Professional Documents
Culture Documents
The Great Divergence - Assessing The Lost Decade of The Latin American Miracle
The Great Divergence - Assessing The Lost Decade of The Latin American Miracle
The Great Divergence - Assessing The Lost Decade of The Latin American Miracle
First version: 14th November 2023/ This version: 21st November 2023
Abstract
The Latin American region has suffered an economic slowdown since the end of the
commodities' boom. This has affected the capacities of the region to tackle inequality and
pending social ills, which has led to protests and the rise of populism. Within this context,
Chile was the poster child of economic growth and development up until 2014. Since then,
Chile has also been trapped in a decade of slow economic growth. Chile's sudden slowdown,
and its recent growth path divergence, have posed a puzzle for economists interested in
economic growth and development. This paper examines this divergence considering Chile’s
recent changes in public policy. We argue that Chile experienced a 'great divergence' from
its recent economic trend largely due to an internal policy regime change. To assess the
timing and quantify this 'great divergence', we use a synthetic control approach, thus
assessing the economic consequences of Chile's recent policies. Our results suggest that,
unlike what is believed, Chile’s recent growth slowdown has been mainly a product of
internal factors, such as its policy regime change, rather than external or regional factors such
as international prices of commodities. This research sheds light on the potential effects of
policy regime changes in affecting long-term economic growth, thus providing valuable
guidelines for development economics and macroeconomic policy.
1. Introduction
From 2012 the Latin American region has suffered for tumultuous political and economic
events, which have hindered its capacity to attain democratic stability, generate economic
growth, and promote widely shared prosperity (Ocampo, 2020). The recent decade has been
categorized as Latin America’s “new lost decade” (Ocampo, 2021). Slowing growth, high
inflation, and global uncertainty due to wars, pandemics, and climate change means that
many people in the region will see their living standards decline and their real wages stagnate.
This has generated increase anxiety about the future, stimulating social conflict, polarization,
and political tensions around the continent. The region has suffered a relevant economic
slowdown since the end of the commodities' boom during 2012-2014, leading to frustration
and social unrest (Alfaro and Jeong, 2019; Ocampo, 2020).
This stagnation has affected the capacities of the region to tackle inequality and many
pending social ills, which has led to protests and new waves of populism across the region
(Edwards, 2019; Justino and Martorano, 2019; Segovia, et al., 2021). The Latin America and
the Caribbean regions (LAC) are highly dependent on commodities and, as such, they
benefited from the ‘commodities’ super-cycle’ since early 2000s up until 2012-2014 (Erten
and Ocampo, 2013; Gruss, 2014). Analogously, LAC also suffered when the commodities’
boom ended, leading to a recent decade of economic stagnation (Ocampo, 2017, 2020). Since
then, countries such as Peru, Chile, Bolivia, Ecuador, Colombia, and others have experienced
protests, political instability, and violence (Segovia, et al., 2021). Analysists have coined the
term “Latin American Spring” to define the recent tumultuous period that followed the end
of the commodities’ boom (Rice, 2020).
Even though the causes of these events and the origins of protests are numerous—
ranging from political polarization, social media networks, and persistent levels of
inequalities (Barrett, et al., 2020; Busso and Messina, 2020; Segovia, et al., 2021)—a key
detonating factor in the “Latin American Spring” has been the sudden economic slowdown
(Cerda and Vergara, 2022; Faiola and Krygier, 2019). In other words, “the protests now
raging across much of Latin America originated from different sparks but are connected by
1
The term ‘great divergence’ came into vogue in economics thanks to K. Pomeranz’s (2000) book,
Great Divergence. The term refers to the growing gap in economic performance between Europe and
China –and by extension the rest of Asia and Arabia—since the late eighteenth century.
10,0% 9,4%
8,2%
8,0%
Real GDP per capita Growth
7,4%
6,1% 6,0%
6,0% 5,6% 5,4% 5,6%
5,2%
4,9% 5,0% 5,1%
4,8% 4,8% 4,8%
4,1%
3,8% 3,8%
4,0% 3,5% 3,6%
2,9% 2,8%
2,5% 2,3%
2,1%
2,0% 2,1%
2,0% 1,7%
1,1%
0,8% 0,6%
0,0%
-0,2%
-1,0%
-2,0% -1,5%
-2,1%
-4,0%
1985 1990 1995 2000 2005 2010 2015
Fig. 1. Chile’s real GDP per capita (constant 2015 US$) growth (1985-2019). Dotted vertical
line marks 2014. Authors’ elaboration based on World Bank data.
Table 1 shows clearly that the last period between 2014-2019 has been the worst set
of years concerning economic growth since the tumultuous years of Allende’s socialist
project (Edwards, 2023); period also in which Chile’s economic growth per-capita got
reduced by more than a fifth in comparison with the ‘miracle’ period of 1985-2000 (Becker,
1997; Escalante, 2022). The data of table 1 also shows that, in between 2014-2019, Chile’s
average per capita GDP growth was a meager 0,6%, significantly below the world average
(Paniagua, 2021). From these stylized facts of Chile’s recent trajectory of economic growth,
three relevant facts standout: first, the period between 2014-2019 has been the period with
the worst performance concerning economic growth since the tortuous 1970s, when
compared against its own average economic performance (Edwards, 2023; Paniagua, 2021).
Second, since 2014 onwards the Chilean economy, on average, is growing severely less than
it used to grow during the golden years of the ‘Chilean Miracle’. Third, and finally, the
Chilean economy in less than a decade went from growing above the world’s average during
1985-2013, from growing significantly below the world’s average during 2014-2019. We
believe this to be quite a puzzle for macroeconomic and development policy. Hence, a
Since the decline in commodity prices during 2012-2014, and the end of the “commodity
super-cycle” (Roch, 2017), a debate amongst economists and other social scientists has
spurred about the causes and consequences of the economic slowdown that the region has
suffered (Aqib, et al., 2016; Didier, et al., 2016; Gruss, 2014; Roch, 2017, 2019; Segovia, et
al., 2021). Didier et al. (2016) argue that the slowdown is synchronous and protracted in
emergent economies, affecting a sizable number of emerging markets, particularly large
ones. The authors suggest that the recent growth slowdown during the last decade has been
driven by both external factors—including weak world trade, low commodity prices, and
tightening financial conditions—and domestic factors, such as slowdown in productivity
growth, policy uncertainty, and an erosion of fiscal buffers (ibid.).
Roch (2019) has argued that shocks through commodity terms of trade are an
important driver of recent business-cycle fluctuations in developing countries. In some cases,
Roch (2019) found that the commodity terms of trade shocks could explain around 30
percent of movements in output. Similarly, Roch (2017) establishes that, in Chile and Peru,
the terms of trade doubled from 2000 to 2011, and in Colombia they increased by 70 percent,
60%
140%
52%
130%
44%
120%
36%
28% 110%
20% 100%
1990 1995 2000 2005 2010 2015 2020 1990 1995 2000 2005 2010 2015 2020
Fig. 2. Chile's GDP per capita PPP (constant 2017 international $) relative
to selected countries (left) and the World (right). Dotted vertical line marks
2014. Authors’ elaboration based on World Bank data.
The reason why the focus of attention has resided in 2014 as a crucial break in Chile’s trend
of economic performance is not only because it is the period that indicates the beginning of
a significant divergence in Chile’s macroeconomic trend (see previous discussion), but also
due to recent historical and public policy reasons. The fact that a potential structural break
occurred in 2014 is consistent with the historical narrative fact that several important policy
related events took place in the country during that year, which represent a relatively major
policy regime change for the country.
In March 2014 President elected Michelle Bachelet arrived for a second time to the
Chilean presidency after a presidential campaign in 2013, promising “a far-reaching reform
program that vowed to initiate a ‘new historical cycle’ in the South American nation”
(Benedikter, et al., 2016, p. 2). Bachelet’s second presidential campaign was marked with a
strong ‘anti-neoliberal’ and ‘anti-market’ rhetoric, aimed at changing the Chilean economic
model of development that the country adopted since 1980s onwards (Edwards, 2023;
10
2
As Bachelet recognized while still in office: “there were some vestiges of the neoliberal model that
we have been putting an end to through the reforms” (La Tercera, 2017).
3
To specify the timing of these reforms during 2014: i) the fiscal and tax reform was announced in
March 2014 and later approved by the Chilean Congress in September 2014; ii) the political reform
that changed the electoral system was announced in April 2014 and later approved by Congress in
November 2014; iii) the educational reform was signed into law by Bachelet during May 2014; 4) the
labour market reform was sent to Congress by President Bachelet in December 2014; v) the pension
reform failed to become a bill during 2014-2018; and, finally, vi) the constitutional reform announced
in October 2014 failed to generate an official document during Bachelet’s presidency, but
spearheaded a long-winded constitutional qualm.
11
4
As pointed out by Claro and Sanhueza (2023), 2014 was the first time in many years in which
Chilean local uncertainty index surpassed international uncertainty index. Additionally, real
investment yearly growth rate went from an average 7.2% during 2000-2013 to -0.6% during 2014-
2018.
5
In the words of Benedikter et al (2016, p. 2): “Bachelet vowed to enact multi-dimensional change
so far-reaching and interdisciplinary in scope and extension that she called it a coordinated array of
‘policies that change cultures’. Overall, Bachelet promised to not only apply sectorial corrections, but
to change the functional, institutional and constitutional basics of the nation to create a ‘new culture’”.
12
6
The tax reform sought to eliminate the Taxable Profits Fund (Fondo de Utilidades Tributables)
known in Chile as FUT; thus, starting to tax accrued income at the partner level instead of withdrawn
income. For a review of Chile’s recent history of tax reforms and the FUT consult: Kinghorn (2016).
13
7
For a review of the literature concerning the effects of fiscal policies and taxes on economic growth
consult: Nijkamp and Poot (2004), Larraín, et al. (2014), and Gemmel, et al. (2011).
8
The tax evidence needs to be taken cum grano salis since different countries have different tax
structures, there are different effects also between personal and corporate taxes, and, finally, the
potential negative effects of tax increases on output could net themselves out with better and more
efficient government spending. Not all taxes might generate distortions or negative effects, and not
all government spending is unproductive or wasteful. Nevertheless, the literature suggests that
corporate taxation is, comparative speaking, the most harmful to growth (Arnold, et al., 2011).
14
Given the state of the debate concerning Latin America’s and Chile’s sudden economic
slowdowns review in the previous sections, we will now explore the hypothesis that Chile
experienced a ‘great divergence’ mainly due to internal factors rather than external ones. We
seek to provide a methodologically plausible answer for the question: how would have
Chile's economic trajectory unfolded if it had not implemented the set of internal reforms in
2014? Addressing this question presents a significant challenge due to the lack of reasonable
counterfactuals. For example, consider the hypothetical scenario in which a country
implements an industrial policy, and experiences a 5% growth in subsequent periods (ceteris
15
16
As a starting point, no individual country has the capacity to approximate and mimic the main
socio-economic indicators of Chile, since the heterogeneity amongst Latin American
countries is simply too high. Therefore, we need a systematic way of choosing which mix of
countries would represent, the best as possible, the pre-treatment period of the Chilean
economy. Following Abadie and Gardeazabal (2003), Absher, et al. (2020), and most recently
Abadie (2021), we choose a pool of “donor” countries that share, in a smaller or greater
degree, similar conditions to the treatment period for the independent variable. Such
conditions could be culture, history, geography, education, language, structural economic
similarities, and overall institutional framework. (Abadie, 2021; Absher, et al., 2020).
We use annual country-level panel data for the period 1990-20199 with two pools of
donors, named ‘group I’ and ‘group II’ as depicted in table 2. The first pool, group I, consists
of most of Latin American countries, with up to 13 donors, therefore capturing similarities in
cultural background, language, geography, history, and some other local aspects. The second
one, group II, includes all the donors of the first pool, but also introduces countries such as
Spain and Portugal, given their past as colonizers of most Latin American countries, and also
some main trading partners and commodity exporters such as China and Australia (Escalante,
2022). The above is done to match the significant copper exporter nature or ‘commodity
9
We considered unwise to extend the analysis further to 2020 and onwards for two main reasons:
first, Chile experienced a severe social unrest marked with street violence and the destruction of
public infrastructure that negatively affected GDP after the treatment period. Second, and right after
March 2020, the Covid-19 pandemic arrived, and Chile followed a very strict logic of national
lockdowns that severely contracted GDP during 2020-2021.
17
10
With one country exception, Venezuela due to lack of available information and data reliability.
11
The Root Mean Square Prediction Error measures the fit (or lack of it) between the actual (Y) and
synthetic (Y!"#$% ) country estimation. The RMSPE is defined as:
)
1
RMSPE = ( ,(Y − w ∗ Y!"#$% ) ( ) +/(
T
$*+
12
As an additional form of robustness, results for group I, are also included in the appendix.
18
13
Currently, Chile only represents near a 5,5% of the total GDP of the entire Latin American and the
Caribbean region, thus, being considered a small economy (Vial, 2018).
19
Finally, just as important as the donor selection, there is the selection of the variables
that the SCM needs to match to build the synthetic Chile. Indicator variables should be known
to be good predictors of the outcome variable, in this case real GDP per capita14. In line with
14
Throughout the paper we use the measure of real GDP per capita at constant 2015 US$ from the
World Bank Database. As further robustness, we provide several estimations for a variety of GDP
measures such as Real GDP per capita at PPP (constant 2017 international $) and GDP per capita at
current US$, see Appendix.
20
15
Our synthetic Chile tracks accurately the actual ‘economic boom’ of Chile during 1990-2014, matching well
also the Asian financial crisis of late 1990s (1997-1999) that affected most of the Latin American region, as
well as the 2007-2008 Great Recession. In the appendix, we provide a similar plot including the samples
average trajectory. The above highlights the importance of this accomplishment.
21
$15.000
Real GDP per capita constant 2015 US$
$14.000
$13.000
$12.000
$11.000
$10.000
$9.000
$8.000
$7.000
$6.000
$5.000
1990 1995 2000 2005 2010 2015
Synthetic Chile Real Chile
Fig. 3. Per-capita income. Note: The solid blue line represents real GDP per capita
(constant 2015 US$) in Chile, 1990-2019; the red line represents the synthetic control.
The vertical black dotted line indicates the end of the pre-treatment years (1990-2013).
Perhaps, a result that is even more meaningful for the long-term economic growth of
Chile since 2014 onwards is the abrupt change in long-run GDP trend depicted in figure 4.
Using a Hodrick-Prescott (HP) filter16 we eliminate the cyclical component of the series to
focus on Chile’s long-run economic growth trend (Hodrick and Prescott, 1997). Figure 4
displays the evolution of such economic trends both for the actual and synthetic Chile. From
the figure we can see that our synthetic counterfactual does a good job in matching very
closely the long-run trend of Chile’s GDP per capita for a period of 24 years in the pre-
16
Sensitivity parameter adjusted to yearly frequency, 𝜆 = 100.
22
$16.000
$15.000
Real GDP per capita constant 2015 US$
$14.000
$13.000
$12.000
$11.000
$10.000
$9.000
$8.000
$7.000
$6.000
$5.000
1990 1995 2000 2005 2010 2015
Synthetic Chile Trend Real Chile Trend
Fig. 4. GDP per-capita trends with HP filter. Note: The solid blue line represents
the real GDP per capita (constant 2015 US$) trend in Chile, 1990-2019; the red
line represents the trend of the synthetic control.
Finally, even though we can find strong causality and identify the crucial role of
internal causes in the Chilean growth slowdown, we cannot pinpoint exactly which policy
changes affected Chile’s economy for better or worse. However, our results are coherent with
the political economy of tax changes reviewed in the previous section (Arnold, et al., 2011;
Barro and Redlick, 2011; Mertens and Ravn, 2013; Romer and Romer, 2010; van der Wielen,
2020; Widmalm, 2001), as well as with the role that political instability, ideology, and
23
4. Robustness tests
The above findings are provocative, but they require robustness checks to be validated. To
evaluate the credibility of our benchmark results we conduct several robustness checks such
as in-time and country placebo tests, both with additional RMSEP analysis, p-values
significance, multiple leave-one-out (Jackknife) permutation tests and, finally, estimating the
SCM while changing the composition-size of our donor’s pool to an alternative group. In this
section we will explore these robustness tests that support our findings.
The in-time placebo test consists in re-estimating the model with an intervention year
other than 2014. If the pre and post treatment periods do not show any significant tracking
difference –that is the placebo intervention had no perceivable effect– there would be a strong
indication that we are correctly estimating the real impact of the Chilean internal set of
reforms that took place in 2014. Finally, we also provide estimations for different placebo
years. Analogously, the country placebo tests consist in estimating the SCM not only to Chile,
but for every other donor country. If the results produce a large placebo estimate, that could
undermine our confidence that the results are indeed indicative of the internal set of reforms
24
17
We select 9 years before the actual intervention of 2014, which is almost half the period of our
pretreatment window, in line with Abadie et al (2015). The results are also robust to alternative
intervention periods. We have also computed similar in-time placebo studies where we reassign the
‘placebo-intervention’ to the years 2000 and 2008, in which results are similar to the one shown here.
25
$14.000
$12.000
$11.000
$10.000
$9.000
$8.000
$7.000
$6.000
$5.000
1990 1995 2000 2005 2010 2015
Synthetic Chile Real Chile
Fig. 5. In-Time placebo test. The solid blue line represents the real GDP per capita
(constant 2015 US$) in Chile, 1990-2019; the red line represents the synthetic control.
The vertical black dotted line indicates the new ‘treatment’ (placebo) year set in 2005.
18
Following the methodological recommendations of Abadie et al. (2010), we drop the observations
that had poor pre-treatment fit for this analysis.
26
5000
Deviations = Effective - Synthetic
4000
3000
2000
1000
-1000
-2000
1990 1995 2000 2005 2010 2015
Control Units Chile
Fig. 6. Real GDP per capita (constant 2015 US$) placebo tests, restricted countries. Note: The
bold line represents the difference between the observed income per-capita in Chile, 1990-
2019, and the synthetic control. Analogously, the gray lines represent the same difference but
for different donors, representing different placebo tests.
Additionally, and related with the above, table 4 provides the RMSPE ratios for all
positive weighted country donors in our estimated model (see list in table 2). These results
show that the RMSPE ratios of the main donors in our pool (i.e., China, Costa Rica, Uruguay
and Australia) have an inferior ratio of that of the synthetic Chile, which is another indication
of the robustness and of the correct measurement of the real effect of the 2014 internal
reforms in Chile.
27
Chile - 28,67
4.3. P-values
Following the country-placebo test, it is standard in the literature to assess if the differences
between the actual and synthetic Chile are in fact statistically significant. To assess the above,
we need to compare how “atypical” is the Chilean synthetic divergence from the rest of the
placebo tests. Figure 7 shows the difference between synthetic Chile and the synthetic
average of all placebos in the pool with its respective p-values.
For the final years of the post-treatment, results are statistically significant under a
90% confidence interval. The above means that we can safely reject the null hypothesis (no
“atypical” divergence), reinforcing the idea that the difference between actual Chile and its
synthetic average are in fact significant and that results are robust in this dimension.
28
.18
-400 .08
.04
-600
-800
-1000
-1200
2015 2016 2017 2018 2019
Post-Treatmenmt Years
Fig. 7. Effects of the treatment and p-values on real GDP per-capita. Bars show the
difference between actual Chile and the placebos’ synthetic average, measuring the
treatment effect for each year. Center values represent the statistical significance level.
29
$15.000
$14.000
Real GDP per capita constant 2015 US$
$13.000
$12.000
$11.000
$10.000
$9.000
$8.000
$7.000
$6.000
$5.000
1990 1995 2000 2005 2010 2015
w/o Costa Rica Real Chile w/o China
w/o Uruguay w/o Australia w/o Panama
Fig. 8. Jackknife multiple permutation test (leave-one-out). Solid blue line plots
actual GDP per-capita for Chile. All the other lines represent the different SCM
estimations without that specific country on the sample.
To conclude, and considering the results of the previous sections, and the several different
robustness tests presented here, we have reasons to believe that our hypothesis is correct, and
that it is safe to say that Chile experienced a ‘great divergence’ in economic growth mainly
due to internal ‘policy’ factors. All in all, there is a strong case to be made that both in terms
of long-term trend in economic growth and real income per capita, the 2014 ‘anti-neoliberal’
set of reforms represented not only a drastic normative shift in the policy regime for Chile,
but also these policies seriously harmed the long-term economic performance of Chile
relative to what it would have occurred with a “business as usual” set of policies (as given
by the predictions of the synthetic control).
30
In this paper we have studied a crucial problem in developing economics and in the recent
debates on economic growth: why countries that appear to be successful and growing rapidly
could suddenly suffer from growth reversals and experience a prolonged divergence in terms
of economic growth (Aiyar, et al., 2013; Eichengreen, et al., 2014). We have focused our
attention on the recent macroeconomic events of Chile, and our analysis suggests that there
was a significant change in Chile’s economic growth trend in 2014: a negative reversal of
fortune that has affected its long-term path of economic growth. Our analysis shows that this
‘great divergence’ is related to endogenous factors, that are most likely attributed to
significant political (and ideological) changes, that drastically altered both the normative
discourse and the content of the overall public policy framework of that country.
We have provided evidence to support the claim that during 2014 there was a change
in Chile’s policy framework, which—alongside a hostile narrative against businesses and
market-based ideas (Edwards, 2023)—became a pivotal factor in Chile’s great divergence
and its enduring economic slowdown that has brought a decade long of weak economic
growth. Although further work is necessary to establish the precise causal mechanism at play
to indicate how changes in the policy framework specifically affected economic growth, our
analysis is consistent both with the political economy of tax shocks and corporate taxes
(Arnold, et al., 2011; Barro and Redlick, 2011; Mertens and Ravn, 2013; Romer and Romer,
2010; van der Wielen, 2020; Widmalm, 2001), as well as with the role that ideology, political
instability and changes in policy might play in affecting economic development (Acemoglu
and Robinson, 2006; Bjørnskov, 2005; Easterly, 1992; Edwards, 2019; Rodrik, 2014).
Empirically, this ‘anti-neoliberal’ shift costed Chile almost 10% of its real GDP per
capita in only five years. This result is both large and strongly significant. This translated in
the fact that the actual Chile is $1.345 (per capita) poorer than what the synthetic
counterfactual predicts. The most noteworthy result of our analysis is the fact that this ‘great
divergence’ in economic growth seems to be persistent over the long-run as our SCM with a
Hodrick-Prescott (HP) filter suggests. These results could be interpreted as the flipside of
Billmeier and Nannicini (2013), by suggesting that all-encompassing ‘anti-neoliberal’
reforms can create large and persistent wealth penalties for those countries seeking to
implement them (Edwards, 2019; Grier and Maynard, 2016; Spruk, 2019). Further work is
31
32
$16.000
$15.000
Real GDP per capita constant 2015 US$
$14.000
$13.000
$12.000
$11.000
$10.000
$9.000
$8.000
$7.000
$6.000
$5.000
1990 1995 2000 2005 2010 2015
Synthetic - Group I Real Chile
Fig. 1 Appendix. Per-capita income. Note: The solid blue line represents real GDP per
capita (constant 2015 US$) in Chile, 1990-2019; the red line represents the synthetic
control only with countries in group I. The vertical black dotted line indicates the end of
the pre-treatment years (1990-2013).
33
34
35
$25.000
Real GDP per capita constant 2015 US$
$23.000
$21.000
$19.000
$17.000
$15.000
$13.000
$11.000
$9.000
1990 1995 2000 2005 2010 2015
Synthetic Chile Real Chile
Fig. 4. Appendix. GDP per capita PPP constant 2017 international $. Note: The solid
blue line represents observed income per-capita in Chile, 1990-2019; the red line
represents the synthetic control. The vertical black dotted line indicates the end of the pre-
treatment years (1990-2013).
36
$25.000
Real GDP per capita constant 2015 US$
$23.000
$21.000
$19.000
$17.000
$15.000
$13.000
$11.000
$9.000
1990 1995 2000 2005 2010 2015
Synthetic Chile Trend Real Chile Trend
Fig. 5 Appendix. GDP per capita PPP constant 2017 international $ trends with
HP filter. Note: The solid blue line represents observed per-capita income trend in
Chile, 1990-2019; the red line represents the trend of the synthetic control.
37
38
39
40
41
42