin-america. Wong, S. A., & Petreski, M. (2014). Dutch Disease in Latin American countries: Deindustrialization, how it happens, crisis, and the role of China. MPRA Paper No. 57056. https://mpra.ub.uni-muenchen.de/57056/.

Appendix A: The Data The data used to estimate the SVAR model are quarterly and cover the period from 2000 - 2014. The following describe each of the variables and sources: • Terms of trade index: the terms of trade index of Cuba y Gonzales (2016)35, the base year is 2006 and the data are quarterly. • Real effective exchange rate index (REER): monthly data calculated by ECLAC using the following expression: 𝑟𝑒𝑟𝑡 =

𝐵𝑂𝐿−𝑗 𝑗 𝜔𝑗 𝐸𝑡 𝑃 40 ∏𝑗=1 ( 𝐵𝑂𝐿 𝑡 ) 𝑃 𝑡

It is a geometric mean of the real exchange rate between Bolivia and its main 40 𝐵𝑂𝐿−𝑗

trading partners. 𝐸𝑡

is the nominal exchange rate in period t between Bolivia

𝑗

and country j. 𝑃𝑡 and 𝑃𝑡𝐵𝑂𝐿 are consumer price indices in country j and Bolivia respectively in period t, with base year 2005. 𝜔𝑗 is country j’s participation in international trade with Bolivia36. The conversion of monthly data to quarterly was achieved through simple arithmetic means. • Entropy of imports: it is a concentration indicator calculated using the following expression: 𝑒𝑛𝑡𝑡 = − ∑𝑁 𝑖=1 𝑝𝑖 ln(𝑝𝑖 ) Where 𝑝𝑖 is the share of item i in the total. We have worked with data from the National Institute of Statistics (INE) from the uniform classification of trade (CUODE) by large items, taking the cif value of imports. • Informality rate: this variable has been estimated based on INE’s37 annual Household Survey. For the missing data from 2004 and 2010 (years for which the survey is not available), we perform a simple interpolation. 10 criteria were used to define an individual as formal or informal, these being: 1. Workers who receive a monthly salary are formal and other workers are informal. 2. Employees who receive a monthly salary, work with more than five workers and work between five and six days a week, while other employees are informal.

For more details, we suggest referring to this work. The authors estimate the terms of trade index using the same methodology described in INE (2010) and report a correlation of 90% between their estimate and the INE official estimate, which stopped being published in 2013. 36 It is the quotient of the sum of Bolivia's exports and imports with country j on Bolivia's total exports and imports. 37 We appreciate the help of CIESS ECONOMÉTRICA for allowing us access to these estimates. Those interested in accessing the information can visit www.ciess-econometrica.com.bo 35

3. Employers (only those registered in EIH surveys) who work with more than five workers are formal and other employers are informal. 4. Remunerated employers (only those registered in MECOVI surveys) who work with more than five workers and receive monthly salary are formal, while other remunerated employers are informal. 5. Unpaid employers (only those recorded in MECOVI surveys) that work with more than five workers and less than six days per week are formal and other unpaid patterns are informal. 6. All cooperatives (only those registered in MECOVI surveys) are informal. 7. All unpaid workers are informal. 8. All household workers are informal. 9. Self-employed workers who have 17 or more years of schooling are formal and other self-employed workers are informal. 10. Qualified independent professionals with 12 or more years of schooling (only those registered in EIH surveys) are formal and other independent professionals are informal. • Unemployment rate: the source is the World Bank's World Development Indicators (WDI) modelled by the International Labour Organization (ILO). Informality and unemployment rates have been quartered by the Dénton-Cholette method in R, using the 1990 base real GDP of the INE as a reference indicator. The results of the paper are robust for each quarter according to the Chow-Lin method. The trends of the variables were extracted using the Christiano-Fitzgerald filter and a four-quarter moving average was also used as a measure of robustness. Finally, the seasonality has been adjusted using the X-13 algorithm.

Appendix B: Impulse response functions Impulse of a structural terms of trade shock Figure 1B Real exchange rate response

Quarters

Figure 2B Import entropy response

Quarters

Confidence intervals at 90%, generated by 2000 bootstrap replicates.

Impulse of a structural terms of trade shock Figure 3B Informality rate response

Quarters

Figure 2B Unemployment rate response

Quarters

Confidence intervals at 90%, generated by 2000 bootstrap replicates.

Appendix C: Granger causality test Table 1C Real exchange rate equation Variables Prob. Chi-squared Terms of trade 0.486 Import entropy 0.001 Informality rate 0.478 Unemployment rate 0.000 All 0.000

Table 2C Import entropy equation Variables Prob. Chi-squared Terms of trade 0.000 Real exchange rate 0.010 Informality rate 0.000 Unemployment rate 0.242 All 0.000

Table 3C Informality rate equation Variables Prob. Chi-squared Terms of trade 0.041 Real exchange rate 0.108 Import entropy 0.000 Unemployment rate 0.147 All 0.000

Table 4C Unemployment rate equation Variables Prob. Chi-squared Terms of trade 0.000 Real exchange rate 0.000 Import entropy 0.363 Informality rate 0.135 All 0.000

Appendix D: Autocorrelation, normality and stability in the SVAR Lag 1 2 3 4 5 6 7 8 9 10

LM Test Prob. Chi-squared 0.01583 0.13783 0.26239 0.01289 0.18409 0.62417 0.08558 0.98466 0.64666 0.88987

Jarque-Bera Test Equation Terms of trade Real exchange rate Import entropy Informality rate Unemployment rate All

Eigenvalues Eigenvalues Modules 0.823 + 0.308𝑖 0.879 0.823 − 0.308𝑖 0.879 0.850 + 0.175𝑖 0.868 0.850 − 0.175𝑖 0.868 0.851 0.851 All eigenvalues fall within the unit circle, satisfying the SVAR stability condition.

Prob. Chisquared 0.541 0.297 0.123 0.377 0.863 0.431

Figures Appendix Figure 1 Price Indices and Volume of Bolivia's Traditional Exports (2006 = 100)

160

Figure 2 Export Price Index of Bolivia (2006 = 100)

600 500

110

400 300

60

200 10

100 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 0

-40 Prices

Volumes

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 Minerals Hydrocarbons Non-traditional Source: UDAPE

Source: UDAPE

Figure 3 Real Exchange Rate Index of Bolivia (2005 = 100) 105 100 95 90 85 80 75

Figure 4

100000000 80000000 60000000 40000000 20000000

70

0 2000

65 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Source: CEPALSTAT

Expoxts and Imports of Bolivia (per thausands of Bolivians)

120000000

Source: INE

2002

2004

2006 Exports

2008 2010 Imports

2012

2014

1,2

Figure 5

Figure 6

Entropy of Exports and Imports

Comsumer Price Index in Tradable and Non-Tradable Goods and Services (2010 = 100)

1,1 160

1 0,9

110

0,8 0,7

60

0,6 0,5

10

0,4 2000

2002

2004

2006 2008 Exports

2010 Imports

2012

2014

2016

Source: Own elaboration, data from INE. See appendix A.

2004

2006

Tradable CPI

2008

2010

2012

Non-tradable CPI

Figure 8

Unemployment and Informality Rate in Bolivia

Ratios of Agriculture and Manufacturing in Terms of Mines and Gas (Real production)

5% 4% 3% 2% 1% 2002

2004

2006

Unemployment Rate Source: Informality rate: CIESS-ECONOMÉTRICA Unemployment rate: WDI – World Bank

2014

Source: CEPALSTAT

2008

2010

2012

90% 80% 70% 60% 50% 40% 30% 20% 10% 0% 2014

180 160

Informality Rate

Unemployment Rate

2002

Figure 7

6%

0% 2000

2000 -40

140 120 100 80 60 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Agriculture / Mines and Gas Ratio

Informal Employment Rate Source: UDAPE

Manufacturing / Mines and Gas Ratio

Population Occupied in Agriculture and Manufacturing (% of Total)

11,5

70

36

50

34

30

32

10

% Manufacturing Industry

Ratios of Agriculture and Manufacturing in Terms of Non-tradable Goods and Services (Real production)

Agricultura

Desv. % de la tendencia

38

Figure 10 50

11

40

10,5

30

10

20

9,5

10

9 30

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 Agriculture / Non-tradable Ratio

0 2000

-10

2001

Manufacturing / Non-tradable Ratio

Source: UDAPE

2002

2003

2005

2006

2007

2008

Manufacturing Industry

2009

2011

2012

Farming

Source: UDAPE

Figure 11

Figure 12

Occupied Population in Mining and Quarrying and Non-Tradable Goods and Services (% of the total) 2,5

Textil Production (In Gross Weight Kg.)

14000000 40 12000000

2

30

1,5 20 1 10

0,5 0

0 2000

2001

2002

2003 2005 Non-Tradable

2006

2007 2008 2009 2011 Mining and Quarrying

2012

Source: UDAPE Non-tradable = Construction + Trade, Restaurants and Hotels + Transport and Communications

% Non-Tradable

% Mining and Quarrying

% Farming

Figure 9

10000000 8000000 6000000 4000000 2000000 0 2000

2002

2004

2006

2008

2010

2012

2014

2016

Source: INE: Division CIIU Rev. 3 Textile Production = Textile Products Manufacturing + Fur Garments Manufacturing, Dressing, and Skin Dyeing.