European Journal of Political Economy 40 (2015) 1–15

Contents lists available at ScienceDirect

European Journal of Political Economy

journal homepage: www.elsevier.com/locate/ejpe

Financial crisis begets financial reform? The origin of the

crisis matters
Sébastien Waelti ⁎
International Monetary Fund, 700 19th St NW, Washington DC 20431, USA

a r t i c l e i n f o a b s t r a c t

Article history: The empirical literature assessing whether crisis begets reform has paid little attention to the
Received 28 March 2014 role of the origin of crises. This paper considers seven dimensions of financial reform, as well
Received in revised form 8 October 2015 as an aggregate index of these dimensions, and focuses on crises defined as sudden stops in fi-
Accepted 10 October 2015
nancial flows. Recent work on gross financial flows provides a suitable strategy to distinguish
Available online 19 October 2015
between domestic crises (sudden flight) and external crises (true sudden stops). While the or-
igin of the crisis plays no role in explaining the likelihood of reform in the case of the aggregate
JEL classification: index of financial reform, different origins of crises affect different individual dimensions of fi-
F32
nancial reform. Thus, the origin of the crisis matters when assessing the nexus between finan-
G28
cial crisis and financial reform. The evidence is particularly compelling for the case of sudden
P16
flight and reform of capital account restrictions, and more tentative for the case of true sudden
Keywords:
stops and reform of banking regulation and supervision. These findings underscore the need to
Reform
pay greater attention to the origin of crises in empirical work focusing on the relationship be-
Crisis
Sudden stop tween financial crisis and financial reform, and also in theoretical models of the political econ-
Financial flows omy of financial reform.
Ordered logit © 2015 Elsevier B.V. All rights reserved.

1. Introduction

The crisis-begets-reform hypothesis postulates that economic policy reforms are more likely to be implemented at times of
economic crisis. A crisis provides policymakers with an opportunity to carry out those reforms that could not be implemented
prior to the crisis. More specifically, the occurrence of a crisis shifts the existing political equilibrium and thereby increases the
likelihood of observing economic policy reforms. Importantly, the deterioration in the relevant economic conditions must be se-
vere enough for reform to become more likely. Thus, there is a sense that the relevant economic conditions must cross some
threshold for crisis to facilitate reform.
The purpose of this paper is to assess whether the origin of crises matters when assessing whether crisis begets reform.
Existing empirical studies have typically identified crises as dummy variables taking a value of one when some macroeconomic
variable crosses a given threshold, and zero otherwise. While this approach captures the idea that economic conditions must de-
teriorate significantly enough for crisis to facilitate reform, it fails to account for the origin of crises. In fact, a macroeconomic var-
iable may cross a given threshold either because of a domestic crisis, that is a home-grown crisis due to the failure of domestic
policies, or because of an external crisis, that is a crisis elsewhere that gets transmitted to the domestic economy through
trade and financial spillovers. These two different origins of crises may play different roles in determining the likelihood of reform
implementation. The possibility that different origins of crises have a different impact on the likelihood of reforms is the central

⁎ Tel.: +1 202 6235850; fax: +1 202 6235858.

E-mail address: swaelti@imf.org.

http://dx.doi.org/10.1016/j.ejpoleco.2015.10.002
0176-2680/© 2015 Elsevier B.V. All rights reserved.
2 S. Waelti / European Journal of Political Economy 40 (2015) 1–15

question of this paper. To my knowledge, this question has not been addressed in the empirical literature on the crisis-begets-
reform hypothesis.
In principle, an assessment of the role of the origin of crises would be based on an empirical strategy that provides for a clean
way to distinguish the domestic and external origins of many kinds of crises. In such a way, the analysis would speak to the broad
hypothesis that crisis begets reform. In practice, however, finding a well-grounded strategy to distinguish domestic and external
crises remains very difficult. For this reason, this paper focuses on the specific case of financial crises in the form of sudden stops
in financial flows. For this specific case, recent work on gross financial flows offers a suitable empirical strategy. The identification
of sudden stops in financial flows has traditionally been based on the behavior of net financial flows (Calvo et al., 2004). However,
Rothenberg and Warnock (2011) and Cowan et al. (2008) have shown that this identification strategy can mask very different
financial flow dynamics. A sudden stop in net financial inflows may arise either from a sudden stop in gross financial inflows
(a true sudden stop), or from a sudden rise in gross financial outflows (a sudden flight). How do these different financial flow
dynamics help distinguish between domestic and external crises? Empirical studies show that sudden flight is relatively more
closely related to domestic policy developments than true sudden stops, whereas true sudden stops are relatively more affected
by external developments than sudden flight. Thus, the relative importance of domestic and external factors in driving gross fi-
nancial flow dynamics underpins the distinction between domestic and external crises in this paper. Episodes of sudden flight
proxy for crises resulting from domestic policy failures, while episodes of true sudden stops proxy for crises resulting from exter-
nal developments, including policy failures elsewhere and global shocks.
The empirical approach to test whether domestic and external crises have a different impact of financial reform makes use of
the seven individual dimensions of financial reform, as well as the aggregate index of financial reform, provided by Abiad et al.
(2008). The seven dimensions and the aggregate index are regressed on both types of crises as well as a number of control var-
iables. The sample consists of 72 advanced and emerging market and developing economies over the period 1980–2005. While
estimation results for the aggregate index of financial reform do not indicate any difference in the effect of true sudden stops
and sudden flight on financial reform, I find that different origins of crises affect different individual dimensions of financial re-
form. The evidence is particularly compelling for the case of sudden flight and reform of capital account restrictions, and more
tentative for the case of true sudden stops and reform of banking regulation and supervision. Thus, the origin of the crisis matters
when assessing the nexus between financial crisis and financial reform. More generally, the results underscore the need to look
beyond the veil of aggregate indices of reform and to conduct a more focused assessment of the crisis-begets-reform hypothesis
by looking at narrower measures of reform.
The remainder of this paper is organized as follows. Section 2 reviews some of the key aspects of the empirical literature on
the crisis-begets-reform hypothesis. Section 3 outlines the empirical approach which borrows extensively from Abiad and Mody
(2005) for reasons outlined therein. Section 4 presents the results. The final section provides concluding remarks.

2. Selective review of the literature

Testing the crisis-begets-reform hypothesis requires measuring both reform and crisis. This section provides a selective review
of the empirical literature on this hypothesis.1 This review does not aim to give an exhaustive survey of the literature. Rather, it
serves as a motivation for some of the choices regarding the measurement of variables and the empirical specification used in this
paper.

2.1. Measuring reform

Reform refers to a process of change. The literature on the political economy of economic reforms does not aim to explain how
things are at a given point in time, but rather how things change (or do not change) over time. Reform can be measured indirect-
ly by looking at the change in some macroeconomic variable (outcome), e.g. economic growth or the inflation rate, or more di-
rectly by looking at the change in some measure of the economic policy regime, e.g. capital account openness or the exchange
rate regime.
The early empirical literature focused on macroeconomic outcomes as an indirect measure of reform given the lack of consis-
tent datasets on economic reforms at the time (Bruno and Easterly, 1998; Drazen and Easterly, 2001). With the recent develop-
ment of large datasets on reforms, including systematic information across countries and over time, more recent empirical testing
of the crisis hypothesis has relied on more direct measures of economic policy regimes. A more direct measure is probably supe-
rior as the mapping between reforms and macroeconomic outcomes is bound to be an imperfect one.
Economic policy regimes are often expressed as indices in recent datasets. Such indices aggregate a large amount of informa-
tion on a varying number of policy dimensions. For example, the Economic Freedom of the World index aggregates five policy
dimensions, including size of government, legal system and property rights, sound money, freedom to trade, and regulation. In
turn, each of these five dimensions comprises numerous components, leading up to a total of 42 distinct variables (Gwartney
et al., 2013). This index has been used by several scholars to test the crisis hypothesis as it captures a broad range of policy

1
Drazen (2000) reviews the theoretical arguments underpinning the crisis-begets-reform hypothesis (Chapter 10, 444–454).
 S. Waelti / European Journal of Political Economy 40 (2015) 1–15 3

changes in a single index (Pitlik and Wirth, 2003; De Haan et al., 2009).2 At the same time, the main advantage of aggregate in-
dices is also their main drawback. Different kinds or origins of crises may affect different policy dimensions over different time
horizons. Therefore, it would seem appropriate as a general empirical approach to consider not only an aggregate index of policy
reforms, but also to examine separately the individual policy dimensions that are used to construct the aggregate index. Indeed,
Rodrik (1996) cautions against lumping together different kinds of reforms and reviews cases where microeconomic policy fail-
ures have been blamed when unsustainable macroeconomic policies were at the root of crises. Similarly, Drazen (2009) stresses
that “for a crisis to trigger reform quickly, the reform must be seen as directly addressing the crisis”. Hence, the analysis in this
paper focuses on an aggregate index of reform as well as each of its individual dimensions. In this way, we are able to establish
whether the origin of crises matters in the sense that different origins of crises affect different policy dimensions.

2.2. Measuring crisis

Rodrik (1996) claims that the crisis-begets-reform hypothesis is a tautological statement insofar as a crisis is simply the logical
consequence of a failure of domestic policy.3 When domestic policy has failed, it will inevitably change. Drazen (2000) addresses
this critique by qualifying the hypothesis. Reform will be facilitated not when economic conditions are just bad, but rather when
they turn bad enough that a crisis occurs. Hence, there is a strong element of a threshold in the crisis-begets-reform hypothesis.
Conveniently, the usual way in which a crisis has been defined in the literature fully incorporates the idea of a threshold. For
example, currency crises are typically said to occur when a calculated exchange market pressure index exceeds some predefined
threshold (Eichengreen et al., 1995). Episodes of debt distress are said to occur when sovereign bond spreads exceed some arbi-
trary level (Pescatori and Sy, 2007). Stock market crashes are typically identified when the return on a given index of stocks falls
below some arbitrary value. Sudden stops are said to occur when net financial flows decrease by, say, more than one standard
deviation below their historical average (Calvo et al., 2004).
While the identification of crises based on thresholds is well established, we know much less about how to identify the origin
of a crisis. As a matter of principle, an assessment of the role played by the origin of crises should be based on an empirical strat-
egy that distinguishes domestic and external crises for many kinds of crises. In such a way, the results of the analysis would speak
to the broad hypothesis that crises beget reform. Unfortunately, to my knowledge such a general strategy is currently not avail-
able. Thus, the analysis in this paper focuses on one particular kind of crisis, namely sudden stops in financial flows, for which a
well-grounded strategy is available.
Sudden stops have typically been identified on the basis of net financial flows (see, for example, Calvo et al., 2004). However,
Rothenberg and Warnock (2011) and Cowan et al. (2008) have shown that this identification strategy may mask important dif-
ferences in financial flow dynamics. Sudden stops based on net flows can arise either from sudden stops in gross financial inflows
(true sudden stops), or from a sudden increase in gross financial outflows (sudden flight). Looking solely at net flows does not
allow us to understand whether the action takes place relatively more in gross financial inflows, reflecting actions by global in-
vestors, or in gross financial outflows, reflecting actions by local investors. For this purpose, looking at gross flows is necessary.
How does the distinction between true sudden stops and sudden flight help us identify the origin of a crisis? An array of em-
pirical studies have found consistent evidence that true sudden stops depend to a relatively greater extent on external factors, and
that sudden flight owes relatively more to domestic policy developments. Cowan et al. (2008) find that the bunching of true sud-
den stops over time suggests that they are driven more by events in international capital markets than sudden flight. Rothenberg
and Warnock (2011) also find bunching of true sudden stops over time, while sudden flight episodes are more dispersed over the
sample period. On this basis, they conclude that true sudden stops are best characterized as having a common component,
reflecting a global factor and perhaps also contagious effects, while sudden flight episodes are more likely to be driven by local
developments.4 Calderon and Kubota (2013) provide more systematic econometric evidence about the determinants of financial
flow episodes. Specifically, they explore whether true sudden stops and sudden flight episodes are explained by the same set of
economic determinants. They find that local macroeconomic conditions affect only episodes of sudden flight, but not true sudden
stops. Overall, therefore, it appears that true sudden stops are driven to a relatively greater extent by external factors, whereas
sudden flight is driven relatively more by local factors. This relative importance provides a natural distinction between domestic
and external crises. Sudden flight episodes proxy for domestic crises, whereas episodes of true sudden stops proxy for external
crises.

3. Empirical approach

The empirical framework used to test whether the origin of a crisis matters for financial crisis to beget financial reform bor-
rows extensively from Abiad and Mody (2005). This choice does not imply that their framework is necessarily and generally

2
De Haan et al. (2006) provide an appraisal of the strengths and weaknesses of the Economic Freedom of the World index. In particular, they emphasize two issues
with the construction of the index: the relevant components of the index, and their aggregation into a single index.
3
Rodrik (1996) offers a colorful analogy by stating that “That policy reform should follow crisis, then, is no more surprising than smoke following fire”.
4
Rothenberg and Warnock (2011) formalize this finding in a multi-asset, noisy rational expectations model including asymmetric information across global inves-
tors and local investors. In this setup, they show that true sudden stops arise because global investors pull out of all markets because of a negative signal about global
markets, or because they misinterpret and overreact to a perceived negative signal about a specific market. In contrast, sudden flight results from locals exiting their
markets because of a negative private signal about the local economy.
4 S. Waelti / European Journal of Political Economy 40 (2015) 1–15

the best specification to test the broader crisis hypothesis. Rather, their framework focuses on the determinants of financial re-
form and as such, it offers a meaningful empirical specification for the specific purpose of this paper.
Abiad and Mody (2005) postulate that the likelihood of reform depends on domestic and regional learning processes, crises,
and political and structural factors. Domestic learning is captured through a reduced-form process whereby financial reform, de-
noted as ΔFLit, is a function of the gap between a given desired level of financial liberalization, denoted as FLit⁎, and the existing
level of financial liberalization, denoted as FLi,t − 1. Thus, ΔFLit = γ(FLit⁎ − FLi,t − 1) + εit. The coefficient γ measures the status
quo bias and is a function of the existing level of financial liberalization, such that γ = αFLi,t − 1.5 Assuming FLit⁎ = 1, domestic
learning can thus be captured by ΔFLit = αFLi,t – 1(1 − FLi,t – 1) + εit. Regional learning (or diffusion) is the idea that countries
within a region may be induced to catch up with the leader(s) in the region, for example due to a reduction in uncertainty re-
garding the benefits of financial reform (Fernandez and Rodrik, 1991). Regional learning is captured by the gap between a region's
leader in financial liberalization, denoted as REGFLi,t − 1, and a country's existing level of financial liberalization, FLi,t – 1. Crisis is
captured by a dummy variable equal to one whenever a crisis occurs, and zero otherwise. Again, the binary measurement of crises
captures the idea that economic conditions must deteriorate enough (beyond a given threshold) for reform to occur. Finally, po-
litical and structural factors are captured by a set of controls.
Put together, the empirical specification of Abiad and Mody (2005) is written as



Δ FLit ¼ α FLi;t−1 1− FLi;t−1 þ β REGFLi;t−1 − FLi;t−1
X
2
þ γk STOP i;t−k ð1Þ
k¼1
þ ϕPOLITICALit þ ψSTRUCTUREit þ ηi þ λt þ εit

where ΔFLit is the change in some index of financial liberalization, the first line captures domestic and regional learning processes,
the second line picks up crises in the specific form of sudden stops in my paper, and the third line includes political and economic
controls. All regression specifications include a set of country fixed effects and time fixed effects. Expressing the dependent var-
iable as the first difference of the index of financial liberalization conveys the sense that financial reform is a process of change.
The set of political variables includes a dummy variable for the first year of government, dummy variables for the political orien-
tation of the executive power (left, right, center omitted), and an index of the level of democracy. The structural factors consist of
trade openness measured as the sum of exports and imports of goods and services over gross domestic product, and a dummy
variable for the presence of an IMF program. Table A2 in the Appendix lists data sources.
Abiad and Mody (2005) construct dummy variables for crises, taking a value of one whenever a crisis occurred during the past
two years. Besides addressing the potential problem of reverse causality, the rationale for lags is that crises may not affect the
adoption of reforms contemporaneously. Drazen (2009) argues that the political equilibrium may take some time to shift, in
other words that support for reform must “percolate”. In contrast to Abiad and Mody (2005), however, both the first lag and
the second lag of the sudden stop variable are entered separately in my analysis, given uncertainty about the appropriate lag
length.
My main contribution is to separate out the STOPit variable to distinguish between the different origins of crises. Specifically,
the econometric model is modified as



Δ FLit ¼ α FLi;t−1 1− FLi;t−1 þ β REGFLi;t−1 − FLi;t−1
X
2 X
2
þ δk TRUESTOP i;t−k þ ωk FLIGHT i;t−k ð2Þ
k¼1 k¼1
þ ϕPOLITICALit þ ψSTRUCTUREit þ ηi þ λt þ εit

where TRUESTOPit is equal to one when a true sudden stop occurs, and zero otherwise, and FLIGHTit is equal to one when sudden
flight occurs, and zero otherwise.
The sample consists of 72 advanced and emerging and developing economies over a time period ranging from 1980 to 2005.6
The reform variable is constructed using the dataset of financial reforms of Abiad et al. (2008). They propose a multi-dimensional
index of financial liberalization including seven dimensions: credit controls and reserve requirements, interest rate controls, entry
barriers, state ownership, policies on securities markets, banking regulation and supervision, and restrictions on the capital ac-
count. Liberalization scores between 0 and 3 are assigned for each individual dimension and then combined into an aggregate
index with a maximum value of 21. The selective review of the empirical literature above made the case for focusing not only
on an aggregate index of financial reform but also on the individual dimensions of financial reform. Thus, in this paper, all

5
One possible theoretical reason why the status quo bias should be strongest when financial liberalization is lowest is proposed by Fernandez and Rodrik (1991).
Initial reforms help identify who are the winners and losers of reforms and thus facilitate the adoption of further reforms. Abiad and Mody (2005) also assert that this
modeling is consistent with the strengthening of outside groups relative to incumbents once the reform process starts, as well as with the need to build up technical and
managerial expertise in reform implementation.
6
See Table A1 in the Appendix for the list of economies grouped by region.
 S. Waelti / European Journal of Political Economy 40 (2015) 1–15 5

seven individual dimensions as well as the aggregate index of financial reform are used, one at a time, to construct eight separate
dependent variables. For the individual dimensions, an increase in FLit indicates a reform; no change in FLit amounts to a status
quo; and a decrease in FLit is a reform reversal. For the aggregate index, an increase in FLit by 3 points or more is a large reform;
an increase in FLit by 1 or 2 points is a reform; no change in FLit amounts to a status quo; a decrease by 1 or 2 points is a reversal;
and a decrease by 3 points or more is a large reversal. Since the liberalization scores have an ordinal (but not a cardinal) inter-
pretation, I follow Abiad and Mody (2005) and estimate the econometric model by the ordered logit method.
The identification of sudden stops and the distinction between true sudden stops and sudden flight is based on the work of
Rothenberg and Warnock (2011). They propose a two-step approach. As a first step, traditional sudden stop episodes are identi-
fied on the basis of net financial flows. Sudden stops occur whenever the change in net financial flows over GDP is below one
standard deviation of their historical average.7 This is the first definition of sudden stops used in my paper. This definition can
be tightened further by adding the constraint that the fall in net financial inflows over GDP exceeds five percent of GDP.
Guidotti et al. (2004) suggest that this extra (yet arbitrary) constraint helps excluding countries that display very small changes
in net financial flows while still exceeding one standard deviation due to historically low volatility. This is the second, more strin-
gent, definition used in my paper as a sensitivity analysis. As a second step, for each country-year for which a traditional sudden
stop has been identified, the behavior of gross financial flows determines whether we have a true sudden stop or sudden flight.
More specifically, true sudden stops occur when the decrease in gross financial inflows outweighs the increase in gross financial
outflows. Reciprocally, sudden flight occurs when the increase in gross financial outflows outweighs the decrease in gross finan-
cial inflows.
Annual data on net and gross financial flows are obtained from the IMF's International Financial Statistics. Financial flows con-
sist of foreign direct investment flows, portfolio debt and equity flows, and other investment flows. Data on financial flows in the
early part of the sample feature a number of gaps. I follow closely Forbes and Warnock (2012) and make a small number of data
adjustments, only where such adjustments are straightforward. First, data gaps surrounded by zeros are filled in with zeros. How-
ever, data gaps occurring at the beginning of the sample are treated as such. Second, some data gaps are filled by consulting
country-specific information, e.g. Korea and Turkey in the 1980s. Finally, some countries are dropped from the sample altogether
as they feature too many missing observations, e.g. Greece.8

4. Results

There has been a steady increase in the degree of financial liberalization across all geographical regions over the sample period
(Abiad et al., 2008). Figs. A1–A8 in the Appendix show that the number of reform episodes exceeds the number of reversal ep-
isodes for each individual policy dimension (and thus the aggregate index) for most of the years in the sample. Thus, the degree
of financial liberalization has increased over time. The incidence of financial reform has generally peaked across all individual re-
form dimensions during the 1990s. All this being said, the data display a strong bias toward the status quo. This evidence is in line
with the extensive literature on the political economy of reform, inaction and delay. This literature has come up with several
strands of theoretical contributions to explain the suboptimal persistence of the status quo (Drazen, 2000).
Sudden stops have not been rare over the sample period. The standard definition of sudden stops based on net financial flows
identifies 220 sudden stops (about 13% of country-year observations), of which 164 are true sudden stops and the remaining 56
correspond to sudden flight. The more stringent definition of sudden stops identifies, as expected, less sudden stops. Specifically,
this latter definition identifies 122 sudden stops, of which 94 are true sudden stops and the remaining 28 represent sudden flight
episodes. Fig. A9 in the Appendix displays the distribution of sudden stops over time. This graphical evidence suggests clustering
of true sudden stop episodes over time. A cursory look would point to the debt crisis in the early 1980s, and emerging market
turbulence in the second half of the 1990s and early 2000s. While true sudden stops occur around specific years, sudden flight
is more scattered over the sample period. This evidence lends support to the empirical strategy used in this paper, whereby
true sudden stops are more closely linked to external developments, while sudden flight is more closely related to domestic
developments.
The key finding emerging from the baseline results presented in Tables 1–2 is that different origins of crises have an effect on
different individual dimensions of financial reform.9 Thus, the origin of the crisis matters when assessing whether financial crisis

7
The use of averages and standard deviations has a long tradition in the crisis identification literature. The empirical literature on sudden stops has followed this tra-
dition. This being said, other approaches could also be explored, such as using median absolute deviations. Alternative approaches are likely to produce similar sets of
crisis episodes as crises are such unusual events. I am grateful to an anonymous referee for suggesting this point. Historical averages and standard deviations are com-
puted over the entire sample period. Alternatively, these could have been calculated using data up to the time period under consideration. Unfortunately, computing
historical averages and standard deviations up to the time period under consideration would imply the exclusion of annual data in the early part of sample period. Doing
this would mean throwing away important variation in the data as there have been many crises and reform episodes across countries in that part of the sample period.
In the end, therefore, considering the entire sample period avoids losing a non-negligible amount of information.
8
The full set of data adjustments is available from the author upon request.
9
Tables 1–2 display odds ratios for the second lag of sudden stops, true sudden stops and sudden flight, as the first lag is generally not statistically different from zero.
Intuitively, odds ratios can be interpreted as the ratio of the probability of reform over the probability of no reform. Odds ratios range between zero and infinity and can
be calculated from estimated coefficients, that is odds ratio = ecoefficient.
6 S. Waelti / European Journal of Political Economy 40 (2015) 1–15

Table 1
Baseline results: aggregate index, capital account restrictions, credit controls, interest rate controlsa.

Aggregate index of reform Capital account restrictions Credit controls Interest rate controls

(I) (II) (I) (II) (I) (II) (I) (II)

Status quo bias 766.899⁎⁎⁎ 754.035⁎⁎⁎ 3.492 3.437 80.963⁎⁎ 80.772⁎⁎ 14,779.34⁎⁎⁎ 13,325.37⁎⁎⁎
(850.538) (836.687) (4.941) (4.908) (141.358) (140.786) (47,189.88) (42,332.37)
Reg diffusion 1310.856⁎⁎⁎ 1318.353⁎⁎⁎ 219.488⁎⁎⁎ 225.746⁎⁎⁎ 95.894⁎⁎⁎ 95.928⁎⁎⁎ 10,248.77⁎⁎⁎ 11,109.41⁎⁎⁎
(858.264) (867.411) (116.369) (120.818) (44.487) (44.529) (13,188.88) (14,732.00)
Sudden stop 1.491⁎⁎ 1.366 1.061 2.121⁎⁎ (0.775)
(0.257) (0.357) (0.310)
True sudden 1.454⁎ 1.130 0.998 (0.339) 2.268⁎⁎ (0.932)
stop (0.287) (0.304)
Sudden flight 1.582⁎ (0.413) 2.789⁎⁎⁎ (1.096) 1.352 (0.539) 1.265 (0.740)
Open 0.996 (0.006) 0.996 (0.006) 0.983⁎⁎ (0.007) 0.983⁎⁎ (0.007) 1.002 (0.010) 1.002 (0.010) 0.994 (0.010) 0.994 (0.011)
First year 1.196 (0.173) 1.198 (0.173) 1.269 (0.306) 1.287 (0.308) 1.380 (0.355) 1.384 (0.357) 1.506 (0.546) 1.493 (0.545)
Left 1.516⁎ (0.339) 1.513⁎ (0.339) 1.980 (0.858) 2.010 (0.874) 0.969 (0.417) 0.977 (0.424) 1.752 (1.151) 1.733 (1.152)
Right 2.080⁎⁎⁎ (0.399) 2.080⁎⁎⁎ (0.398) 1.559 (0.544) 1.573 (0.555) 1.392 (0.551) 1.395 (0.553) 3.634⁎⁎ (1.923) 3.559⁎⁎ (1.906)
Democ 0.990 (0.027) 0.989 (0.027) 1.027 (0.035) 1.029 (0.035) 0.990 (0.043) 0.990 (0.043) 1.011 (0.066) 1.007 (0.066)
IMF 2.059⁎⁎⁎ (0.446) 2.053⁎⁎⁎ (0.443) 1.723 (0.590) 1.704 (0.575) 0.969 (0.269) 0.965 (0.269) 2.550⁎⁎ (1.102) 2.556⁎⁎ (1.120)
Observations 1554 1554 1554 1554 1554 1554 1554 1554
Countries 72 72 72 72 72 72 72 72
Time effects Yes Yes Yes Yes Yes Yes Yes Yes
Fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Pseudo R2 stat 0.138 0.138 0.206 0.208 0.203 0.203 0.407 0.408
p-Value for – 0.77 – 0.02 – 0.53 – 0.43
or equality

Country-clustered standard errors in parentheses. The p-value in the last line refers to the test for the equality of odds ratios for true sudden stops and sudden
flight.
a
All odds ratios are obtained using the ordered logit estimator.
⁎ Significant at 10% level.
⁎⁎ Significant at 5% level.
⁎⁎⁎ Significant at 1% level.

begets financial reform. First, considering the estimation results for the aggregate index of reform, we find that traditional
sudden stops increase the likelihood of financial reform (column (I)) with no significant difference between true sudden
stops and sudden flight (column (II)). The p-value for the test of equality of odds ratios largely exceeds 0.05. On this basis,

Table 2
Baseline results: entry barriers, banking regulation and supervision, state ownership, policies on securities marketsa.

Entry barriers Regulation and supervision State ownership Policies on securities markets

(I) (II) (I) (II) (I) (II) (I) (II)

Status quo bias 6394.543⁎⁎⁎ 6283.37⁎⁎⁎ 0.986 (2.163) 0.890 (1.988) 67.526⁎⁎ 59.678⁎⁎ 2163.774⁎⁎⁎ 2109.115⁎⁎⁎
(18,462.26) (18,262.21) (136.025) (118.070) (5086.561) (4872.616)
Reg diffusion 17,686.51⁎⁎⁎ 17,334.40⁎⁎⁎ 13,200.77⁎⁎⁎ 12,562.07⁎⁎⁎ 373.516⁎⁎⁎ 381.152⁎⁎⁎ 308,704.0⁎⁎⁎ 303,487.8⁎⁎⁎
(29,539.71) (28,826.05) (14,112.46) (13,449.43) (266.220) (275.620) (545,487.4) (528,353.2)
Sudden stop 1.191 (0.468) 2.157⁎⁎ (0.692) 1.784⁎ (0.548) 1.155 (0.627)
True sudden 1.520 (0.655) 2.694⁎⁎⁎ (0.860) 1.673 (0.548) 1.078 (0.700)
stop
Sudden flight 0.379 (0.345) 0.868 (0.808) 2.123 (1.180) 1.409 (1.191)
Open 0.995 (0.017) 0.995 (0.016) 0.991 (0.011) 0.990 (0.011) 1.000 (0.014) 1.000 (0.014) 1.003 (0.014) 1.003 (0.014)
First year 1.738⁎ (0.571) 1.715 (0.566) 1.264 (0.360) 1.244 (0.359) 0.728 (0.221) 0.747 (0.225) 0.914 (0.376) 0.923 (0.373)
Left 1.783 (1.025) 1.686 (0.976) 1.221 (0.694) 1.163 (0.652) 1.342 (0.524) 1.326 (0.519) 0.641 (0.438) 0.645 (0.442)
Right 2.627 (1.640) 2.622 (1.617) 1.539 (0.801) 1.504 (0.769) 3.343⁎⁎⁎ (1.432) 3.359⁎⁎⁎ (1.414) 0.877 (0.483) 0.881 (0.484)
Democ 0.937 (0.054) 0.933 (0.053) 0.947 (0.061) 0.945 (0.060) 0.991 (0.036) 0.990 (0.035) 1.038 (0.073) 1.038 (0.073)
IMF 3.464⁎⁎ (1.771) 3.408⁎⁎ (1.761) 1.455 (0.671) 1.430 (0.658) 1.839⁎ (0.632) 1.825⁎ (0.610) 1.369 (0.661) 1.358 (0.650)
Observations 1554 1554 1554 1554 1554 1554 1554 1554
Countries 72 72 72 72 72 72 72 72
Time effects Yes Yes Yes Yes Yes Yes Yes Yes
Fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Pseudo R2 stat 0.364 0.367 0.295 0.297 0.235 0.240 0.363 0.363
p-Value for OR – 0.15 – 0.23 – 0.17 – 0.79
equality

Country-clustered standard errors in parentheses. The p-value in the last line refers to the test for the equality of odds ratios for true sudden stops and sudden
flight.
a
All odds ratios are obtained using the ordered logit estimator.
⁎ Significant at 10% level.
⁎⁎ Significant at 5% level.
⁎⁎⁎ Significant at 1% level.
 S. Waelti / European Journal of Political Economy 40 (2015) 1–15 7

we would conclude that the origin of the crisis does not play any role in assessing whether financial crisis begets financial
reform. Second, however, turning to the estimation results for the individual dimensions of financial reform, we find that
true sudden stops have a statistically significant effect on the likelihood of reform of interest rate controls as well as reform
of banking regulation and supervision, while sudden flight has a statistically significant effect on the likelihood of reform of
capital account restrictions. 10 It follows that the origin of the crisis matters importantly since different origins of crises —
whether domestic or external — explain different individual financial reforms. More generally, this result underscores the
need to look beyond the veil of aggregate reform indices. The evidence is especially strong for the case of capital account re-
strictions, for which the null hypothesis of equality of odds ratios is clearly rejected at the 5% level, and somewhat weaker for
the cases of interest rate controls and banking regulation and supervision, for which this same null hypothesis cannot be
rejected. This finding may not be entirely surprising in the context of sudden stops in cross-border financial flows. Indeed,
capital account restrictions are the dimension which is probably most closely linked to sudden stops among the seven indi-
vidual dimensions of financial reform.
Interestingly, the evidence really points out to a higher likelihood of financial reform and not some form of panic reaction to
the financial crisis. In response to true sudden stops, policymakers have generally reduced interest rate controls, and introduced
sounder regulatory and supervisory measures for the banking sector. Also, in response to sudden flight, policymakers have relaxed
restrictions on the capital account. The direction of these responses indicates financial reform. A panic reaction to the crisis would
have entailed increased interest rate controls, regulatory and supervisory forbearance, and tighter restrictions on the capital
account.11
To get an idea about the economic significance of the results, Specification (II) for capital account restrictions is used as an
illustration. The estimated odds ratio for sudden flight is equal to 2.789. In words, this means that for a unit increase in sud-
den flight (going from 0 to 1), the odds of reform versus the other two categories are about 2.8 times greater, keeping all
other variables constant. This result points to strong economic significance on top of strong statistical significance.
The estimated odds ratios for the other explanatory variables behave largely as expected. For the aggregate index as well as
some individual dimensions, the positive and statistically significant estimate on the variable picking up status quo bias means
that the larger the existing level of financial liberalization, the greater is the likelihood of observing a change in financial lib-
eralization. In other words, once the process of financial reform has started, it becomes more and more likely that it will con-
tinue. For the aggregate index and all individual dimensions, the positive and statistically significant odds ratio for regional
diffusion is consistent with the idea that countries in a given geographical region that are farther away from the maximum
level of financial liberalization in that region are more likely to adopt financial reform, other things equal.12 Estimated odds
ratios for the other explanatory variables are generally not significant.13 There is some evidence that right-wing governments
are more likely to reduce interest rate controls and also to reduce state ownership in the financial sector.14 The presence of an
IMF program increases the likelihood of reducing interest rate controls and also of lowering entry barriers into the financial
sector.
The key finding that the origin of the crisis matters when assessing the nexus between financial crisis and financial reform is
generally robust to using the more stringent definition of sudden stop episodes, constraining the fall in net financial inflows
over GDP to exceed the arbitrary threshold of five percent of GDP. Table 3 shows that true sudden stops have a statistically sig-
nificant impact on the likelihood of reform of banking regulation and supervision (but not of interest rate controls as in the
baseline results), while sudden flight affects the likelihood of reform of capital account restrictions. Thus, we continue to
find evidence that different origins of crises have an effect on different individual dimensions of financial reform. This being
said, this evidence is not as strong as in the baseline results given that the null hypothesis of equality of odds ratios for capital
account restrictions can be rejected only at the 10% level of significance. In this case, however, it is worth keeping in mind that
we have only 28 episodes of sudden flight (out of 1554 observations) for the more stringent definition. As a result, the precision
of the estimates may be reduced to a significant extent. Finally, we also find evidence using the more stringent definition that
sudden flight reduces the likelihood of lower entry barriers into the financial sector. This result, which is not robust across dif-
ferent definitions of sudden stop episodes, is probably not saying much about reform but probably reflects some kind of panic
reaction to the financial crisis.

10
All episodes of true sudden stops and sudden flight are included for the estimations. There is a small number of cases where episodes occur consecutively in a given
country. In most of these cases, a true sudden stop (sudden flight) is followed by another true sudden stop (sudden flight) in the following year. In only four cases, a
sudden flight is followed by a true sudden stop (three episodes) or a true sudden stop is followed by a sudden flight (one episode). As a robustness check, I have esti-
mated my baseline specifications such that whenever an episode in a given year is followed by another episode in the following year, the latter episode is excluded. My
baseline results remain unchanged.
11
I am grateful to an anonymous referee for suggesting this point to me.
12
This result pointing out the significant role of regional diffusion is in line with the evidence in Pitlik (2007) and Gassebner et al. (2011). In particular,
Gassebner et al. (2011) find that geography plays an important role in regional diffusion, thereby lending support to the measure of regional diffusion of Abiad
and Mody (2005).
13
Measuring trade as the sum of exports and imports over GDP is a simple measure of economic openness. Other measures of globalization, such of the KOF Global-
ization Index, are also available. The KOF indices cover three dimensions: economic globalization, social globalization and political globalization. I computed my baseline
results with each of these three dimensions separately, as well as with an aggregate index, as a robustness check. The results remain unchanged.
14
Abiad and Mody (2005) enter political controls into the specification contemporaneously. Entering these with one lag does not affect my results.
8 S. Waelti / European Journal of Political Economy 40 (2015) 1–15

Table 3
More stringent definition of sudden stopsa.

Aggregate Capital account Credit Interest rate Entry barriers Regulation and State Securities
index of reform restrictions controls controls supervision ownership markets

(II) (II) (II) (II) (II) (II) (II) (II)

Status quo 753.223⁎⁎⁎ 3.561 75.714⁎⁎ 14421.64⁎⁎⁎ 6367.830⁎⁎⁎ 0.859 (1.856) 54.348⁎⁎ 2191.415⁎⁎⁎
(835.937) (5.089) (132.019) (46,565.48) (18,905.12) (108.102) (5151.123)
Reg diffusion 1306.030⁎⁎⁎ 219.527⁎⁎⁎ 95.512⁎⁎⁎ 9902.938⁎⁎⁎ 19,174.75⁎⁎⁎ 12,672.94⁎⁎⁎ 374.048⁎⁎⁎ 348,473.7⁎⁎⁎
(867.331) (117.411) (44.349) (13,104.14) (32,227.98) (13,584.22) (278.868) (634,882.3)
True sudden 1.438 (0.372) 1.382 (0.478) 1.222 (0.487) 1.767 (0.928) 2.113 (1.497) 3.348⁎⁎⁎ (1.513) 1.628 (0.694) 0.949 (0.770)
stop
Sudden flight 1.916⁎ (0.663) 3.740⁎⁎ (1.948) 1.909 (1.097) 1.182 (1.344) 0.062⁎⁎⁎ (0.039) 0.875 (0.951) 2.349 (1.773) 0.540 (0.658)
Open 0.996 (0.005) 0.982⁎⁎ (0.007) 1.002 (0.010) 0.995 (0.010) 0.998 (0.017) 0.990 (0.011) 1.001 (0.014) 1.003 (0.015)
First year 1.188 (0.172) 1.275 (0.303) 1.381 (0.352) 1.558 (0.545) 1.740⁎ (0.562) 1.298 (0.370) 0.720 (0.212) 0.898 (0.368)
Left 1.524⁎ (0.345) 2.028 (0.879) 0.974 (0.426) 1.800 (1.199) 1.793 (1.046) 1.257 (0.707) 1.304 (0.497) 0.609 (0.415)
Right 2.092⁎⁎⁎ (0.400) 1.581 (0.553) 1.403 (0.559) 3.529⁎⁎ (1.918) 2.574 (1.604) 1.601 (0.808) 3.263⁎⁎⁎ (1.327) 0.844 (0.461)
Democ 0.989 (0.027) 1.030 (0.034) 0.990 (0.043) 1.008 (0.066) 0.930 (0.054) 0.945 (0.057) 0.991 (0.036) 1.039 (0.072)
IMF 2.077⁎⁎⁎ (0.438) 1.653 (0.559) 0.959 (0.270) 2.629⁎⁎ (1.164) 3.542⁎⁎ (1.826) 1.415 (0.653) 1.926⁎⁎ (0.637) 1.382 (0.672)
Observations 1554 1554 1554 1554 1554 1554 1554 1554
Countries 72 72 72 72 72 72 72 72
Time effects Yes Yes Yes Yes Yes Yes Yes Yes
Fixed effects Yes Yes Yes Yes Yes Yes Yes Yes
Pseudo R2 stat 0.137 0.208 0.205 0.406 0.372 0.297 0.243 0.364
p-Value for 0.46 0.10 0.49 0.74 0.00 0.22 0.66 0.69
OR equality

Country-clustered standard errors in parentheses. The p-value in the last line refers to the test for the equality of odds ratios for true sudden stops and sudden
flight.
a
All odds ratios are obtained using the ordered logit estimator.
⁎ Significant at 10% level.
⁎⁎ Significant at 5% level.
⁎⁎⁎ Significant at 1% level.

5. Concluding remarks

The empirical literature on the political economy of reform has paid little attention, if any, to the role of the origin of crises.
Different origins of crises may affect different dimensions of reform. The specific purpose of this paper is to assess the extent
to which the origin of crises matters when assessing whether financial crisis begets financial reform. This assessment is carried
out by using seven individual dimensions of financial reform, together with an aggregate index of these dimensions. The analysis
focuses on the particular case of sudden stops in financial flows as recent work on gross financial flows provides a well-grounded
strategy to distinguish domestic crises and external crises.
The key finding is that the origin of the crisis matters when we assess the relationship between financial crisis and financial reform.
While estimation results for an aggregate index of financial reform suggest no difference in the effect of domestic and external crises
on financial reform, I also find that different origins of crises affect different individual dimensions of financial reform. The evidence is
especially compelling for the case of sudden flight and reform of capital account restrictions, and more tentative for the case of true
sudden stops and reform of banking regulation and supervision. These results generally underscore the need to extend the analysis of
the crisis-begets-reform hypothesis to narrower measures of reform. More broadly, my results add to the existing evidence in the lit-
erature supporting the idea that crises beget reform.
The fact that the respective actions of local investors and global investors have a different effect on different dimensions of
financial reform calls for greater attention to this issue also in theoretical work. Closed-economy models of the political economy
of reform are ill-equipped to deal with certain kinds of crises, such as sudden stops in financial flows. But even open-economy
models in this realm may be further developed to better understand the different role that local as opposed to global investors
play in fostering financial reform in the aftermath of financial crises.
Clearly, more work remains needed to broaden the analysis of the origin of crises to other types of crises and other types of reform.
This paper focuses squarely on sudden stops in financial flows as this is one type of crisis for which we have a well-grounded strategy
to identify the origin of crises. In this sense, this paper provides a first step. Further steps will be needed to determine whether the role
played by the origin of crises extends to other types of crises and reform. I leave these further steps for future research.

Acknowledgments

Part of the work on this paper was carried out while the author was at the Swiss National Bank. The views expressed herein
are those of the author and should not be attributed to the IMF, its Executive Board or its management, or to the Swiss National
Bank. I am very grateful for helpful comments and suggestions to the Editor, three anonymous referees, Hans Genberg, Daniel
Heller, and participants at the Ninth Annual NBP–SNB Joint Seminar. The usual disclaimer applies.
 S. Waelti / European Journal of Political Economy 40 (2015) 1–15 9

Appendix A

Table A1
Sample of 72 countries.

Advanced economies Emerging Asia Latin America Sub-saharan Africa Transition economies MENA countries

Australia Bengladesh Argentina Cameroon Bulgaria Egypt

Austria China Bolivia Cote d’Ivoire Czech Republic Jordan
Canada India Brazil Ghana Estonia Morocco
Denmark Indonesia Chile Kenya Hungary Pakistan
Finland Korea Colombia Mozambique Latvia Tunisia
France Malaysia Costa Rica Senegal Lithuania Turkey
Germany Philippines Dominican Republic South Africa Poland
Ireland Singapore Ecuador Tanzania Romania
Israel Sri Lanka El Salvador Uganda Russia
Italy Thailand Guatemala Ukraine
Japan Jamaica
Netherlands Mexico
New Zealand Nicaragua
Norway Paraguay
Portugal Peru
Spain Uruguay
Sweden Venezuela
Switzerland
United Kingdom
United States

Table A2
Data sources.

Financial reform indicators Abiad, Detragiache and Tressel (2008)

Gross capital flows IMF International Financial Statistics (various issues)
Gross domestic product IMF World Economic Outlook database
Trade openness World Developments Indicators, World Bank
First year of government in office Database of Political Institutions (DPI2012)
Party orientation (left, center, right) Database of Political Institutions (DPI2012)
Democracy index Polity IV Project, Center for Systemic Peace
IMF program IMF History of Lending Arrangements
Globalization index KOF Swiss Economic Institute
10 S. Waelti / European Journal of Political Economy 40 (2015) 1–15

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Large reversal Reversal Status quo Reform Large reform

Fig. A1. Aggregate index of financial reform.

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Reversal Status quo Reform

Fig. A2. Capital account restrictions.

 S. Waelti / European Journal of Political Economy 40 (2015) 1–15 11

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Reversal Status quo Reform

Fig. A3. Credit controls and reserve requirements.

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Reversal Status quo Reform

Fig. A4. Interest rate controls.

12 S. Waelti / European Journal of Political Economy 40 (2015) 1–15

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Reversal Status quo Reform

Fig. A5. Entry barriers.

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Reversal Status quo Reform

Fig. A6. Banking regulation and supervision.

 S. Waelti / European Journal of Political Economy 40 (2015) 1–15 13

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Reversal Status quo Reform

Fig. A7. State ownership.

100%

90%

80%

70%

60%

50%

40%

30%

20%

10%

0%

Reversal Status quo Reform

Fig. A8. Policies on securities markets.

14 S. Waelti / European Journal of Political Economy 40 (2015) 1–15

(a) Standard definition

20

18

16

14

12

10

True sudden stops Sudden flight

(b) More stringent definition

20

18

16

14

12

10

True sudden stops Sudden flight

Fig. A9. The distribution of sudden stops over time.

 S. Waelti / European Journal of Political Economy 40 (2015) 1–15 15

References

Abiad, A., Mody, A., 2005. Financial reform: what shakes it? What shapes it? Am. Econ. Rev. 95, 66–88.
Abiad, A., Detragiache, E., Tressel, T., 2008. A new database of financial reforms. IMF Working Paper 08/266.
Bruno, M., Easterly, W., 1998. Inflation crises and long-run growth. J. Monet. Econ. 41, 3–26.
Calderon, C., Kubota, M., 2013. Sudden stops: are global and local investors alike? J. Int. Econ. 89, 122–142.
Calvo, G., Izquierdo, A., Mejia, L.F., 2004. On the empirics of sudden stops: the relevance of balance-sheet effects. NBER Working Paper 10520.
Cowan, K., De Gregorio, J., Micco, A., Neilson, C., 2008. Financial diversification, sudden stops, and sudden starts. In: Edwards, S., Valdes, R., Cowan, K. (Eds.), Current
Account and External Financing. Central Bank of Chile, Santiago, Chile, pp. 159–194.
De Haan, J., Lundström, S., Sturm, J.E., 2006. Market-oriented institutions and policies and economic growth: a critical survey. J. Econ. Surv. 20, 157–191.
De Haan, J., Sturm, J.E., Zandberg, E., 2009. The impact of financial and economic crises on economic freedom. In: Gwartney, J., Lawson, R., Hall, J. (Eds.), Economic Free-
dom of the World: 2009 Annual Report. Fraser Institute, pp. 25–36.
Drazen, A., 2000. Political Economy in Macroeconomics. Princeton University Press, Princeton.
Drazen, A., 2009. Financial market crisis, financial market reform: why hasn't reform followed crisis. DG ECFIN Sixth Annual Research Conference, Brussels.
Drazen, A., Easterly, W., 2001. Do crises induce reform? Simple empirical tests of conventional wisdom. Econ. Polit. 13, 129–157.
Eichengreen, B., Rose, A., Wyplosz, C., 1995. Exchange market mayhem: the antecedents and aftermath of speculative attacks. Econ. Policy 21, 249–312.
Fernandez, R., Rodrik, D., 1991. Resistance to reform: status quo bias in the presence of individual-specific uncertainty. Am. Econ. Rev. 81, 1146–1155.
Forbes, K., Warnock, F., 2012. Capital flow waves: surges, stops, flight and retrenchment. J. Int. Econ. 88, 235–251.
Gassebner, M., Gaston, N., Lamla, M., 2011. The inverse domino effect: are economic reforms contagious? Int. Econ. Rev. 52, 183–200.
Guidotti, P., Sturzenegger, F., Villar, A., 2004. On the consequences of sudden stops. Economia 4, 171–214.
Gwartney, J., Lawson, R., Hall, J., 2013. Economic Freedom of the World: 2013 Annual Report. Fraser Institute.
Pescatori, A., Sy, A., 2007. Are debt crises adequately defined? IMF Staff Papers. 54, pp. 306–337
Pitlik, H., 2007. A race to liberalization? Diffusion of economic policy reform among OECD economies. Public Choice 132, 159–178.
Pitlik, H., Wirth, S., 2003. Do crises promote the extent of economic liberalization?: an empirical test. Eur. J. Polit. Econ. 19, 565–581.
Rodrik, D., 1996. Understanding economic policy reform. J. Econ. Lit. 34, 9–41.
Rothenberg, A., Warnock, F., 2011. Sudden flight and true sudden stops. Rev. Int. Econ. 19, 509–524.