sunk costs and exporting behavior: a sectoral analysis
TRANSCRIPT
1
Sunk Costs and Exporting Behavior: A Sectoral Analysis
Kurmaş Akdoğan1 & Laura M. Werner2,3
11 June 2019
ABSTRACT:
This article examines the hysteresis behavior due to sunk costs in exports of the Turkish
manufacturing sector. The results of the analysis using the Preisach method for 2006Q1 to 2018Q2
reveal hysteresis for only one sector: The manufacturing of wearing apparel, dressing and dyeing of
fur (clothing). To shed more light on this result we provide detailed information on the multi-layered
production structure of the clothing sector. We argue that the sub-contracting capacity of
intermediaries with their previous export experience and established connections, low importance of
plant size in the entry decision, easier financing conditions and price advantage due to a real
exchange rate depreciation are the main determinants of relatively lower sunk costs in this sector.
KEYWORDS: Nonlinearity, Path-dependency, Exports.
JEL: C19, F14, L60
1 Central Bank of the Republic of Turkey, Structural Economic Research Department, Ulus, 06050, Ankara, Turkey, [email protected]. 2 FernUniversität in Hagen, Faculty of Business Adminstration and Economics, 58084 Hagen, Germany, [email protected]. 3 The views and opinions presented in this study belong to the authors and do not necessarily represent those of the Central Bank of the Republic of Turkey. The authors would like to thank to Cihan Yalçın, Şeref Saygılı, Etkin Özen, Orhun Sevinç, Hülya Saygılı, Saygın Şahinöz and Yusuf Emre Akgündüz as well as the seminar participants at Central Bank of the Republic of Turkey and Eastern Mediterranean University.
2
I. Introduction
In much of the literature on foreign trade, export demand is determined as a function of the relative
prices and external demand. The former determinant could be captured by prices at the product
level (in domestic or foreign currency, depending on the specification) or by using the real effective
exchange rate (REER) at the aggregated (sectoral or country) level. An increase in REER implies lower
competitiveness and hence is expected to decrease exports. However, as documented in the
previous literature, this presumed theoretical negative relationship between exports and the real
exchange rate does not hold (or is not stable over time) for many countries. Instead, aggregate
exports are mostly driven by external demand conditions.
Among alternative rationales provided by the literature to account for the aforementioned inability
of real exchange rates in explaining exports, this article highlights the sunk costs of entry and exit. As
the argument goes, existence of sunk costs would imply threshold level(s) of exchange rate for the
firms to enter into (and exit from) the export market. In between these thresholds, there is a band of
inaction where the firm does not change its export status. For example, an exporting firm with high
sunk costs might bear with temporary losses as long as the variable costs are covered. This wait-and-
see behavior of individual firms would result in hysteresis in export markets at the aggregated level.
In economics, hysteresis would suggest permanent effects of temporary shocks and is usually
characterized by path-dependent multiple equilibria. Nevertheless, measuring hysteresis is not
straightforward since the adjustment of exports could differ in size and speed depending on the firm
characteristics. Firms might have different exchange rate thresholds beyond which their export
market activity would change. This article employs Preisach method to aggregate the impact of the
aforementioned wait-and-see behavior of individual firms on exports, taking cognizance of different
threshold levels.
The empirical exercise includes 17 subsectors of the Turkish manufacturing sector. In this sense and
to the best of our knowledge, this study is the first one that tests for the existence of export market
hysteresis in Turkey. Turkey is a developing country with increasing export orientation in the last
couple of decades and provides an interesting case study for the aforementioned export market
hysteresis phenomenon. On the one hand, higher integration of Turkish firms into the global value
chains increases their export survival rate and suggests a weaker role for exchange rates in their
entry-exit decisions to the export market. On the other hand, the enduring depreciation of the
domestic currency in real terms in the last years (Figure 1) renewed interest on the relationship
between the exchange rate and exports at the sub-sectoral level. In particular, this exercise puts into
test whether taking cognizance of the hysteresis behavior in the export market would help us signify
the relationship between exports and exchange rates in certain sectors.
INSERT FIGURE 1 ABOUT HERE
Our empirical exercise includes two-stages. In the first stage, we conduct the conventional export
equation with REER and a global growth variable and show that REER could not explain exports in any
subsectors. In the second stage, we replace the REER with the Preisach variable (PV), a variable that
filters the small changes in the exchange rate calculated á la Piscitelli et al. (2000). A significant PV
would imply the existence of hysteresis in the export market. The results of this second stage point
out a single subsector for which such a nonlinear filter improves the significance of the exchange rate
in export estimations: The manufacturing of wearing apparel, dressing and dyeing of fur (henceforth
3
referred to as clothing). 4 This result indicates that once the hysteresis behavior in the export market
is taken into account, the theoretical relationship between exports and the exchange rate holds in
this sector.
To shed more light on the results, the dynamics of the production and export behavior in the clothing
sector is analyzed in depth. Consequently, a number of factors are suggested for the hysteresis in the
Turkish clothing sector: the sub-contracting capacity of intermedieries determined with their
previous export experience and established connections, the low importance of plant size in entry
decision, easier financing conditions and price advantage due to a real exchange rate depreciation.
The plan of the study is as follows. Next section presents a review of the literature on the export
market hysteresis methods as well as a brief review of the historical evolution of Turkish export
dynamics followed by the corresponding literature on the Turkish exports. The third section
describes the Preisach method in detail. The fourth section describes the data and the fifth section
documents the results. The sixth section discusses the policy implications of the results and
concludes.
II. Literature Review on Export Market Hysteresis and the Dynamics of the Turkish
Exports
II.a. Literature on export market hysteresis
There are many occasions a firm may face sunk costs when entering an export market. For example
information about foreign demand, or health and security standards of destinations have to be
gathered (Bernhard and Wagner, 2001; Roberts and Tybout, 1997). Transporting, distribution and
selling have to be organized (Baldwin, 1990; Bernhard and Wagner, 2001). There may be costs for
advertising and establishing a brand name as well as for hiring and training additional workers
(Baldwin, 1990). Exiting a market can also involve sunk costs as e.g. severance payments. Dixit
(1989a), Dixit (1989b), and Baldwin and Krugman (1989) studied the impacts of sunk costs for market
entry and exit theoretically. They lay also the foundation for the hysteresis literature in international
trade.
Roberts and Tybout (1997) test if sunk cost hysteresis effects the exporting decisions of Colombian
firms and indeed find evidence that prior market experience influences the export decision. Other
factors which affect exporting behavior are macroeconomic conditions, observable plant costs and
demand variables as well as unobserved time-invariant plant heterogeneity. Lieberman et al. (2015)
show that sunk costs for market entry and exit decrease if the investment, such as a new build plant
or specially trained workers, can be used in a related existing business earned by the investor.
Timoshenko (2015) begins her framework with the same hysteresis model of non-ideal relay as we
do. She aims to identify the source of state dependence and therefore distinguishes between sunk
costs and learning-by-exporting which could both be reasons for hysteresis. Studying Colombian
plant-level data from 1979 to 1989 she finds that learning, i.e. being exporting in previous periods,
has a stronger effect on export persistence than sunk costs especially in differentiated-products
industries. She argues that exporters continue exporting because export experience depreciates
rapidly. Roberts and Tybout (1997) also state that a Colombian firm with two years absence of
exporting has to pay similar reentry costs than a new exporter.
Meinen (2015) controls for aspects like learning and finds that destination specific sunk costs matter.
However, they matter differently across sectors. In addition, import experience from a specific
4 Both “clothing” or “garment” terms are used in a mixed manner for this sector in previous studies.
4
market is able to facilitate exports to this destination. Studying the Danish furniture industry, Meinen
(2015) shows that a firm which already exports has a higher probability of a further export market
expansion than a non-exporting firm. However, the latter result depends on the characteristics and
the number of developed export markets. Gullstrand and Persson (2015) distinguish core and
peripheral export markets and give reasons for core markets having higher entry costs, e.g. for
marketing. Thus, they find theoretical and empirical evidence, testing Swedish food chain data from
1997 to 2007, that firms stay longer on core markets in line with hysteresis literature but are more
willing to exit peripheral markets as is analyzed by trade duration literature. Padmaja and Sasidharan
(2017) analyze Indian firm-level data of manufacturing firms and find also evidence that sunk costs
matter for the export participation decision. They are able to control for firm characteristics and
show that large firms, foreign owned firms and multiproduct firms face less sunk costs than small
firms, single product firms or firms owned by Indians. Beneath a dynamic discrete choice model they
also apply discrete-duration survival analysis and find persistence in exports. The longer a firm
exports, the lesser is the risk of exit from the export market.
Kemp and Wan (1974) lay the foundation of hysteresis in trade studies. They found out that adding
adjustment costs of hiring and firing induce multiple long-run equilibria in a closed economy, two
industries model with only one mobile factor, labor. Thus, they found hysteresis as it is known in
physics e.g. in magnetics. Hysteresis in trade, on the other hand, is usually caused by sunk costs. It is
referred to as export persistence, meaning that exporters stay in export markets despite unfavorable
conditions such as home currency appreciations. A further characteristic of a hysteresis system is
that large shocks which neutralize each other do not bring the system back to its initial point. This
characteristic property is called remanence (Piscitelli et al. 2000). The starting point for research in
the eighties (Baldwin and Krugman, 1989; Baldwin, 1990; Dixit 1989a; Dixit 1989b) was that an
appreciation of the US dollar from 1980 to 1985 resulted in market entries in the USA of foreign firms
which did not abandon the US market after the exchange rate shock was over. These foreign firms
stayed on the US market because they had already paid sunk market entry costs, e.g. invested in
distribution networks and marketing. On the other hand, US firms abandoned export markets during
this time span and did not re-enter these markets only because of the following dollar depreciation
(Baldwin and Krugman, 1989). Other competitors may have filled the gap and a re-entry may have
been costly. Thus, the exchange rate period of dollar appreciation changed the market structure of
US and former US export markets. A temporary shock had long-lasting effects. Thus, path-
dependence combined with non-linearity is another characteristic property of hysteresis.
After Roberts and Tybout (1997) showed that sunk costs influence the decision to export, Belke and
Göcke (1999) examine the effects of exchange rate changes on employment if costs for hiring and
firing matter. In Belke and Göcke (2001) they provide an estimation procedure with which they also
test for hysteresis in trade (Belke et al. (2013); Belke et al. (2014); Belke et al. (2015)). They derive a
hysteresis variable from the exchange rate and include it in an empirical estimation model. The
advantage of this approach is the sufficiency of aggregated export data which are more available
than firm-level data. Piscitelli et al. (2000) apply another approach which builds on the algorithm of
Preisach (1935) to derive a hysteresis variable. In Hallett and Piscitelli (2002) both methods are
compared and the latter one is favored. However, in Belke et al. (2013), Belke et al. (2014), and Belke
et al. (2015) an improved version of the algorithm is used and hysteresis is found for German and
other European Area member countries’ exports. Werner (2017) examines European wine exports to
the US applying the method of Belke and Göcke (2001) as well as the Preisach method published by
Piscitelli et al. (2000) and receives similar results with both approaches. De Prince and Kannelbey
Junior (2013) study hysteresis in prices and quantities of Brazilian imports and combine the Piscitelli
(2002) method with panel cointegration testing.
5
Using Spanish firm-level data, Campa (2004) and Manez et al. (2008) find hysteresis in Spanish
manufacturing exports due to sunk costs which affect small firms in particular.
Other researchers use time series methods to search for hysteresis. Kannebley (2008) applies e.g.
threshold cointegration analysis and identifies hysteresis in Brazilian exports. However, many time
series analysts define hysteresis as zero-root dynamics. In this case all past events influence the
current state of the output variable. In contrast to this we use the term hysteresis as it is defined in
physics which means there is a selective memory. Thus, the output depends only on the non-
dominated past extremum values of the input. This phenomenon will be described later in section IV.
Amable et al. (1994), Amable et al. (2004), O’Shaughnessy (2000) or Setterfield (2009) discuss the
differences of these approaches in more detail.
II.b. Historical review of Turkish exports dynamics and the corresponding literature on
export market entry-exit decisions
The Turkish industrialization strategy during early 1930s to 1980s could be characterized by import-
substitution policies with protection of certain sectors aiming to expand their industrial base. While a
handful of export promotion schemes were adopted especially after 1960s, the industrial and trade
policies as a whole did not create a favorable environment for enhancing exports and the country has
suffered serious balance of payments problems in the 1970s. Accordingly, 1980s marked a shift from
the earlier policies of reducing imports to the low level of exports towards a broader policy of import
liberalization along with export promotion. These policies were accompanied by a more liberalized
exchange rate regime during the decade whereas the capital account has been fully liberalized in
1989. This outward-oriented policy initially led to a sharp increase in exports of the manufacturing
sector at the first half of 1980s, partially owing to the excess capacity of the sector following the
import shortages and recessions of the previous decade (Şenses, 1989). However, these export
promotion policies were not accompanied with sufficient investments in the manufacturing sector
and hence could not generate a profound accumulation pattern for sustainable growth for the
following decades5 (Özcan et al., 2001). Accordingly, the productivity gains in the leading export
sectors of 1980s were relatively limited (Voyvoda and Yeldan, 2001). Then again, lower labor costs
due to wage suppression and real depreciation of the domestic currency led to a competitive export
market in most parts of 1980s and 1990s.
Turkey signed a customs union agreement with the European Union in 1996. Afterwards, the export
performance in the first decade of the 2000s were relatively strong compared to the previous
decades. Exports show a fivefold increase from 2002 to 2017 in US dollar terms while their share of
GDP oscillates between 20 and 25 percent during the same period (Figure 2). Moreover, the
composition of exports has shifted from the sectors that produce consumption goods towards the
sectors that produce intermediate goods (Figure 3).
INSERT FIGURE 2 AND FIGURE 3 ABOUT HERE
An assessment of the structural change in Turkish exports in the last two decades should also take
cognizance of the change in world trade patterns in this same period. According to OECD (2018a), 70
percent of the current global trade consists of production through global value chains (GVC’s) where
means of production are exchanged across countries during the different stages of production. The
trade in value added (TİVA) analysis which considers this fragmented production structure suggests
5 In their comparative study where they focus on Turkish and Korean export-oriented growth strategies, Onaran and Stockhammer (2006) show that investments are stimulated by export competitiveness in Korea while they are driven mostly by domestic demand in Turkey.
6
that the foreign content of Turkish exports fluctuates between 15 to 20 percent in the last decade,
which is lower than the OECD average of 25 percent (OECD, 2018b).
The increasing role of GVCs is particularly relevant for our original question for a couple of reasons.
First, higher participation in GVC increases the export survival rate for firms thanks to the lower
uncertainty, higher cooperation, strategic partnership among foreign firms as well as knowledge of
foreign markets (Diaz et al., 2018a)6. Accordingly, Türkcan (2016) shows higher export survival rates
for Turkish firms producing machinery equipment with higher engagement in GVCs. Second, as OECD
(2018a) points out, even small trade barriers (such as a low tariff rate) could have recurring patterns
along the value chains, hence could accumulate into substantial costs. This could be considered as a
factor increasing the entry costs to the export market. Third, this recurrence in production also
implies a lower exchange rate elasticity for exports [Ahmed et al. (2015), Soyres et al. (2018)]7.
Recent studies on Turkish exports suggest that the external demand is the main determinant of
exports while relative prices are mostly insignificant, in line with the literature on many other
countries, as discussed in the introduction. Uz (2010) and Saygılı and Saygılı (2011) studies both show
that exchange rate sensitivity of Turkish exports is very low. The latter study further argues that the
impact of external demand is not stable over time. In particular, foreign demand elasticity of exports
is higher for the 2000-2008 period in comparison to the 1987-2000 period. Bozok et al. (2015) use
disaggregation among export regions and show that while income is significant for all regions,
relative prices are only significant for selected regions. Similarly, Çulha and Kalafatçılar (2014) show
that exports to developed countries have a significant relationship with foreign income while exports
to emerging markets are more responsive to real exchange rate changes. Berument et al. (2014)
focus on the variation in income elasticities among sectors. They suggest that the income elasticity is
high in Motor Vehicles, Basic Metals and Radio-TV while it is either insignificant or low for food
products and the beverages sector8.
The literature on the export market participation decisions of Turkish manufacturing sector consists
of a number of sectoral as well as firm-level analyses. Özler et al. (2009) examines Turkish
manufacturing firms for the 1990-2001 period and show that sunk costs of entry are higher than that
of re-entry, indicating a positive but diminishing effect of the export history on export entry
decisions. Aldan and Günay (2008) results provide support for the self-selection hypothesis in the
sense that the presence of larger and more productive firms in export markets would be an outcome
of their higher capability to bear the sunk costs of entrance. Recently, Demirhan (2016a) further
underlines the existence of a learning effect, in addition to the self-selection hypothesis for Turkish
exporters. Corroborating with the findings of Özler et al. (2009), she also suggests the significance of
previous export experience in export propensity.
Demirhan (2016b) further delves into the entry and exit decisions of exporting firms in Turkish manufacturing sectors, using duration models. She shows that firms waiting time to be an exporter gets smaller with size, productivity, quality-orientation, ease of financing and capital intensity. Interestingly, profitability is suggested to lower the export incentive which is suggested as an implication of the risk-averse behavior of these firms. However, partially in contrary to these results,
6 Diaz et al. (2018b) also shows that manufacutring export flows with higher foreign services lead to longer export duration. 7 In line with this view, Eichengreen and Gupta (2013) argues that the exchange rate elasticity of services is higher than the manufacturing sector since the service sector employ fewer imported inputs. 8 One contrary evidence to the insignificant relationship between exchange rate and exports is recently provided by Toraganlı and Yalçın
(2016). They show that the firms with higher foreign exchange denominated debt to exports are more sensitive to the changes in the
exchange rate, pointing out the importance of liability dollarization and currency mismatch in financing decisions of, in particular, the small
and medium sized firms.
7
Gezici et al. (2018) argues that financing constraints of Turkish manufacturing firms do not present a significant obstacle for export market entry. Demirhan and Ercan (2018) analyze the impact of economic crises on export behavior of the Turkish
manufacturing firms. According to their results, export propensity increased in 1994 due to
devaluation and contracting demand. However, while similar conditions resulted in an increase in
export volume, accompanying credit crunch was a major obstacle for new entrants in 2001 crisis. The
2008 crisis, on the other hand, highlights a contraction both in export propensity and export volume
due to the collapse in global trade.
III. Data
The quarterly sectoral export volume indices which are classified in Broad Economic Classification are
taken from Turkish Statistical Institute for 2006Q1-2018Q2 period. The manufacturing sector has 17
subsectors as presented in Table 4.
The CPI based real effective exchange rate (REER) is measured as the weighted geometric average of
the domestic prices relative to the prices of the principal trade partners and taken from the Central
Bank of the Republic of Turkey (CBRT) database. An increase in REER suggests appreciation of the
domestic currency in real terms, indicating higher value of Turkish goods in terms of foreign goods.
Hence, the expected sign of the coefficient in the export specification is negative.
The foreign demand variable is the export-weighted global growth. This index is calculated by
multiplying the real growth of country i with the weights of this country in Turkish exports (wi) at
time t (Çıplak et al., 2011).
𝐺𝐺𝑒𝑥𝑝,𝑡 = ∑ 𝑤𝑖𝑦𝑡,𝑖
𝑛
𝑖=1
In the export estimation, higher global demand would indicate higher exports and hence the
expected sign of this coefficient would be positive.
The quarterly real GDP is calendar adjusted, measured as a chain linked volume index and provided
by Turkish Statistical Institute (TURKSTAT).
Most of our series suffer from the unit root problem and display seasonal patterns. Taking
cognizance of these problems, we employed year-on-year changes for all dependent and
independent variables in our estimations.
IV. Method
We apply the Preisach procedure (Preisach, 1935) provided by Piscitelli et al. (2000) to derive a
hysteresis variable, namely, Preisach variable, (PV). This variable is kind of a filtered exchange rate
which only reflects the large changes. More precisely, the non-dominated local minima and maxima
are described by this variable.
To derive the PV variable, we start with the non-ideal relay which displays the simplest hysteresis
model. We assume that the depending variable can take only two states: exporting (1) and not
exporting (0). The independent variable which causes these two states is the exchange rate.
Fluctuations of the exchange rate are expressed by back and forth movements on the horizontal axis.
Movements to the right are interpreted as depreciations of the home currency. Therefore if the
exchange rate comes from a low value and increases steadily, it will, at some point in time, reach the
8
export market entry trigger 𝛼 thus incentivize the firm to start exporting (Figure 4). We assume that a
firm which enters an export market has to pay sunk costs. If the exchange rate increases/depreciates
further, the firm will stay in the export market. However, despite the exchange rate appreciates
afterwards and falls below the entry trigger the firm will stay in the export market because it has
already paid the irrevocable market entry costs. However, if the exchange rate decreases more and
more, there will be a value at which the variable costs of exporting are not covered anymore and the
firm will pay the market exit costs and abandon the market. This value is the exit trigger 𝛽 at which
the firm switches from state 1 to state 0 (Figure 4). Thus, between the exit trigger and the entry
trigger there is a band of inaction. Knowing that the exchange rate is currently in this band does not
suffice to determine if the firm is exporting or not. It is important to know in which state the firm has
been in the previous period because if the firm has been in state 0, a movement in the band of
inaction which does not exceed the entry trigger, lets the firm stay in its non-exporting state.
Analogous considerations can be done if the exchange rate alters within the band of inaction and the
firm was in state 1 in the previous period. As long as the exchange rate is not less than the exit
trigger, the firm still exports. Mathematically we can express the non-ideal relay 𝐹𝛼,𝛽(𝑥(𝑡)) which
depends on the market entry trigger 𝛼 and the market exit trigger 𝛽 < 𝛼, as well as on the exchange
rate x(t) at time t as:
𝐹𝛼,𝛽(𝑥(𝑡)) = {1, if 𝑥(𝑡) ≥ 𝛼 − (𝛼 − 𝛽)𝐹𝛼,𝛽(𝑥(𝑡 − 1))
0, otherwise,
see Timoshenko (2015).
INSERT FIGURE 3 AROUND HERE
Therefore, the non-ideal relay is able to model the exporting hysteresis behavior of one firm in a
simple way. The next step is to aggregate many heterogeneous firms of which everyone has different
entry and exit trigger values.
𝑃𝑉(𝑥(𝑡)) = ∬ 𝜔(𝛼, 𝛽)𝐹𝛼,𝛽(𝑥(𝑡))𝑑𝛼𝑑𝛽𝛼≥𝛽
This aggregation procedure was invented by Preisach (1935). We assume that the different entry and
exit triggers of the firms are distributed uniformly among the Preisach triangle which is depicted in
Figure 4. This assumption is technically convenient as we assume that the weight function 𝜔(𝛼, 𝛽) ≡
1 for all 𝛼 and 𝛽, but does not alter the results meaningfully as was shown by Piscitelli et al. (2000).
We write the exit trigger values on the horizontal axis and the entry trigger values on the vertical
axis. Then, all firms lie in the Preisach triangle which is bordered by the entry = exit trigger line, the
vertical axis and the maximum of the exchange rate in the considered period because for all firms the
entry trigger exceeds the entry trigger i.e. 𝛽 < 𝛼. To illustrate the aggregation process, let us assume
we start at a low value of the exchange rate at which no firms export. An increase of the exchange
rate up to a local maximum M1 therefore will exceed entry triggers of some firms. These firms will
now export and they can be identified by the triangle which arises when we move upwards on the
vertical axis (Figure 5a).
Next, the exchange rate will decrease to a local minimum value m1. Firms whose exit triggers are
undercut will exit the export market. They can be found in Figure 4 by projecting the previous local
maximum M1 from the vertical axis by the entry = exit trigger line to the horizontal axis. Next, the
movement from this local maximum value to the local minimum value is retraced on horizontal axis.
9
The exiting firms are represented by the small triangle which is cut from the previous triangle of
active firms. i.e. the active firms are now depicted by a trapezoid (Figure 5b).
INSERT FIGURE 5 AROUND HERE
The next example shows how local maxima and minima are erased from the memory process. A
strong increase of the exchange rate, retraced by a vertical move on the vertical axis up to a higher
local maximum than the last one M2 > M1, erases all previous local maxima and minima from the
memory process. A large upwards movement means a huge shift to the right in all firm’s non-ideal
relays which means that all the firms which entry triggers are exceeded will now start to export or
remain in the export market, see Figure 5c. The next decrease of the exchange rate results in a
trapezoid of active firms as described above, see Figure 5d. The following increase which is assumed
to be not as strong as the second one up to M3, adds a further triangle to the trapezoid when moving
upwards on the vertical axis again. Following fluctuations of the exchange rate results in the end in a
staircase function which divides the Preisach triangle in two parts. In the upper part 𝑆− lie the firms
which are not active in the export market whereas in the lower part 𝑆+ all exporting firms are
pictured. As 𝐹𝛼,𝛽(𝑥(𝑡)) = 0 for all inactive firms in 𝑆−, it is sufficient to integrate over all active firms
in 𝑆+ where 𝐹𝛼,𝛽(𝑥(𝑡)) = 1, thus:
𝑃𝑉(𝑥(𝑡)) = ∬ 𝜔(𝛼, 𝛽)𝐹𝛼,𝛽(𝑥(𝑡))𝑑𝛼𝑑𝛽𝑆+
+ ∬ 𝜔(𝛼, 𝛽)𝐹𝛼,𝛽(𝑥(𝑡))𝑑𝛼𝑑𝛽𝑆−
= ∬ 𝜔(𝛼, 𝛽)𝑑𝛼𝑑𝛽𝑆+
≈ ∑ ∬ 𝜔(𝛼, 𝛽)𝑑𝛼𝑑𝛽𝑄𝑘(𝑡)
𝑛(𝑡)
𝑘=1= ∑ 𝑄𝑘(𝑡)
𝑛(𝑡)
𝑘=1
Every step of the staircase function is built by a trapezoid 𝑄𝑘(𝑡), thus the Preisach variable PV at time
t is the sum of all 𝑛(𝑡) trapezoids which represent the active firms at time t. Only non-dominated
local extremum values matter for the memory process which is selective, non-linear and with
remanence (Hallett and Piscitelli, 2002).
In our analysis, the set of equations are estimated for each subsector of the Turkish manufacturing
sector:
Xti, yoy = C + α1 REERt ,yoy+ α2 GGt,yoy + α3 GDPt-1,yoy + εti (1)
Xti,yoy = C + α1 PVt,yoy + α2 GGt ,yoy+ α3 GDPt-1,yoy + εti (2)
In the equations, X stands for the export volume, C is the constant term, REER is the real effective
exchange rate, GG is the export-weighted global growth, GDP is the gross domestic product, PV is the
Preisach variable and ε is the error term where subscript t and i denote the time and sector
components. The impacts of crises are captured by two dummies in our estimations. One of them is
the 2008 global crises (denoted by FC) which reduced both the export volume and new entrance in
export markets (Demirhan and Ercan, 2008) as mentioned in the literature section. We also use a
dummy for 2016Q4 to capture the impact of the Russia-Turkey jet crisis: When Turkey shot down a
Russian jet, Russia decided to ban some Turkish export products until the issue has been solved
10
through diplomacy9. This dummy is shown as DP16_4 in the estimations. For each subsector, the first
equation is the benchmark equation with REER. In the following equation, we replace REER with the
aforementioned PV variable.
The motivation for using the explanatory variables, REER and GG, are discussed in previous sections
in line with the previous literature. The literature also suggests including variables capturing
domestic growth measuring the impact of two counteracting forces in export supply estimations
(Goldstein and Khan, 1985). On the one hand, an increase in trend income could result in an increase
in total factor productivity or would indicate better infrastructure. Moreover, assuming that there is
excess capacity for exporting firms, higher income would lead to more abundant factor supplies. All
these factors would result in higher supply of exports. On the other hand, if domestic demand is the
leading factor for higher income, then exporting firms might prefer to direct their sales towards the
domestic market to reap potential profits, resulting in lower exports. Hence, the coefficient of GDP
could be positive or negative depending on which of these counteracting factors would dominate.
There are two nuisances that should be handled in PV estimations as stated by previous literature.
First, as documented above, all of our dependent and independent variables suffer from non-
stationarity indicating time-dependent means or variances. However, usual remedy of taking
differences is problematic in PV analysis since the procedure deals with path-dependent effects
determined by the levels of the forcing variable (exchange rate) (Belke et al, 2013). One solution to
this problem is using fully modified least squares (FM-OLS) proposed by Phillips and Hansen (1990),
and implemented by Mota et al. (2012). However, as Belke et al. (2013) reports the problem still
remains for identification of play width in this framework. Instead, in our analysis we employ a two-
stage process, different than the previous literature. In the first stage, the PV variable is derived from
the level of the exchange rate. In the second stage, we take year-on-year (y-o-y) difference of this
variable and use it in our estimation. The y-o-y difference would also help us to account for the
seasonality problem in the export series.
The second nuisance is on the correlation between the PV variable and the forcing variable. As
described above, the PV variable is a filtered version of the exchange rate and could reveal high
correlation if the band of inaction is small. Taking cognizance of the impact of correlation between
independent variables on the results, unlike the previous literature, we do not use REER and PV
variables in the same equations.
V. Results
The estimation results for the manufacturing sector and its subsectors are provided in Tables 1-3. For
each subsector, the first column includes the estimations with REER and the second column the
estimations with PV.
The estimation results for the manufacturing sector as a whole (Columns 1 and 2 of Tables 1-3)
suggest that the global growth variable is significant while the domestic demand indicator is
insignificant. On the other hand, the real effective exchange rate coefficient has an unexpected
positive sign. These results are in line with the previous literature stating that the main determinant
of the exports in the manufacturing sector is the global growth and the exchange rate is mostly
insignificant (Uz, 2014; Saygılı and Saygılı, 2011; Bozok et al., 2015; Çulha and Kalafatçılar, 2014;
Berument et al., 2014.).
9 See news in the link https://www.bbc.com/news/world-europe-35209987.
11
The rest of Tables 1-3 present the sectoral results. The REER is either positive or not significant in all
subsectors, as shown in the first columns of each subsector. At this point we ask whether this result
changes once we consider the band of inaction due to sunk costs. To answer this question, we
replace the REER with the PV variable. The results are documented in the second column of each
subsector. The results indicate that for only two sectors, this new filtered exchange rate (PV variable)
is significant: manufacturing of wearing apparel, dressing and dyeing of fur; and manufacture of
radio, television and communication equipment and apparatus (abbreviated as clothing and radio,
respectively for the rest of the text). In our analysis, we only focus on the clothing sector,
disregarding the relatively unreliable results for radio. For the radio sector the insignificance of the
global growth variable and the significance of the crisis dummy suggest that the estimation would
take into account the structural breaks in the period. However, without a longer data set, the result
of such an analysis would not be reliable.
First of all, the absence of the significance of PV in most sectors could be the result of very large sunk
costs (hence very large band of inaction). As the argument goes, the main determinant of export
entry behavior for many developing countries is the “trust” of the foreign correspondent to the
domestic firm for sustainable production. If the foreign buyer could not tolarate an interruption in
any stage of her production, she would be very selective on including a domestic firm into the
production chain. This kind of confidence could only be provided by large and experienced firms.
These firms, on the other hand, are usually less credit-constrained, in the sense that they can raise
funds in foreign currency and could hedge themselves against changes in the exchange rate. Hence,
the production processes for many of these large and capital-intensive firms are less dependent on
exchange rate changes. This argument is in line with the self-selection hypothesis in export entry
decision as described in the second section (Aldan and Günay, 2008; Demirhan 2016a).
The significance of the PV variable for the clothing, on the other hand, could be motivated with an in-
depth analysis of the production layers specific to this sector. To this aim, we first provide a general
description of the sector and depict the historical developments over time. Later on, we discuss the
presence and the scope of sunk costs in this particular sector in four premises.
As of 2017, clothing constitutes the sixth largest subsector of the total manufacturing sector,
producing 6 percent of the total manufacturing sector value added. On the other hand, the share of
employment in clothing sector in total manufacuring sector is 18 percent while the average wages in
clothing is 28 percent lower than the average wage level in Turkey. 10 These figures indicate that
clothing is a relatively labor-intensive sector with low productivity.11
Initially, the aforementioned comparative advantage in clothing (and textiles as well) makes it one of the locomotive sectors for the export boom that started in the 1980s. However, the 2000s revealed a global shift of production towards China and neighbouring Asian developing economies in these sectors due to relatively lower production costs and preferential trade agreements with major importer economies (MD, 2014). The share of Turkey in total world clothing exports was around 3.5 percent between 2006-201412. However, as Figure 6 depicts, the share of exports of clothing in the
10 Employment and wage figures are taken from 2016 Labor Force Survey (LFS). Value added is the 2017 figure. The source for both datasets are TURKSTAT. 11 Unfortunately, all around the world, this sector is one of the most problematic ones in terms of the working conditions. OECD (2018c)
provides a due dilligance report specifically designed for the enterprises and subcontractors in this sector to meet their responsibilities
against their workers and the society. 12 The data source is ITCTradeMap (https://www.trademap.org).
12
exports of the total manufacturing sector in Turkey displayed a steady decline in the last two decades, going down from 24 percent in 1996 to 7 percent in 2018. In contrast to the lowering share of clothing in total manufacturing exports, Figure 7 shows that the
number of firms in the clothing sector has a steady increase with the exception of two years after the
GFC. This surge is consistent with the result of Özler et al. (2009) indicating that sunk costs are lower
in textile and clothing industries relative to other sectors. This provides support for our hypothesis in
the following manner: We argue that low costs of this sector let the exchange rate to be determinant
of entry and exit in this sector. If the exchange rate depreciates above a certain level, the firm might
find it profitable to enter in the market in textile and clothing sector. However, the sunk costs are
extremely high in other sectors that even a big change in the exchange rate would not matter much
for the entry-exit behavior. Below, we discuss the entry-exit behavior of the exporters in the clothing
sector in relation with the previous literature.
INSERT FIGURE 6 AND FIGURE 7 ABOUT HERE
First, the textile and clothing sector has a multi-layered production structure in Turkey. Many major foreign brands have strong connections with some middle / large sized Turkish firms. These firms with large-scale production units also act as intermediaries which might, at times, extend the production process to some subcontractors in their region. Once the foreign demand increases, these large firms can either increase production via intensive margin or pass some of the excess demand to these subcontractors. If the foreign demand is high enough, these intermediaries, with their expertise and network connections, could initiate the establishment of new small enterprises with employment below 20 workers. Furthermore, most of the employment of the new subcontractors consists of previous workers / employers in the sector. Hence, as Özler et al. (2009) and Demirhan (2016a) suggest, some of these new establishments could be a re-entry in the sector indicating that previous export experience is important in export propensity. As discussed before, textile and clothing were the main sectors of the export boom starting at 1980s. Furhtermore, as Şenses (1989) argues, the government was active in these years in bilateral trade agreements to increase exports of the manufacturing sector which helped these firms to establish relations with foreign firms. Hence, most of these firms have very long experience and network connections (foreign as well as domestic) in this sector, supporting our hypothesis that these intermediaries could help to initiate new companies against a rise in foreign demand. Second, compared to other sectors, according to Özler et al. (2009), the importance of plant size is relatively lower in export propensity in the textile and clothing industry in comparison to others. This increases the probability of establishing a new small firm to benefit from exporting. Third point that would help us to motivate lower sunk costs in the last decade would be related to financing conditions. In fact, Özler et al. (2009) show that the role of imported machinery and equipment is relatively important in the textile and clothing sector for capital stock, in comparison to other sectors. This dependency on foreign inputs was less of a significant obstacle for Turkish firms in the last decade for two reasons. The first one is the presence of a relatively low interest rate environment in the period which might have led to easier financing conditions for new firms. The second one is the availability of leasing opportunities which constitutes around 6-7 percent of the total machinery and equipment in textile sector. 13 The fourth point is directly related to our analysis of exchange rates. The significance of the PV variable suggests that the exporting behavior in the intensive and extensive margin depends on the
13 Detailed data on sectoral leasing is available at Association of Financial Institutions (www.fkb.org.tr).
13
exchange rate, once we consider the sunk costs. The increase in foreign demand might be a result of the price advantage due to a depreciation of Turkish lira in real terms over the last decade (Figure 1). Note that, in order to benefit from this price advantage, the foreign firms should have alternative producers / intermediaries in different countries. As discussed before, after the increase in the share of Asian countries in textile and clothing production around the world in the first decade of the century, many important brands have suppliers in Asian countries in addition to the previous exporters such as Turkey. These brands observe the exchange rate developments all over the world and easily direct their production from one country to the other by their already established intermediary contractors in these countries and their market power in this low cost, labor intensive industries.
VI. Conclusion, Discussion and Policy Recommendations
To sum up, this article argues that taking cognizance of sunk costs would help us to observe the
impact of exchange rate changes in the export behavior of Turkish firms in the clothing industry. The
potential determinant of this relationship is suggested as the sub-contracting capacity of experienced
intermediaries with strong connections; insignificance of plant size in the establishments of new
firms, better financing conditions in the period and a price advantage observed by a depreciation of
the domestic currency.
One problem for developing countries is that the deferred consumption might result in higher
domestic demand once the economy is above its long-term trend. In an economy mainly led by
domestic demand, many firms can make profit by producing for the domestic market and not have
strong incentives to export. However, moving to export markets in good times would be easier since
credit conditions are usually more relaxed in the higher phase of the economic cycles. On the other
hand, when the economy turns into a negative state, usually domestic demand goes down and the
firms might have higher benefits from exporting. However, the financing conditions also deteriorate
in these periods so it is harder to cover the costs of entry into export markets. This suggests that the
exporting behavior should follow a countercyclical pattern: Invest in exporting during “good times”
when the marginal benefit from exporting is lower compared to the “bad times”, but when the
financing conditions allow you to cover the cost of entry.
The aforementioned countercyclical behavior could also be supported by government incentives.
However, this would require a detailed analysis in many different respects. Our analysis tells that the
level of the exchange rate could determine the affordability of sunk costs. However, one caution
could be on the optimal plant size for the future productivity of this sector. A recent report published
by Turkish Clothing Manufacturing Association (TGSD, 2016) documents that while the need for
flexibility and speed justifies the need for small firms, the value-added of middle / large size plants is
higher in the sector. They suggest that the investments would also be channeled towards increasing
the average firm-size in the sector. Obviously, increasing automation in the last decade would have
immediate impacts on employment in this sector, once such a path would be followed. Hence, the
optimal policy would require a comprehensive analysis including the impact of incentives on both the
production and employment in the sector.
14
References
Ahmed, S., Appendino, M., & Ruta, M. (2015). Depreciations without exports? Global value chains and the exchange rate elasticity of exports. The World Bank. Aldan, A., & Günay, M. (2008). Entry to Export Markets and Productivity: Analysis of Matched Firms in Turkey. Central Bank of the Republic of Turkey Working Papers, 08/05. Amable, B., Henry, J., Lordon, F., & Topol, R. (1994). Strong hysteresis versus zero-root dynamics. Economics Letters, 44(1-2), 43-47. Amable, B., Henry, J., Lordon, F., & Topol, R. (2004). Complex remanence vs. simple persistence: are hysteresis and unit-root processes observationally equivalent?. In Economic Complexity (pp. 67-89). Emerald Group Publishing Limited. Baldwin, R. (1990). Hysteresis in trade. In Hysteresis Effects in Economic Models (pp. 19-34). Physica-Verlag HD. Baldwin, R., & Krugman, P. (1989). Persistent trade effects of large exchange rate shocks. The Quarterly Journal of Economics, 104(4), 635-654. Belke, A., Göcke, M., & Günther, M. (2013). Exchange rate bands of inaction and play‐hysteresis in German exports—Sectoral evidence for some OECD destinations. Metroeconomica, 64(1), 152-179. Belke, A. H., Göcke, M., & Werner, L. (2014). Hysteresis Effects in Economics–Different Methods for Describing Economic Path-dependence. Belke, A., Göcke, M., & Werner, L. (2015). Exchange Rate Volatility and other Determinants of Hysteresis in Exports-Empirical Evidence for the Euro Area. Review of Economic Analysis, 7(1), 24-53. Bernard, A. B., & Wagner, J. (2001). Export entry and exit by German firms. Weltwirtschaftliches Archiv, 137(1), 105-123. Berument, M. H., Dincer, N. N., & Mustafaoglu, Z. (2014). External income shocks and Turkish
exports: A sectoral analysis. Economic Modelling, 37, 476-484.
Bozok, İ., Dogan, B. S., & Yunculer, C. (2015). Estimating Income and Price Elasticity of Turkish Exports
with Heterogeneous Panel Time-Series Methods (No. 1526).
Campa, J. M. (2004). Exchange rates and trade: How important is hysteresis in trade?. European
Economic Review, 48(3), 527-548.
Çıplak, U., Eren, O., Öğünç, F., & Saygılı, M. (2011). Turkiye’nin Ihracati Icin Alternatif Dis Talep Gostergeleri (No. 1107). CBRT Economic Notes, Central Bank of the Republic of Turkey.
Çulha, O. Y., & Kalafatçılar, M. K. (2014). A Glance at Income and Price Elasticity of Turkey‟ s Exports:
the Importance of Regional Disparities. CBT Research Notes in Economics, 5, 1-14.
Demirhan, A. A. (2016a). Export Behavior of the Turkish Manufacturing Firms. Emerging Markets Finance and Trade, 52(11), 2646-2668.
15
Demirhan, A. A. (2016b). To be exporter or not to be exporter? Entry–exit dynamics of Turkish manufacturing firms. Empirical Economics, 51(1), 181-200. De Prince, D., & Kannebley Junior, S. (2013). Strong hysteresis in Brazilian imports: a panel cointegration approach. Journal of Economic Studies, 40(4), 528-548. Díaz-Mora, C., Gandoy, R., & Gonzalez-Diaz, B. (2018a). Strengthening the stability of exports through GVC participation: The who and how matters. Journal of Economic Studies, 45(3), 610-637. Díaz-Mora, C., Gandoy, R., & González-Díaz, B. (2018b). Looking into global value chains: influence of foreign services on export performance. Review of World Economics, 154(4), 785-814. Dixit, A. (1989a). Hysteresis, import penetration, and exchange rate pass-through. The Quarterly Journal of Economics, 104(2), 205-228. Dixit, A. (1989b). Entry and exit decisions under uncertainty. Journal of political Economy, 97(3), 620-638. Eichengreen, B., & Gupta, P. (2013). The real exchange rate and export growth: are services different?. The World Bank. Gezici, A., Orhangazi, Ö., & Yalçın, C. (2018). Exports and Financing Constraints Evidence from Turkey. Review of Middle East Economics and Finance, 14(1). Goldstein, M., & Khan, M. S. (1985). Income and price effects in foreign trade. Handbook of
international economics, 2, 1041-1105.
Gullstrand, J., & Persson, M. (2015). How to combine high sunk costs of exporting and low export
survival. Review of World Economics, 151(1), 23-51.
Hallett, A. H., & Piscitelli, L. (2002). Testing for hysteresis against nonlinear alternatives. Journal of Economic Dynamics and Control, 27(2), 303-327. Kannebley Jr, S. (2008). Tests for the hysteresis hypothesis in Brazilian industrialized exports: a threshold cointegration analysis. Economic Modelling, 25(2), 171-190. Kemp, M. C., & Wan, H. Y. (1974). Hysteresis of long-run equilibrium from realistic adjustment costs. In Trade, Stability, and Macroeconomics (pp. 221-242). Lieberman, M. B., Lee, G. K., & Folta, T. B. (2017). Entry, exit, and the potential for resource redeployment. Strategic management journal, 38(3), 526-544. Máñez, J. A., Rochina-Barrachina, M. E., & Sanchis, J. A. (2008). Sunk costs hysteresis in Spanish manufacturing exports. Review of World Economics, 144(2), 272-294.
MD (Ministry of Development). (2014). Report on Textile-Fur and Wearing Apparel Prepared for the
Tenth Development Plan of Turkey, Ankara, ISBN 978-605-9041-08-9
Meinen, P. (2015). Sunk costs of exporting and the role of experience in international trade. Canadian Journal of Economics/Revue canadienne d'économique, 48(1), 335-367.
16
Metin-Ozcan, K., Voyvoda, E., & Yeldan, A. E. (2001). Dynamics of macroeconomic adjustment in a globalized developing economy: growth, accumulation and distribution, Turkey 1969–1999. Canadian Journal of Development Studies/Revue canadienne d'études du développement, 22(1), 219-253. OECD (2018a), Trade Policy Implications of Global Value Chains, OECD Publishing, Paris (available at http://www.oecd.org/sti/ind/measuring-trade-in-value-added.htm) ______(2018b), Trade in Value Added: Turkey, OECD Publishing, Paris (available at http://www.oecd.org/sti/ind/measuring-trade-in-value-added.htm) OECD (2018c), OECD Due Diligence Guidance for Responsible Supply Chains in the Garment and Footwear Sector, OECD Publishing, Paris. http://dx.doi.org/10.1787/9789264290587-en Onaran, Ö., & Stockhammer, E. (2006). Do profits affect investment and employment? An empirical test based on the Bhaduri-Marglin model. In Wages, Employment, Distribution and Growth (pp. 206-222). Palgrave Macmillan, London. O’Shaughnessy, T. (2000). Hysteresis in an open economy model. Scottish Journal of Political Economy, 47(2), 156-182. Özler, Ş., Taymaz, E., & Yilmaz, K. (2009). History matters for the export decision: Plant-level evidence from Turkish manufacturing industry. World Development, 37(2), 479-488. Padmaja, M., & Sasidharan, S. (2017). Sunk Costs, Firm Heterogeneity, Export Market Entry and Exit: Evidence from India. Journal of Quantitative Economics, 15(2), 367-393. Piscitelli, L., Cross, R., Grinfeld, M., & Lamba, H. (2000). A test for strong hysteresis. Computational Economics, 15(1-2), 59-78. Preisach, F. (1935). Über die magnetische Nachwirkung. Zeitschrift für Physik, 94(5-6), 277-302.
Roberts, M. J., & Tybout, J. R. (1997). The decision to export in Colombia: an empirical model of entry
with sunk costs. The American Economic Review, 545-564.
Saygılı, H., & Saygılı, M. (2011). Structural changes in exports of an emerging economy: Case of
Turkey. Structural Change and Economic Dynamics, 22(4), 342-360.
Setterfield, M. (2009). Path dependency, hysteresis and macrodynamics. In Path dependency and macroeconomics (pp. 37-79). Palgrave Macmillan, London. Soyres, D., Raphael, F. M. M., Frohm, E., Gunnella, V., & Pavlova, E. (2018). Bought, Sold, and Bought Again: The Impact of Complex Value Chains on Export Elasticities. World Bank Policy Research Working Paper, (8535). Şenses, F. (1989). The Nature and Main Characteristics of Recent Turkish Growth in Export of Manufactures. The Developing Economies, 27(1), 19-33.
TGSD (Turkish Clothing Manufacturing Association), 2016. Horizon 2030: A Toadmap for Turkish
Clothing Sector, In Turkish), İstanbul
17
Timoshenko, O. A. (2015). Learning versus sunk costs explanations of export persistence. European Economic Review, 79, 113-128.
Toraganlı, N., & Yalçın, C. (2016). Exports, Real Exchange Rates and External Exposures: Empirical Evidence from Turkish Manufacturing Firms, Central Bank of the Republic of Turkey Working Papers, 16/24. Türkcan, K. (2016). On the role of vertical differentiation in enhancing survival of export flows: Evidence from a developing country. Uz, I. (2010). Bilateral trade elasticities of Turkey. International Journal of Applied Economics, 7(1), 28-46. Voyvoda, E., & Yeldan, A. E. (2001). Patterns of productivity growth and the wage cycle in Turkish manufacturing. International Review of Applied Economics, 15(4), 375-396. Werner, L. M. (2017). Identifying hysteresis non-linearities in trade: Evidence from European wine exports to the U.S.. AMERICAN ASSOCIATION OF WINE ECONOMISTS Working paper No. 215, 1-37. Werner, L. M. (2018). Hysteresis losses in the Preisach framework. Empirical Economics, 1-30. (https://doi.org/10.1007/s00181-018-1547-z)
18
Figure 1: Real Effective Exchange Rate (8-quarters moving average)
Figure 2: Total Exports
Source: Turkstat
Figure 3: Percentage Share of Sectors in Total Exports
Source: Turkstat
60
70
80
90
100
110
120
130
06
-4
07
-3
08
-2
09
-1
09
-4
10
-3
11
-2
12
-1
12
-4
13
-3
14
-2
15
-1
15
-4
16
-3
17
-2
18
-1
19
Figure 4: Non-ideal relay hysteresis model
not exporting 0
exporting 1
state of exporting
exchange rate
export
market entry
trigger
export
market exit
trigger
band of inaction
𝛼 𝛽
20
Figure 5: Preisach triangle and aggregation procedure
(a) (b)
(c)
(e) (f)
m2
M3
m1 M1
M1
entry = exit trigger
entry trigger
exit trigger
M1
entry = exit trigger
entry trigger
exit trigger
M1
entry = exit trigger
entry trigger
exit trigger
M2
M1
entry = exit trigger
entry trigger
exit trigger M1 M2 m2 m1
M1
entry = exit trigger
entry trigger
exit trigger
M1
entry = exit trigger
entry trigger
exit trigger
(d)
𝑆+
𝑆+ 𝑆+
𝑆+ 𝑆+
𝑆+
𝑆−
21
Figure 6: Share of Exports of Clothing in Exports of Total Manufacturing Sector
Source: Turksat
Figure 7: Number of Firms in Clothing Industry
Source: Ministry of Industry, Enterprise Information System
12000
13000
14000
15000
16000
17000
18000
19000
20000
2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016
0
5
10
15
20
25
19
96
19
97
19
98
19
99
20
00
20
01
20
02
20
03
20
04
20
05
20
06
20
07
20
08
20
09
20
10
20
11
20
12
20
13
20
14
20
15
20
16
20
17
20
18
22
Table 1: Outcomes of hysteresis estimations of manufacturing and five subsectors
Dependent variable:
manufacturing manuFoodBev manuTobacco manuTextiles manuDressing manuWood
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
REER 0.206*
-0.164
-0.057
0.211**
0.232*
-0.229
(0.110)
(0.127)
(0.380)
(0.097)
(0.136)
(0.270)
PV
0.009
0.001
-0.017
0.003
-0.023***
-0.017
(0.006)
(0.007)
(0.021)
(0.006)
(0.007)
(0.015)
GG 1.449** 1.413** -0.663 -1.023 0.923 1.669 0.516 0.787 0.208 1.868*** 2.046 2.460*
(0.543) (0.604) (0.624) (0.695) (1.870) (2.031) (0.478) (0.548) (0.667) (0.681) (1.326) (1.442)
GDP 0.164 0.140 0.611* 0.777** 0.819 0.548 0.441 0.300 0.447 -0.222 -0.319 -0.433
(0.305) (0.317) (0.350) (0.365) (1.048) (1.066) (0.268) (0.287) (0.374) (0.357) (0.743) (0.757)
FC 1.758 1.985 4.850** 4.899** 9.346 8.847 -0.029 0.022 -0.508 -1.258 4.376 3.902
(1.637) (1.676) (1.881) (1.929) (5.632) (5.636) (1.440) (1.520) (2.009) (1.889) (3.994) (4.001)
DP16_4 0.226 0.158 5.364 6.790* -26.010** -28.586*** 1.492 0.337 1.942 -4.135 1.372 0.132
(2.988) (3.129) (3.434) (3.601) (10.281) (10.521) (2.630) (2.838) (3.668) (3.526) (7.292) (7.469)
Observations 50 50 50 50 50 50 50 50 50 50 50 50
R2 0.617 0.603 0.600 0.586 0.329 0.337 0.523 0.476 0.263 0.358 0.250 0.258
Adjusted R2 0.574 0.559 0.556 0.540 0.254 0.263 0.469 0.417 0.182 0.287 0.167 0.176
Residual Std. Error (df = 45) 6.664 6.778 7.660 7.800 22.932 22.789 5.865 6.146 8.180 7.638 16.264 16.177
F Statistic (df = 5; 45) 14.468*** 13.680*** 13.526*** 12.725*** 4.408*** 4.577*** 9.850*** 8.165*** 3.218** 5.016*** 3.006** 3.135**
Note: *p<0.1; **p<0.05; ***p<0.01. Numbers in parentheses are standard errors.
23
Table 2: Outcomes of hysteresis estimations of six further subsectors
Dependent variable:
manuPaper manuCoke manuChemicals manuRubber manuOther manuMetals
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
REER 0.296*
-0.060
0.103
0.066
0.201*
0.293
(0.160)
(0.380)
(0.140)
(0.119)
(0.107)
(0.708)
PV
0.016*
0.026
-0.002
-0.001
0.014**
0.037
(0.009)
(0.021)
(0.008)
(0.007)
(0.006)
(0.040)
GG 0.561 0.306 5.163*** 3.677* 1.281* 1.571** 1.608*** 1.798*** 0.356 0.050 1.161 -0.197
(0.788) (0.863) (1.870) (2.012) (0.689) (0.758) (0.584) (0.640) (0.526) (0.566) (3.485) (3.780)
GDP 0.500 0.542 -0.945 -0.373 0.355 0.226 0.299 0.215 0.072 0.150 -0.922 -0.463
(0.442) (0.453) (1.048) (1.056) (0.386) (0.398) (0.327) (0.336) (0.295) (0.297) (1.954) (1.984)
FC 7.151*** 7.598*** -3.622 -2.803 1.079 1.010 1.247 1.200 1.023 1.406 8.832 9.928
(2.373) (2.394) (5.635) (5.583) (2.076) (2.102) (1.760) (1.777) (1.586) (1.570) (10.498) (10.490)
DP16_4 3.037 3.655 -24.466** -19.166* 5.809 4.688 0.533 -0.201 5.083* 5.972** -12.980 -8.471
(4.332) (4.470) (10.286) (10.422) (3.791) (3.925) (3.212) (3.318) (2.895) (2.931) (19.165) (19.583)
Observations 50 50 50 50 50 50 50 50 50 50 50 50
R2 0.612 0.610 0.306 0.328 0.579 0.574 0.655 0.653 0.279 0.304 0.029 0.044
Adjusted R2 0.569 0.567 0.229 0.253 0.532 0.527 0.616 0.614 0.199 0.226 -0.078 -0.062
Residual Std. Error (df = 45) 9.662 9.681 22.942 22.574 8.455 8.502 7.164 7.186 6.457 6.348 42.745 42.417
F Statistic (df = 5; 45) 14.198*** 14.105*** 3.967*** 4.393*** 12.363*** 12.128*** 17.070*** 16.910*** 3.491*** 3.925*** 0.273 0.416
Note: *p<0.1; **p<0.05; ***p<0.01. Numbers in parentheses are standard errors .
24
Table 3: Outcomes of hysteresis estimations of six further subsectors
Dependent variable:
manuFabricMetal manuMachinery manuElectric manuRadio manuVeh manuFurniture
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12)
REER 0.183
0.057
0.019
-0.301
0.359
0.145
(0.133)
(0.108)
(0.143)
(0.526)
(0.223)
(0.459)
PV
0.012
-0.003
0.004
-0.073**
0.009
0.008
(0.008)
(0.006)
(0.008)
(0.028)
(0.013)
(0.026)
GG 0.458 0.227 1.074** 1.324** 2.006*** 1.821** -1.118 2.063 4.367*** 4.606*** -1.746 -1.850
(0.655) (0.712) (0.533) (0.584) (0.701) (0.764) (2.588) (2.649) (1.099) (1.228) (2.257) (2.468)
GDP 0.934** 0.987** 0.855*** 0.750** 0.203 0.270 1.935 0.789 0.239 0.083 1.071 1.084
(0.367) (0.374) (0.299) (0.306) (0.393) (0.401) (1.451) (1.390) (0.616) (0.644) (1.265) (1.295)
FC -1.218 -0.899 -1.228 -1.318 -1.600 -1.472 -14.191* -16.374** -4.951 -4.733 3.966 4.172
(1.973) (1.977) (1.607) (1.620) (2.112) (2.121) (7.797) (7.351) (3.310) (3.408) (6.801) (6.849)
DP16_4 4.585 5.227 2.396 1.460 -5.463 -4.829 0.777 -10.159 9.274 8.089 6.340 6.567
(3.601) (3.691) (2.933) (3.024) (3.856) (3.959) (14.234) (13.724) (6.043) (6.362) (12.415) (12.786)
Observations 50 50 50 50 50 50 50 50 50 50 50 50
R2 0.602 0.606 0.750 0.750 0.512 0.515 0.094 0.206 0.690 0.676 0.086 0.086
Adjusted R2 0.558 0.562 0.722 0.722 0.458 0.461 -0.007 0.118 0.656 0.640 -0.016 -0.016
Residual Std. Error (df = 45) 8.032 7.994 6.543 6.550 8.601 8.576 31.748 29.725 13.479 13.781 27.690 27.694
F Statistic (df = 5; 45) 13.608*** 13.826*** 27.011*** 26.929*** 9.457*** 9.564*** 0.935 2.332* 20.068*** 18.808*** 0.847 0.843
Note: *p<0.1; **p<0.05; ***p<0.01. Numbers in parentheses are standard errors.
25
Table 4: Description of abbreviations
Abbreviation (Sub-)sector
manufacturing Manufacturing manuFoodBev Manufacture of food products and beverages manuTobacco Manufacture of tobacco products manuTextiles Manufacture of textiles manuDressing Manufacture of wearing apparel; dressing and dyeing of fur manuWood Manufacture of wood and of products of wood and cork, except furniture;
manufacture of articles of straw and plaiting materials. manuPaper Manufacture of paper and paper products manuCoke Manufacture of coke, refined petroleum products and nuclear fuel manuChemicals Manufacture of chemicals and chemical products manuRubber Manufacture of rubber and plastics products manuOther Manufacture of other non-metallic mineral products manuMetals Manufacture of basic metals manuFabricMetal Manufacture of fabricated metal products, except machinery and equipment manuMachinery Manufacture of machinery and equipment n.e.c. manuElectric Manufacture of electrical machinery and apparatus n.e.c. manuRadio Manufacture of radio, television and communication equipment and apparatus manuVeh Manufacture of motor vehicles, trailers and semi-trailers manuFurniture Manufacture of furniture; manufacturing n.e.c.