TY - JOUR
AU - , Di Maio, Michele
AB - Abstract This article investigates whether conflict induces distortions in the functioning and accessibility of markets for production inputs and in their allocation among firms. We study firm operations and outcomes in the context of Palestine during the Second Intifada. We analyse input usage over time across districts experiencing differential changes in conflict intensity. Conflict induces firms to substitute domestically produced materials for imported ones. Counterfactual analyses show that this mechanism can account for more than 70% of the fall in the output value of firms in high conflict districts. Well‐functioning markets are conducive to an efficient allocation of resources. In the case of production inputs, the presence of market imperfections may prevent resources from accruing to those sectors and firms which value them the most at the margin. This negatively affects aggregate total factor productivity (Hsieh and Klenow, 2009). There is still limited research examining the specific sources of market distortions and their relative importance, especially in developing countries. Filling this gap is essential in order to design policies that might remove these constraints to the functioning of markets and increase aggregate productivity and income. Violent conflict is one possible source of market distortions. For instance, the increase in violence and insecurity can lead to an increase in workers’ absence, limiting firms’ access to labour (Ksoll et al., 2015). At the same time, the increase in uncertainty related to conflict is likely to discourage the establishment or continuation of lending relationships and limit firms’ access to capital. Finally, conflict can affect the sustainability and scope of international trading relationships, thus constraining firms’ access to foreign inputs. These distortions are likely among the microeconomic mechanisms behind the negative relationship between conflict and aggregate economic activity (Alesina et al., 1996; Collier et al., 2003; Blattman and Miguel, 2010). Understanding the relative importance of distortions in the accessibility of each production input is therefore of primary importance in the design of conflict recovery policies. In this article, we quantify the extent of conflict‐induced distortions in the context of the Occupied Palestinian Territories (OPT) during the Second Intifada. The unique features of the Israeli–Palestinian conflict make it particularly suitable for the analysis of the effects of a violent conflict on firms’ access to markets. First, yearly establishment‐level data for a representative sample of firms in the OPT are available for the entire conflict period and before, with detailed information on input usage. Second, the conflict has been characterised by meaningful variation in violence across time and space, about which detailed information is available for the entire period. Third, the economy never collapsed in either the West Bank or the Gaza Strip during the Second Intifada, even if its functions were severely affected. This setting is therefore representative of those situations where economic activity and continuous low intensity violence take place side by side. This is an increasingly common scenario with high policy relevance (World Bank, 2011). We think about conflict as affecting the functioning and accessibility of markets where firms buy their production inputs and/or sell their final goods. We develop this intuition within the conceptual framework proposed by Hsieh and Klenow (2009). In their formalisation of the economy, firms in the same sector are endowed with the same production technology. In the absence of distortions, all firms use inputs in the same proportions, while differences in total factor productivity determine the size of the firm. Firm‐level distortions in the accessibility of markets change the relative demand for inputs and their marginal product. Those firms which find it harder to access the market for one specific input use that input less intensively in production. It follows that heterogeneity arises within sectors in the proportions in which firms combine their inputs. Therefore, within‐sector differences in the production choices of firms which are differentially exposed to conflict can be informative of the relative extent of conflict‐induced distortions in the accessibility of markets. We take these arguments to the data by combining detailed establishment‐level information from the OPT during the Second Intifada (2000–6) with information on conflict intensity, as proxied by conflict‐related Palestinian fatalities. In the first part of the article, we compare the production choices of firms in the same sector across districts experiencing differential changes in conflict intensity. This allows us to net out both overall time trends and unobserved time‐invariant sources of heterogeneity in firms’ operations at the district level, possibly correlated with conflict incidence. Therefore, we identify the impact of conflict intensity out of yearly shocks that are differential at the district level. We find that, within the same sector, firms more exposed to conflict substitute domestically produced materials for imported ones. Our estimates indicate that conflict induces distortions in the accessibility of markets for imported material inputs which are more than three times bigger than the distortions for markets for domestically produced materials. They are also significantly higher than those for labour and capital markets. We show that results are not confounded by changes in relative prices, in the population of active or surveyed firms, omitted variables such as firm localisation, and reverse causality, namely the possibility that workers are themselves involved in conflict. Our results are robust to a number of checks. First, our findings do not change if we use different proxies of conflict intensity, including per capita measures and conflict data from alternative sources. Second, violations of one assumption of the proposed conceptual framework, namely the homotheticity of production functions, do not confound our interpretation of results. We use data from the period before the Second Intifada to identify those sectors for which this assumption does not hold and we show that results are unchanged when we exclude them from our analysis. Finally, we check whether internal and external mobility restrictions imposed by the Israeli Defense Forces (IDF) explain away our results. Evidence shows that the number of days of border closure correlates significantly with input usage but the estimated coefficient associated with our proxy for conflict intensity is not affected. This indicates that the relationship we find between conflict intensity and input usage holds independently from border closures. In the second part of the article, we investigate the mechanisms behind the observed pattern of input substitution. We explore in detail the specific nature of conflict‐induced distortions. The relevant sources of market distortions include trade regulations, transportation obstacles and transaction costs. In particular, evidence shows that importing firms in high conflict localities pay a higher percentage of their inputs before delivery. This is consistent with the hypothesis that the increase in uncertainty due to conflict decreases the bargaining power of firms in their contractual relationship with foreign suppliers, possibly explaining the change in input usage. Several pieces of qualitative evidence further support this hypothesis. Finally, in the third part of the article, we explore the extent to which the conflict‐induced substitution of foreign with domestically produced materials accounts for the fall in output value of firms in high conflict districts. Building upon our conceptual framework, we use data from the no‐conflict period and structurally estimate the factor‐share parameters of the production function. Combining the latter with our previous estimates of the magnitude of conflict‐induced distortions, we perform a counterfactual policy analysis to derive the value of firm‐level output that we would have observed in the absence of conflict. Our estimates suggest that the value of output would have been 6.4% higher for the average firm in the period, with conflict‐induced distortions explaining more than 70% of the drop in the output value of firms in high conflict districts. Our article builds upon and contributes to several strands of the literature. The first refers to those studies which investigate the effects of violent conflict on economic performance. Evidence robustly shows that violent conflict is associated with output fall (Cerra and Saxena, 2008; Chen et al., 2008), lower investment (Eckstein and Tsiddon, 2004) and lower growth (Alesina et al., 1996). A few studies investigate the effect of a violent conflict at the micro level, looking at outcomes such as firm stocks (Abadie and Gardeazabal, 2003; Guidolin and La Ferrara, 2007), investment (Singh, 2013) and firm exit (Camacho and Rodriguez, 2013). Collier and Duponchel (2013) use firm‐level survey data from Sierra Leone and show that conflict reduces the number of employees and their income. Ksoll et al. (2015) use detailed firm‐level export data from Kenya and show that the ethnic violence of 2007 negatively affected export volumes and revenues primarily through an increase in workers’ absence. Studying the representative case of a flower‐packaging plant, Hjort (2014) shows that this same increase in interethnic violence led to higher discrimination among coworkers and lower allocative efficiency within the plant. Finally, Klapper et al. (2013) use census data from Ivory Coast and show that firm productivity decreased with conflict following the coup d’état in 1999. They provide suggestive evidence at the sector level that the increase in firm operating costs, including cost of imported inputs, may drive the results. Our article improves on the existing literature on the microeconomics of conflict in three ways. First, while the majority of previous studies have considered only one sector or some specific group of firms, we use data on a representative sample of the whole population of establishments in the manufacturing sector. Second, our detailed establishment‐level data allow us to look at a wide range of firm‐level figures, including foreign and domestic input usage. Third, we investigate the salience of conflict‐induced distortions in the functioning and accessibility of markets for inputs. We are able to provide direct evidence of the nature of these distortions and estimate their impact on output value. We also contribute to the empirical literature on factor misallocation. Several papers have investigated how market frictions and distortions can affect aggregate output and productivity (Wasmer and Weil, 2004; Restuccia and Rogerson, 2008; Hsieh and Klenow, 2009). A number of studies focus on capital market distortions (Midrigan and Xu, 2014), while others address the specific impact of labour and size‐dependent policies (Guner et al., 2008; Garicano et al., 2016). We contribute to this literature by identifying conflict as an additional determinant of factor misallocation, with a specific focus on developing countries. Moreover, the suggestive evidence we present on the relationship between conflict, uncertainty, bargaining power with foreign suppliers and use of imported inputs is also in line with recent contributions in the literature which specifically investigate the impact of uncertainty on resource misallocation in the economy (Bloom, 2009; Asker et al., 2014; David et al., 2016). Given the salience of our results on imported inputs, our article also relates to the literature which links international trade and firms’ performance. Several theoretical papers have emphasised the importance of intermediate inputs in generating productivity gains (Ethier, 1982; Kugler and Verhoogen, 2012). These predictions are confirmed by robust empirical evidence (Schor, 2004; Amiti and Konings, 2007; Kasahara and Rodrigue, 2008; Topalova and Khandelwal, 2011). In particular, Boehm et al. (2015) use the 2011 Tōhoku earthquake in Japan as a shock to the availability of imports for Japanese affiliates in the United States. They find that output falls roughly one‐for‐one with declines in imports. Although we examine a very different context, our results are consistent with theirs, as we find that conflict‐induced distortions in the accessibility of foreign inputs account for more than 70% of the fall in output value. Finally, our article contributes to the literature on the effect of the Israeli–Palestinian conflict on the Palestinian economy. Previous contributions analysed the impact of the conflict on a number of different outcomes: labour market (Miaari and Sauer, 2011; Cali and Miaari, 2013; Abrahams, 2015), child labour and health (Di Maio and Nandi, 2013; Mansour and Rees, 2012), education (Brück et al., 2014), and welfare (Etkes and Zimring, 2015). While several reports have discussed the aggregate economic impact of the Second Intifada on the Palestinian economy and firms in particular (see for instance World Bank, 2004), there are no empirical estimates of this effect at the micro level, which we examine in relation to both the West Bank and the Gaza Strip. Background: The Israeli–Palestinian Conflict and the Second Intifada The Israeli–Palestinian conflict dates back to 1948, making it one of the longest and most violent conflicts in the world. In 1967, the Six‐Day War ended with the Israeli occupation of the West Bank and the Gaza Strip, previously parts of Jordan and Egypt respectively. In the ensuing years, the conflict went through different phases, each characterised by different levels of violence. Between 1967 and 1993, Israel held the West Bank and the Gaza Strip under military rule. The Israeli occupation led in 1987 to an unarmed but violent and widespread Palestinian uprising. The First Intifada came to an end in 1993, when the Oslo Accord created the Palestinian National Authority (PNA) and gave it limited control over some civilian matters (e.g. education, health and taxation) in both the West Bank and the Gaza Strip. The Israeli authorities maintained control over some strategic issues such as security, border controls and foreign trade between the OPT and Israel, Jordan and Egypt. In the years immediately after the Oslo Accord, violence decreased. This relatively peaceful period ended in September 2000 with the beginning of the so‐called Second Intifada. Also called the Al‐Aqsa Intifada, this period heralded intensified violence between the occupying IDF and the Palestinians. Violence raged on both sides, including Palestinian attacks in Israel, assassination of Palestinian leaders in Palestine and demolitions of Palestinian houses by the IDF. Clashes in the OPT between Palestinians and the IDF have often culminated with the death of both civilians and soldiers. The causes of these clashes have varied; they include communication misunderstandings between Palestinian civilians and IDF at the checkpoints, skirmishes between young Palestinians throwing stones and the IDF, and actual armed fighting between Palestinian militants and the IDF (Sletten and Pedersen, 2003). Given that the Second Intifada has been essentially a period of violent resistance of different sectors of the Palestinian population against the Israeli occupying force, it is not surprising that violence has been highly asymmetrical. Between 2000 and 2006, Palestinians killed 234 Israeli civilians and 226 IDF personnel in the OPT while the IDF caused more than 4,000 Palestinian fatalities, the majority of them non‐combatants (B’Tselem, 2007). The conflict persisted during the whole period but the intensity of violence varied substantially over time and space in both the West Bank and the Gaza Strip. There have been periods of relative calm in different areas. The Palestinian economy never completely collapsed, as opposed to what often happens to countries experiencing genocide episodes or interstate wars. In order to enhance security and control in the OPT during the Second Intifada, the IDF severely scaled up the restrictions on the mobility of goods and people within the OPT as well as across the borders with Israel, Jordan and Egypt. Internal and external movement and access restrictions have been identified as key constraints of Palestinian economic development (World Bank, 2004,8007). Movement restrictions imposed by Israeli authorities stifle economic activity by increasing uncertainty, raising transaction costs, inflating the cost of imported inputs and reducing output. Since there are no ports or airports in the OPT, import and export goods need to travel through Israel, Jordan or Egypt. Israel currently still controls all trade access routes; thus Palestinian foreign trade flows heavily depend on the state of the conflict with Israel. Foreign trade constitutes about 80% of the Palestinian economy gross domestic product, of which 80% is trade with Israel (UNCTAD, 2006). The negative impact of trade restrictions is thus likely to be very sizable. Operating costs of firms in the OPT also increased; 24% of firms in the West Bank and Gaza report political instability as the biggest obstacle to their operations, right after macroeconomic instability at 30%, and before transportation at 9% (World Bank, 2013). Arguably, the Second Intifada never ended. However, violence decreased substantially after 2006. The 2006 elections caused a de facto division of OPT into a Fatah‐controlled West Bank and a Hamas‐controlled Gaza Strip. Israel imposed a complete blockade on the Gaza Strip in 2007 as a sanction against Hamas. The West Bank and the Gaza Strip – which until then had similar institutions and very similar trends in prices and consumption – started to diverge in economic and political terms (Etkes and Zimring, 2015). Conceptual Framework Conflict affects economic activity by inducing distortions in the functioning and accessibility of markets. As a consequence, it changes the relative demand for inputs and their marginal product. On the one hand, the conflict may make it more difficult to access those markets where firms sell their final products and services. Such distortion acts like a tax on the value of output, thus reducing firm size: the demand for all inputs decreases accordingly, and their marginal product increases. On the other hand, conflict may generate (or exacerbate) distortions that are heterogeneous across inputs. Accessing markets for some production inputs may become more difficult than others. In this case, differential distortions across inputs will have a differential impact on input usage: for each pair of inputs, a larger distortion for one input will lead to a decrease in demand and an increase in its marginal product relative to the other. The relative amount of inputs used in production will change accordingly. The way we describe firm‐level distortions is close to Hsieh and Klenow (2009). We build upon their formalisation of the economy to provide the conceptual framework for our analysis. Let the aggregate final output in the economy be produced by a single representative firm that produces a single final good Y with price P. Good Y is produced using a Cobb–Douglas production technology by aggregating the output Ys from all S sectors in the economy, that is: Y=Πs=1SYsθs, (1) with ∑s=1Sθs=1 ⁠. Taking the price P of the final good as given, cost minimisation implies PsYs=θsPY for all s. This set of S first‐order conditions determines the allocation of demand across sectors. Production in each sector s is carried out by a single representative firm that aggregates ns differentiated input products by means of a CES (constant elasticity of substitution) production function. Each input for sector s is supplied by a firm i producing output Ysi at price Psi, and operating under monopolistic competition. Production in each sector s is thus given by: Ys=∑i=1nsYsiσ−1σσσ−1, (2) with σ > 1. Cost minimisation determines the allocation of sector‐level demand Ys across firms. The first‐order conditions imply: Ysi=YsPsiPs−σ⇔Psi=PsYsYsi1σ, (3) for each firm i in sector s. Each firm produces output by means of a Cobb–Douglas production function using as inputs capital K, labour L, and materials M. The production function of firm i is given by: Ysi=AsiKsiαsLsiβsMsi1−αs−βs, (4) so that the output value of the firm is given by: PsiYsi=PsiAsiKsiαsLsiβsMsi1−αs−βs. (5) Here, we depart from the basic framework of Hsieh and Klenow (2009) in that we include materials as input. The empirical analysis will also further differentiate between imported and domestically produced materials, keeping the Cobb–Douglas formulation of the production function. This means we assume that the elasticity of substitution between foreign and domestically produced materials is equal to one. In Section A.2 of Appendix A, we structurally estimate the elasticity of substitution in each sector and provide evidence that supports this hypothesis. We capture distortions faced by firm i in the accessibility of markets for output and inputs using τYi and τXi respectively, where X is one of the production inputs (capital, labour, or materials). Inputs are traded in a centralised market, with firms taking prices as given and equal to w for labour, R for capital, and z for materials. Profits of firm i are given by: (1−τYi)PsiYsi−w(1+τLi)Lsi−R(1+τKi)Ksi−z(1+τMi)Msi. (6) Given product differentiation, in monopolistic competition each firm enjoys a certain degree of market power, so that Psi is endogenous to Ysi. Since Ps and Ys are exogenous to firm i and determined by the allocation of demand at the sector level, we can substitute Psi = Ps(Ys/Ysi)(1/σ) in the firm's profits expression in (6) and maximise with respect to each input. From the corresponding first order conditions we get: Ksi=σ−1σαsPsiYsiR(1+τKi)(1−τYi),Lsi=σ−1σβsPsiYsiw(1+τLi)(1−τYi),Msi=σ−1σ(1−αs−βs)PsiYsiz(1+τMi)(1−τYi). (7) Equation (7) shows that output and input distortions have a different impact on the demand for each input and their marginal product. An increase in output distortion τYi, such as restricted access to the market for final goods, proportionally decreases the demand for all inputs and increases their marginal product. While the firm becomes smaller, the relative marginal products and demand for each input do not change. On the contrary, an increase in the distortion faced by input X (τXi), such as restricted access to the input X market, reduces the demand for that input only, and increases its marginal product. Rearranging (7), we obtain the following expressions for the ratios of input values: RKsiwLsi=αsβs1+τLi1+τKi,RKsizMsi=αs1−αs−βs1+τMi1+τKi,wLsizMsi=βs1−αs−βs1+τMi1+τLi. (8) These equations provide a number of useful results for our analysis. First, they show that input value ratios are invariant with respect to output distortion τYi, but not to input distortions τXi. Therefore, the relative value of inputs used in production is informative of the relative size of distortions in the functioning and accessibility of markets for inputs, independently of distortions related to markets for final goods and services. Second, in the absence of distortions, input value ratios are the same across firms within sectors, as they are uniquely determined by the factor share parameters in the production function. Third, input value ratios are invariant to the firm‐level price Psi. This implies that they do not depend on the competition environment faced by the firm, and thus do not depend on the market structure of each sector s. Equation (8) show that we can infer firm‐level conflict‐induced distortions in the functioning and accessibility of markets for production inputs by comparing input value ratios across firms operating in the same sector which are differentially exposed to conflict. Indeed, any systematic relationship between conflict intensity and input value ratios across firms within sectors would provide evidence of conflict‐induced relative input distortions. For example, if the input value ratio between capital and labour (RKsi/wLsi) was systematically higher for firms operating in conflict areas as compared to other firms in the same sector, this would indicate that conflict increases relatively more firm‐level distortions in the market for labour with respect to capital as measured by (1+τLi/1+τKi). As a final step, we derive firm i's output value. As in Hsieh and Klenow (2009), the optimal firm‐level output price under monopolistic competition is a constant mark‐up over the marginal cost of production. The price is given by: Psi=σσ−11Asi(1−τYi)R(1+τKi)αsαsw(1+τLi)βsβsz(1+τMi)1−αs−βs1−αs−βs. (9) An increase in any firm‐level distortion increases the optimal firm‐level price. Using the within‐sector demand allocation condition in (3), we can rewrite input levels as a function of Psi only and derive the firm‐level demand of inputs given sector‐level production and prices. Substituting into (5), the output value for firm i in sector s can be finally be written as: PsiYsi=σσ−111−τYi1+τKiαsαs1+τLiβsβs1+τMi1−αs−βs1−αs−βs(RKsi)αs(wLsi)βs(zMsi)1−αs−βs. (10) The framework we adopt is static in nature. Firms take distortions in the accessibility of markets as given and choose the amount of inputs to use in production accordingly. Output and prices are also determined simultaneously. We conceptualise conflict as having an impact on firm activity in the current period and on the intensive margin only along the specific dimension of input usage. We therefore abstract from dynamic considerations and, in particular, from the possibility that conflict affects firm entry and exit. In our empirical analysis, we explicitly question the validity of this approach and provide evidence in favour of its underlying assumptions. Data For the purpose of our analysis, we combine two main data sources. The first source of information is the Industry Survey, a yearly representative survey of Palestinian establishments in the manufacturing sector designed and administered by the Palestinian Central Bureau of Statistics (PCBS, 2007). In addition to the information contained in the publicly available version of the data set, we received confidential information as to the district of location of each establishment. We are thus able to map each of the surveyed establishments in each of the 16 Palestinian districts. Moreover, we have information on the ISIC two‐digit sector of each establishment. We can therefore explore the relationship between our firm‐level variables of interest within and between both sectors and districts over time for the years 1999–2006. Our final sample comprises 14,287 establishment observations spanning 8 years. The main variables we use in the empirical analysis are output value, the value of capital and labour, and the value of imported and domestically produced materials used during the year. We measure conflict intensity using the yearly number of Palestinians fatalities caused by the IDF at the district level. Data on fatalities are collected and distributed by the Israeli NGO B’Tselem (B’Tselem, 2007). Both the Israelis and the Palestinians consider these data, which are based on a number of sources and validated by several cross‐checks, to be accurate and reliable. Other scholars studying the Israeli–Palestinian conflict have therefore used them (see for instance Jaeger and Paserman, 2008; Mansour and Rees, 2012). The number of conflict‐related Palestinian fatalities provides the most accurate description of conflict intensity in the OPT during the Second Intifada. The B’Tselem data set provides a rich set of information, such as age, gender and place of residence of the killed, as well as the date, place and a description of the circumstances of the event. This allows us to determine the number of fatalities in each of the 16 Palestinian districts for each year. Table A1 in Appendix A shows the summary statistics of the variables used in the empirical analysis. We observe meaningful variation across establishments in the variables of interest and, in particular, in output value and input value ratios. Figure A1 in Appendix A provides further information on the distribution of some of these variables across Palestinian firms. More than 80% of establishments have fewer than six employees and an output value of less than 400,000 NIS (approximately US$ 50,000). This indicates that small and medium enterprises (SMEs) carry out the largest part of Palestinian manufacturing production. Establishments appear to be evenly distributed across districts, although some of the smallest sectors are clustered in a few districts. As for conflict intensity, the period 2000–6 recorded an average number of 35 Palestinians fatalities per district per year. The standard deviation is equal to 42, meaning that we have considerable variation across the 112 district‐year observations. Empirical Strategy The conceptual framework in Section 2 illustrates how within‐sector differences in input usage across firms can be informative of conflict‐induced distortions in the functioning and accessibility of markets for inputs. Bringing this argument to the data poses several challenges. A simple correlation between conflict intensity and input value ratios cannot be interpreted as evidence of a causal link between conflict and market distortions. Unobservable omitted variables may generate a spurious correlation between the two. Even within the same sector, firms’ access to a given input may be systematically lower in some districts, and these same districts may also be more prone to conflict. For example, conflict may be higher in those districts where the banking sector is less developed. Firms in these districts would have a systematically lower access to capital. We would therefore find the value of capital relative to other inputs to be negatively correlated with conflict intensity, even in the absence of a causal relationship between conflict and access to capital markets. Similarly, the tightening of both household and corporate credit in a given year would simultaneously affect both firms’ access to capital and the opportunity cost of fighting, increasing conflict intensity (Dube and Vargas, 2013). We address these issues by combining cross‐district and time variation in conflict intensity. Including sector, district and year‐fixed effects in our regression specification, we can net out a large fraction of unobservable determinants of establishment‐level outcomes, possibly correlated with conflict intensity. For identification, we therefore rely on yearly shocks to conflict intensity that are differential at the district level. This strategy requires that the number of fatalities – our proxy for conflict intensity – exhibits meaningful variation both across and within districts over time. Figure 1 plots the average number of Palestinians fatalities over time across two subsamples of districts. The continuous line refers to those 25% of districts which recorded the highest number of fatalities in the peak fatalities year (2002), while the dashed line shows the same variable for all other districts. Conflict intensity exhibits meaningful variation over time, with changes being heterogeneous across the two groups of districts. The maps in Figure A2 in Appendix A also confirm cross and within‐district variation in conflict intensity. They show each district classified according to their quintile in the distribution of the yearly level of Palestinian fatalities and their two‐year changes. Fig. 1 View largeDownload slide Cross‐district and Time Conflict Variability Notes. The Figure plots the average number of Palestinians fatalities over time in districts as divided according to the number of fatalities in 2002. Sources. B’Tselem (2007). Colour Figure can be viewed at wileyonlinelibrary.com. Fig. 1 View largeDownload slide Cross‐district and Time Conflict Variability Notes. The Figure plots the average number of Palestinians fatalities over time in districts as divided according to the number of fatalities in 2002. Sources. B’Tselem (2007). Colour Figure can be viewed at wileyonlinelibrary.com. Starting from the conceptual framework, the identification concerns discussed above guide us in the choice of the regression specification to implement. Taking logs of (8), we get: lnRKsizMsi=lnαs1−αs−βs+ln1+τMi1+τKi,lnwLsizMsi=lnβs1−αs−βs+ln1+τMi1+τLilnRKsiwLsi=lnαsβs+ln1+τLi1+τKi,, (11) where RKsi, zMsi, and wLsi are the value of capital, materials and labour used by firm i in sector s. For every pair of inputs (Xsi1,Xsi2) with corresponding prices (p1, p2), we implement: lnp1Xsi1p2Xsi2gt=δt+γg+φs+λ12fatalitiesgt+Zisgt′ρ+εisgt, (12) where p1Xsi1 and p2Xsi2 are the value of input X1 and X2 respectively for firm i operating in sector s surveyed in time t and located in district g, and fatalitiesgt is the number of Palestinians fatalities in year t in the same district, measured in standard deviation units from the district‐year distribution. This allows us to make coefficient estimates directly interpretable as the increase in the dependent variable associated with a one standard deviation increase in fatalitiesgt. The set of sector fixed effects φs captures average differences in input usage across firms at the 2‐digit sector‐level. That is, we net out differences in production technologies, matching the sector‐specific factor shares in the conceptual framework. As anticipated, we also net out average time‐invariant differences across districts and overall time trends by including the full set of district and year fixed effects, γg and δt respectively. In our most demanding specification we interact the two and include sector‐year fixed effects, which net out differential time trends in technology (and input usage) across sectors. Zisgt is a vector of establishment‐specific controls which can proxy for unobserved differences in technology across firms. Finally, εisgt captures residual determinants of input usage. The coefficient of interest λ12 captures systematic differences in the corresponding input value ratio across firms which are differentially exposed to conflict. Results Conflict‐induced Distortions in Input Usage Table 1 reports in each row the corresponding estimates of λ from specification 12 for each of the input value ratios. Standard errors are clustered along both sector‐year and district‐year categories. This allows the residuals uisgt belonging to establishment observations located in the same district and year to be correlated, and the same for the residuals belonging to establishments surveyed in the same year and operating in the same sector. Column (1) shows estimates from a specification where sector, district and year fixed effects are included, together with our main variable of interest fatalitiesgt. Rows (a)–(c) consider the relative value ratios of capital RK, labour wL and materials zM used in production. We find that the input value ratios between the three inputs do not differ across firms which are differentially exposed to conflict, with estimates of the λ coefficient being close to zero and insignificant. Nonetheless, conflict exposure is instead systematically correlated with the relative use of material inputs according to their foreign or domestic origin. In rows (d)–(h), we consider imported materials Mf and domestically produced materials Md separately. Results in row (d) show that a one standard deviation increase in the number of fatalities is associated with a 1.2 increase in the value of domestically produced materials used in production relative to imported ones, with this estimate being significant at the 1% level. Indeed, the value of capital and labour with respect to imported materials increases significantly with conflict intensity (rows (e) and (f)), while the ratio of capital and labour value over the value of domestically produced materials decreases significantly (rows (g) and (h)). All these estimates are significant at the 1% level. Table 1 Input Distortions – Regression Coefficients Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.005 0.003 0.008 0.006 (0.043) (0.043) (0.044) (0.046) (b) ln wLsi/zMsi 0.025 0.022 0.024 0.010 (0.039) (0.037) (0.037) (0.040) (c) ln RKsi/wLsi −0.018 −0.014 −0.015 −0.000 (0.040) (0.039) (0.039) (0.041) (d) lnzdMsid/zfMsif 1.216*** 1.216*** 1.234*** 1.243*** (0.272) (0.271) (0.270) (0.270) (e) lnRKsi/zfMsif 0.523*** 0.521*** 0.538*** 0.551*** (0.122) (0.121) (0.119) (0.127) (f) lnwLsi/zfMsif 0.471*** 0.467*** 0.466*** 0.484*** (0.138) (0.140) (0.140) (0.150) (g) lnRKsi/zdMsid −0.690*** −0.692*** −0.692*** −0.690*** (0.171) (0.171) (0.171) (0.164) (h) lnwLsi/zdMsid −0.668*** −0.671*** −0.668*** −0.662*** (0.184) (0.182) (0.182) (0.182) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.005 0.003 0.008 0.006 (0.043) (0.043) (0.044) (0.046) (b) ln wLsi/zMsi 0.025 0.022 0.024 0.010 (0.039) (0.037) (0.037) (0.040) (c) ln RKsi/wLsi −0.018 −0.014 −0.015 −0.000 (0.040) (0.039) (0.039) (0.041) (d) lnzdMsid/zfMsif 1.216*** 1.216*** 1.234*** 1.243*** (0.272) (0.271) (0.270) (0.270) (e) lnRKsi/zfMsif 0.523*** 0.521*** 0.538*** 0.551*** (0.122) (0.121) (0.119) (0.127) (f) lnwLsi/zfMsif 0.471*** 0.467*** 0.466*** 0.484*** (0.138) (0.140) (0.140) (0.150) (g) lnRKsi/zdMsid −0.690*** −0.692*** −0.692*** −0.690*** (0.171) (0.171) (0.171) (0.164) (h) lnwLsi/zdMsid −0.668*** −0.671*** −0.668*** −0.662*** (0.184) (0.182) (0.182) (0.182) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes *p‐value < 0.1; **p‐value < 0.05; ***p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities variable. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table 1 Input Distortions – Regression Coefficients Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.005 0.003 0.008 0.006 (0.043) (0.043) (0.044) (0.046) (b) ln wLsi/zMsi 0.025 0.022 0.024 0.010 (0.039) (0.037) (0.037) (0.040) (c) ln RKsi/wLsi −0.018 −0.014 −0.015 −0.000 (0.040) (0.039) (0.039) (0.041) (d) lnzdMsid/zfMsif 1.216*** 1.216*** 1.234*** 1.243*** (0.272) (0.271) (0.270) (0.270) (e) lnRKsi/zfMsif 0.523*** 0.521*** 0.538*** 0.551*** (0.122) (0.121) (0.119) (0.127) (f) lnwLsi/zfMsif 0.471*** 0.467*** 0.466*** 0.484*** (0.138) (0.140) (0.140) (0.150) (g) lnRKsi/zdMsid −0.690*** −0.692*** −0.692*** −0.690*** (0.171) (0.171) (0.171) (0.164) (h) lnwLsi/zdMsid −0.668*** −0.671*** −0.668*** −0.662*** (0.184) (0.182) (0.182) (0.182) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.005 0.003 0.008 0.006 (0.043) (0.043) (0.044) (0.046) (b) ln wLsi/zMsi 0.025 0.022 0.024 0.010 (0.039) (0.037) (0.037) (0.040) (c) ln RKsi/wLsi −0.018 −0.014 −0.015 −0.000 (0.040) (0.039) (0.039) (0.041) (d) lnzdMsid/zfMsif 1.216*** 1.216*** 1.234*** 1.243*** (0.272) (0.271) (0.270) (0.270) (e) lnRKsi/zfMsif 0.523*** 0.521*** 0.538*** 0.551*** (0.122) (0.121) (0.119) (0.127) (f) lnwLsi/zfMsif 0.471*** 0.467*** 0.466*** 0.484*** (0.138) (0.140) (0.140) (0.150) (g) lnRKsi/zdMsid −0.690*** −0.692*** −0.692*** −0.690*** (0.171) (0.171) (0.171) (0.164) (h) lnwLsi/zdMsid −0.668*** −0.671*** −0.668*** −0.662*** (0.184) (0.182) (0.182) (0.182) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes *p‐value < 0.1; **p‐value < 0.05; ***p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities variable. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large In columns (2) and (3), we start including establishment‐specific controls as additional regressors. These are meant to proxy and control for unobserved differences in technology across firms, possibly correlated with conflict intensity. One possibility is that those districts which experience a surge in conflict may also be those in which micro and small family‐owned enterprises are more prevalent. These are likely to employ a systematically different production technology, more intense in domestically produced materials. In order to proxy and condition on these differences, we include as controls the fraction of family workers and that of proprietors over the total number of labour units. Finally, in column (4), we also include the full set of district‐year fixed effects ϕst in order to control for sector‐specific trends in input usage. Estimates for all input value ratios are stable across all specifications. Our results show that the within‐district and within‐sector variation over time in the relative value of inputs used by Palestinian establishments is systematically correlated with conflict intensity. We interpret this as evidence that conflict induces distortions which are differential across inputs: the relative value of imported materials is systematically lower for firms located in high conflict environments. Our estimates indicate that when conflict intensity increases by one standard deviation the average value share of imported materials used in production falls from 8.5% of the total value of materials to 3%. Using our conceptual framework, we can also derive for each input the relative size of the distortions associated with conflict. Equation (11) shows that, for every pair of inputs (X1, X2), the relative amount of distortions induced by a one‐standard deviation increase in conflict intensity is given by: expλ^12=1+τXi21+τXi1. (13) Our calculations reveal that one standard deviation increase in conflict intensity is associated with a significant 3.5 increase in the relative distortions faced by firms in accessing the market for imported materials as compared to the one for domestically produced inputs. By the same token, conflict‐induced distortions in imported materials are 1.7 and 1.6 significantly higher than those in capital and labour. Our results shows that firms in high conflict districts use a relatively lower value of foreign produced materials and a relatively higher value of domestically produced ones in production. Given that the relative total value of materials as compared to labour and capital does not change (see Table 1 rows (a), (b), and (c)), we infer that conflict distortions lead firms to substitute domestically produced materials for imported ones. The two inputs are likely not perfect substitutes. Evidence from the trade literature shows how access to imported inputs increases firm productivity (Schor, 2004; Amiti and Konings, 2007; Kasahara and Rodrigue, 2008; Topalova and Khandelwal, 2011). In this sense, our results provide evidence of a specific trade‐related supply‐side mechanism through which conflict may negatively affect output. We explore this possibility in more detail in Section 7. In accordance with the micro‐level evidence above, aggregate trade data show that conflict intensity is associated with changes in Palestinian foreign trade, particularly imports. Figure A3 in Appendix A shows that changes in the number of fatalities over the period are positively correlated with changes in the net balance of trade. This indicates that the (negative) change in the value of imports was much more sizable than the change in the value of exports. The composition of imports also changes differentially. Figures A4 and A5 in Appendix A show the import and export composition respectively in 1999 – the year prior to the outbreak of the Second Intifada – and 2002 – the year with the highest death toll – at the ISIC 1‐digit sector level. While export composition does not appear to be different across the two years, import composition shows meaningful changes. The categories that suffer the largest reduction in import share are: miscellaneous manufacturing articles; manufactured goods classified by material and machinery and transportation equipments. These three categories account for about 55% of the total value of imports in 1999 and only 37% in 2002. Figure A6 also shows the Israeli import value share for each category. Israel is the main supplier of imports in all categories except for miscellaneous manufactures articles. Comparing 1999 and 2002, we see that the Israeli shares of import change proportionally to total import shares, suggesting no differential pattern according to country of origin. Identification threats and sources of variation Our results show that the value of domestically produced materials relative to imported ones used in production increases with conflict. We interpret this as evidence that conflict induces firms to substitute domestically produced inputs for imported ones, with a change in relative quantities. There are five main threats to this interpretation of results. First, holding quantities constant, the relative value of domestic vs. imported inputs used in production may increase because of changes in input prices, and, in particular, a decrease in the relative price of imported inputs. Although firm‐level data on input prices do not exist, several studies indicate that the costs of imports increased considerably during the Second Intifada (Akkaya et al., 2011; World Bank, 2006b). PCBS data show that wholesale price of imports increases during the period, and correlates negatively with conflict intensity. Figure A8 in Appendix A shows that there is no clear pattern of correlation between the ratio of import to domestic wholesale prices and the number of Palestinian fatalities. If anything, the estimated correlation is positive. These results leads us to exclude the possibility that a decrease in the relative price of imported inputs is responsible for our findings. On the contrary, they support the hypothesis that a change in relative quantities is behind the conflict‐induced increase in the relative value of domestic vs. imported inputs we document in Table 1. A second threat to the proposed interpretation of our findings is that conflict may affect firms’ activity on the extensive margin. That is, conflict can affect the size and characteristics of the population of active firms. We directly test for this hypothesis by implementing our baseline regression specification (12) replacing as outcome the (weighted and unweighted) number of firms surveyed in each sector, district and year. In addition, we test for a systematic relationship between conflict intensity and all available firm‐level outcomes measured at the beginning of the year: initial value of capital, inventory of output, and inventory of materials, all in logs. Table 2 reports for each outcome the coefficient estimates from the two most demanding regression specifications. Results show that there is no evidence of a systematic relationship between conflict intensity and the number of operating firms and their baseline characteristics. These findings support the hypothesis that conflict affects firm activity on the intensive margin, with no impact on the population of active firms and their characteristics. Table 2 Conflict and Firm Selection Number of firms Number of firms (weighted) Initial capital Initial Inventory of output Initial inventory of materials (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Fatalities 0.178 0.215 2.116 2.449 −0.055 −0.056 −0.090 −0.123 0.015 0.035 (0.191) (0.266) (1.413) (1.551) (0.040) (0.039) (0.154) (0.155) (0.135) (0.134) Family workersTotal −1.044*** −1.032*** −1.710*** −1.647*** −1.682*** −1.675*** (0.152) (0.157) (0.222) (0.220) (0.279) (0.274) ProprietorsTotal −2.433*** −2.442*** −2.339*** −2.336*** −2.830*** −2.838*** (0.169) (0.171) (0.258) (0.256) (0.243) (0.233) District FE Y Y Y Y Y Y Y Y Y Y Year FE Y n.a. Y n.a. Y n.a. Y n.a. Y n.a. Sector FE Y n.a. Y n.a. Y n.a. Y n.a. Y n.a. Sector × Year FE N Y N Y N Y N Y N Y Observations 1,599 1,599 1,599 1,599 12,449 12,449 10,039 10,039 12,407 12,407 R2 0.000 0.000 0.001 0.001 0.093 0.093 0.034 0.034 0.053 0.053 Number of firms Number of firms (weighted) Initial capital Initial Inventory of output Initial inventory of materials (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Fatalities 0.178 0.215 2.116 2.449 −0.055 −0.056 −0.090 −0.123 0.015 0.035 (0.191) (0.266) (1.413) (1.551) (0.040) (0.039) (0.154) (0.155) (0.135) (0.134) Family workersTotal −1.044*** −1.032*** −1.710*** −1.647*** −1.682*** −1.675*** (0.152) (0.157) (0.222) (0.220) (0.279) (0.274) ProprietorsTotal −2.433*** −2.442*** −2.339*** −2.336*** −2.830*** −2.838*** (0.169) (0.171) (0.258) (0.256) (0.243) (0.233) District FE Y Y Y Y Y Y Y Y Y Y Year FE Y n.a. Y n.a. Y n.a. Y n.a. Y n.a. Sector FE Y n.a. Y n.a. Y n.a. Y n.a. Y n.a. Sector × Year FE N Y N Y N Y N Y N Y Observations 1,599 1,599 1,599 1,599 12,449 12,449 10,039 10,039 12,407 12,407 R2 0.000 0.000 0.001 0.001 0.093 0.093 0.034 0.034 0.053 0.053 Notes *p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is reported on top of each column. In columns (1)–(4), data are collapsed at the two‐digit sector level in each district and year. The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). In all columns but (1), observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table 2 Conflict and Firm Selection Number of firms Number of firms (weighted) Initial capital Initial Inventory of output Initial inventory of materials (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Fatalities 0.178 0.215 2.116 2.449 −0.055 −0.056 −0.090 −0.123 0.015 0.035 (0.191) (0.266) (1.413) (1.551) (0.040) (0.039) (0.154) (0.155) (0.135) (0.134) Family workersTotal −1.044*** −1.032*** −1.710*** −1.647*** −1.682*** −1.675*** (0.152) (0.157) (0.222) (0.220) (0.279) (0.274) ProprietorsTotal −2.433*** −2.442*** −2.339*** −2.336*** −2.830*** −2.838*** (0.169) (0.171) (0.258) (0.256) (0.243) (0.233) District FE Y Y Y Y Y Y Y Y Y Y Year FE Y n.a. Y n.a. Y n.a. Y n.a. Y n.a. Sector FE Y n.a. Y n.a. Y n.a. Y n.a. Y n.a. Sector × Year FE N Y N Y N Y N Y N Y Observations 1,599 1,599 1,599 1,599 12,449 12,449 10,039 10,039 12,407 12,407 R2 0.000 0.000 0.001 0.001 0.093 0.093 0.034 0.034 0.053 0.053 Number of firms Number of firms (weighted) Initial capital Initial Inventory of output Initial inventory of materials (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Fatalities 0.178 0.215 2.116 2.449 −0.055 −0.056 −0.090 −0.123 0.015 0.035 (0.191) (0.266) (1.413) (1.551) (0.040) (0.039) (0.154) (0.155) (0.135) (0.134) Family workersTotal −1.044*** −1.032*** −1.710*** −1.647*** −1.682*** −1.675*** (0.152) (0.157) (0.222) (0.220) (0.279) (0.274) ProprietorsTotal −2.433*** −2.442*** −2.339*** −2.336*** −2.830*** −2.838*** (0.169) (0.171) (0.258) (0.256) (0.243) (0.233) District FE Y Y Y Y Y Y Y Y Y Y Year FE Y n.a. Y n.a. Y n.a. Y n.a. Y n.a. Sector FE Y n.a. Y n.a. Y n.a. Y n.a. Y n.a. Sector × Year FE N Y N Y N Y N Y N Y Observations 1,599 1,599 1,599 1,599 12,449 12,449 10,039 10,039 12,407 12,407 R2 0.000 0.000 0.001 0.001 0.093 0.093 0.034 0.034 0.053 0.053 Notes *p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is reported on top of each column. In columns (1)–(4), data are collapsed at the two‐digit sector level in each district and year. The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). In all columns but (1), observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large A third threat is related to the identification strategy: firms located in districts where conflict increases disproportionally more could have already been on a different trajectory in terms of production choices and input usage. However, Figures 1 and A2 indicate that identification in this context does not stem from divergent linear trends but rather originates from yearly shocks in conflict intensity at the district level. The actual evolution of conflict intensity mitigates the concern that pre‐existing trends may be generating our results. Nonetheless, we address this concern by including in our main regression specification both the current and next year level of conflict intensity, still measured by the number of fatalities. Table A8 in Appendix A shows the corresponding results. Input value ratios are significantly correlated with current but not with next year conflict intensity. This allows us to conclude that pre‐existing movements in input usage do not with conflict intensity, validating our approach to identification. A fourth concern is that our results are only capturing systematic differences related to firm localisation. Firms located closer to the borders are likely to be more intensive in imported material inputs with respect to other establishments in the same sector. By the same token, districts near the border with Israel may experience higher variation in the number of Palestinian fatalities over the period. We condition on the role of distance from the border by saturating the input value ratio regression specifications (12) with a full set of year fixed effects interacted with a measure of the road distance of the district capital from the closest entry passage. Corresponding estimates are reported in Table A9 in Appendix A. Point estimates are very similar to those reported in Table 1. This suggests firms’ localisation does not confound our results. A final concern is that reverse causality could drive our empirical results. An independent shock to the availability of foreign inputs at the district level may lead to changes in input usage, possibly diminishing the marginal product of labour and thus wages. This would decrease the opportunity cost of fighting, thus increasing conflict intensity. However, our results show that the relative use of labour and number of fatalities are not correlated (see Table 1, rows (b) and (c)), in contrast with what one would expect if workers played an important role in the conflict. Still, holding constant the total amount of labour, it is possible that the loss of workers involved in the conflict lead to changes in workforce composition. However, this change would generate the result we find on input value ratios (substitution of foreign for local material inputs) only under a very precise assumption: workers substituting those participating in the conflict are more complementary to locally produced materials than to imported materials. No evidence supports this assumption. Robustness Checks Alternative measures of conflict We have obtained our results using one specific measure of conflict intensity: the number of Palestinian fatalities. As explained in Section 4, in order to make coefficient estimates readily interpretable, we measure the number of fatalities in standard deviation units. In evaluating the robustness of results, we first implement our main regression specification (12) using as a measure of conflict intensity the actual number of Palestinian fatalities in district d and time t. Table A4 in Appendix A shows the corresponding results, which confirm our previous findings. As a second alternative proxy of conflict intensity, we use the per capita number of Palestinian fatalities in district d and time t. Table A5 in Appendix A reports the corresponding results. As before, results indicate a strong systematic relationship between conflict exposure and the relative use of foreign vs. domestically produced material inputs. To corroborate our estimates further, we derive a third proxy of conflict intensity from a separate source: the Integrated Crisis Early Warning System (ICEWS) data set (Shilliday and Lautenschlager, 2012). This proxy counts the number of hostile political interactions occurred in each district in the OPT in each year during the Second Intifada. For consistency with our previous analysis, we restrict our attention to those hostile events triggered by the IDF. As before, we standardise the count variable and divide it by the standard deviation of the corresponding district‐year distribution. Table A6 in Appendix A shows the corresponding estimates. Remarkably, using a completely separate source of data, we unveil the same empirical regularities. Coefficient magnitudes indicate that conflict‐induced distortions in relative input usage are highest for foreign vs. domestically produced materials. The corresponding coefficient is lower than in our baseline results but significant at the 10% level. We interpret the smaller size of the effect as following from the additional noise that characterises this measure of conflict intensity, as hostile events do not always result in Palestinian fatalities. This may result in classical measurement error, biasing our estimates towards zero. Despite this, all coefficient magnitudes are ordered as in Table 1, confirming our previous findings. The same holds when we divide the number of ICEWS hostile events by population, as shown by Table A7 in Appendix A. Non‐homothetic production functions and demand‐side effects The validity of our interpretation of empirical results rests on the assumptions of Hsieh and Klenow (2009). In particular, the model assumes homothetic production functions. While firms in the same sector can be heterogeneous in terms of total factor productivity, this assumption ensures that – in the absence of distortions – they will all use inputs in the same proportion. It follows that within‐sector differences in input value ratios which relate systematically to conflict exposure can be interpreted as evidence of the relative amount of distortions induced by the conflict in the accessibility of markets for inputs. This is no longer the case if production functions are non‐homothetic. When differences in factor shares are correlated with firm's output, changes in input usage could be the result of conflict‐induced changes in demand. In particular, if firms with lower output were to employ relatively more domestically produced materials than imported ones, output fall might drive the observed increase in the amount of domestically produced materials used in production, with possibly no role played by conflict‐induced distortions in the accessibility of markets. To rule out this possibility, we use data from year 1999 – the year before the start of the Second Intifada – and directly test whether firm size correlates with input value ratios. This is based on our conceptualisation of 1999 as a benchmark economy with no conflict‐induced distortions. Figure 2(a) plots the relationship between the (log of) value ratio of domestically produced materials over imported ones and the (log of) output value in 1999, averaging out sector‐level means. For any given level of output value, we observe substantial heterogeneity across firms. The line fitting the scatterplot is downward sloping, with the corresponding coefficient being significant at the 5% level. This means that, before the start of the conflict and within sectors, firms with higher output value employed relatively fewer domestically produced materials with respect to imported ones. While these results may suggest that demand‐driven mechanisms may be at work, further analysis of the data reveals that before the start of the conflict the relationship between input value ratio and output value is non‐significant for 15 out of the 25 sectors (to which 66% of surveyed establishments belong), as shown in Figure 2(b). We thus re‐estimate the coefficient λ from our regression specification using only the observations belonging to this restricted sample for which the homotheticity assumption finds support in the data. Table A10 of Appendix A reports the corresponding estimates. Results are almost exactly the same as the baseline ones in Table A1. Under the assumption that the within‐sector relationship between factor shares and output value remained constant over time, this suggests that violations of the homotheticity assumption in the data do not confound our interpretation of the empirical findings: the observed changes in input usage are attributable to conflict‐induced distortions in the supply side of the economy rather than to a fall in demand. Fig. 2 View largeDownload slide Within‐sector Heterogeneity in Technology and Output Value. (a) All Sectors. (b) Restricted Sample Notes. The left and right Figures plot the within‐sector residual log of the ratio between the value of domestically produced materials and imported materials used over the residual log of output value for firms in 1999. Circle size correspond to the observation's weight in the sample. The top Figure shows the relationship of interest using all available observations, while the bottom Figure considers only those sectors for which the relationship between the two variables is non‐significant. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Fig. 2 View largeDownload slide Within‐sector Heterogeneity in Technology and Output Value. (a) All Sectors. (b) Restricted Sample Notes. The left and right Figures plot the within‐sector residual log of the ratio between the value of domestically produced materials and imported materials used over the residual log of output value for firms in 1999. Circle size correspond to the observation's weight in the sample. The top Figure shows the relationship of interest using all available observations, while the bottom Figure considers only those sectors for which the relationship between the two variables is non‐significant. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Internal and external mobility restrictions As we discussed in Section 1, one of the distinctive features of the Israeli–Palestinian conflict is the presence of restrictions imposed by the IDF on the mobility of goods and people within the OPT as well as across the border with Israel. In order to interpret our findings correctly, it is thus important to explore the extent to which changes in the number of Palestinian fatalities correlate with changes in internal and external mobility restrictions. Internal mobility restrictions include checkpoints, roadblocks and barriers between villages within the OPT put in place by the IDF. These measures may increase travel time and thus affect the accessibility of both foreign and domestic inputs. However, the data needed to estimate their impact on firm input choice suffer from a serious limitations: data are missing for the most violent year (2002) and for all districts in the Gaza Strip for all years. This implies that it is not possible to quantify the role of internal mobility restriction in our framework precisely using existing data. External mobility restrictions take the form of closure of the border between Israel and the OPT imposed by the IDF. During border closure days, movements of workers and import and export of goods are interrupted (World Bank, 2008). This represents a negative shock to the accessibility of foreign markets for Palestinian firms. If such shock was correlated with conflict intensity, we would be attributing to local conflict conditions (proxied by the number of Palestinian fatalities) the negative effect of border closures on the availability of foreign produced materials. Figure 3 shows the evolution of the total number of Palestinian fatalities over time in each quarter from 2000 to 2006, together with the quarterly number of days of border closure in the same period. A visual investigation reveals no systematic relationship between the two. Indeed, the estimated correlation is equal to 0.12 and insignificant. This means that the variation in the incidence of border closures does not overlap with the one captured by our proxy for conflict intensity, and that the changes in input usage we document are independent from border closure. Moreover, notice that, as long as the impact of closures is uniform across Palestinian districts, the year fixed effects in our specification already take that into account. Fig. 3 View largeDownload slide Conflict and Border Closures Notes. The Figure plots the number of days of border closures in each quarter between the 3rd quarter of 2000 and the end of 2006, together with the total number of Palestinians fatalities over time. Sources. Palestinian Central Bureau of Statistics, B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Fig. 3 View largeDownload slide Conflict and Border Closures Notes. The Figure plots the number of days of border closures in each quarter between the 3rd quarter of 2000 and the end of 2006, together with the total number of Palestinians fatalities over time. Sources. Palestinian Central Bureau of Statistics, B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Yet, closures may affect firms located in different districts in different ways. Firms that are located near the Israeli border may find it easier to cope with closures, as they can take advantage of non‐closure days to acquire foreign inputs. Or, firms next to the border may be more vulnerable to border closures and adjust their input usage to a larger extent. In both cases, if the largest changes in conflict intensity were in districts next to the border, the differential effect of closures would be nested in our estimate of the impact of conflict on input usage. We provide direct evidence of the impact of border closures by including the interaction of the yearly number of closure days with the road distance of firm location from the closest entry gate as an additional regressor in our specification. We therefore allow the effect of border closures on input usage to be heterogeneous according to the distance from the border, capturing either scenario described above. Table 3 reports the estimated coefficients of the variables of interest from this augmented specification. The coefficient of the interaction variable is significant: the impact of border closures is systematically higher for firms located farther away from the border. The higher the number of closure days and the higher the distance from the entry gates, the more firms substitute domestically produced inputs for imported ones. Perhaps more importantly, when compared to the baseline results in Table 1, the coefficients of the fatalities variable are almost unaffected. Table 3 Input Distortions, Fatalities and Border Closures Regression Coefficients (1) (2) (3) (4) Panel (a) Dependent variable: lnzdMsid/zfMsif fatalities 1.263*** 1.263*** 1.279*** 1.290*** (0.247) (0.247) (0.247) (0.246) closuredays × dtpassage 0.010** 0.010** 0.010** 0.010** (0.004) (0.004) (0.004) (0.004) Panel (b) Dependent variable: lnRKsi/zfMsif fatalities 0.547*** 0.545*** 0.562*** 0.575*** (0.115) (0.114) (0.112) (0.120) closuredays × dtpassage 0.005** 0.005** 0.005** 0.005** (0.002) (0.002) (0.002) (0.002) Panel (c) Dependent variable: lnwLsi/zfMsif fatalities 0.499*** 0.494*** 0.492*** 0.515*** (0.116) (0.119) (0.119) (0.127) closuredays × dtpassage 0.006** 0.006** 0.006** 0.006*** (0.002) (0.002) (0.002) (0.002) Panel (d) Dependent variable: lnRKsi/zdMsid fatalities −0.713*** −0.715*** −0.715*** −0.713*** (0.157) (0.157) (0.157) (0.151) closuredays × dtpassage −0.005* −0.005* −0.005* −0.005* (0.002) (0.002) (0.002) (0.002) Panel (e) Dependent variable: lnwLsi/zdMsid fatalities −0.694*** −0.698*** −0.694*** −0.690*** (0.169) (0.166) (0.166) (0.166) closuredays × dtpassage −0.006** −0.006** −0.006** −0.006** (0.002) (0.002) (0.002) (0.002) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y (1) (2) (3) (4) Panel (a) Dependent variable: lnzdMsid/zfMsif fatalities 1.263*** 1.263*** 1.279*** 1.290*** (0.247) (0.247) (0.247) (0.246) closuredays × dtpassage 0.010** 0.010** 0.010** 0.010** (0.004) (0.004) (0.004) (0.004) Panel (b) Dependent variable: lnRKsi/zfMsif fatalities 0.547*** 0.545*** 0.562*** 0.575*** (0.115) (0.114) (0.112) (0.120) closuredays × dtpassage 0.005** 0.005** 0.005** 0.005** (0.002) (0.002) (0.002) (0.002) Panel (c) Dependent variable: lnwLsi/zfMsif fatalities 0.499*** 0.494*** 0.492*** 0.515*** (0.116) (0.119) (0.119) (0.127) closuredays × dtpassage 0.006** 0.006** 0.006** 0.006*** (0.002) (0.002) (0.002) (0.002) Panel (d) Dependent variable: lnRKsi/zdMsid fatalities −0.713*** −0.715*** −0.715*** −0.713*** (0.157) (0.157) (0.157) (0.151) closuredays × dtpassage −0.005* −0.005* −0.005* −0.005* (0.002) (0.002) (0.002) (0.002) Panel (e) Dependent variable: lnwLsi/zdMsid fatalities −0.694*** −0.698*** −0.694*** −0.690*** (0.169) (0.166) (0.166) (0.166) closuredays × dtpassage −0.006** −0.006** −0.006** −0.006** (0.002) (0.002) (0.002) (0.002) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities variable and the interaction of the yearly number of days of border closure with the road distance of the district capital from the closest entry passage as measured in 10 kilometres units. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The variable closure days captures the yearly number of days of border closure, while dtpassage measures road distance of the district capital from the closest entry passage as measured in 10 kilometre units. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table 3 Input Distortions, Fatalities and Border Closures Regression Coefficients (1) (2) (3) (4) Panel (a) Dependent variable: lnzdMsid/zfMsif fatalities 1.263*** 1.263*** 1.279*** 1.290*** (0.247) (0.247) (0.247) (0.246) closuredays × dtpassage 0.010** 0.010** 0.010** 0.010** (0.004) (0.004) (0.004) (0.004) Panel (b) Dependent variable: lnRKsi/zfMsif fatalities 0.547*** 0.545*** 0.562*** 0.575*** (0.115) (0.114) (0.112) (0.120) closuredays × dtpassage 0.005** 0.005** 0.005** 0.005** (0.002) (0.002) (0.002) (0.002) Panel (c) Dependent variable: lnwLsi/zfMsif fatalities 0.499*** 0.494*** 0.492*** 0.515*** (0.116) (0.119) (0.119) (0.127) closuredays × dtpassage 0.006** 0.006** 0.006** 0.006*** (0.002) (0.002) (0.002) (0.002) Panel (d) Dependent variable: lnRKsi/zdMsid fatalities −0.713*** −0.715*** −0.715*** −0.713*** (0.157) (0.157) (0.157) (0.151) closuredays × dtpassage −0.005* −0.005* −0.005* −0.005* (0.002) (0.002) (0.002) (0.002) Panel (e) Dependent variable: lnwLsi/zdMsid fatalities −0.694*** −0.698*** −0.694*** −0.690*** (0.169) (0.166) (0.166) (0.166) closuredays × dtpassage −0.006** −0.006** −0.006** −0.006** (0.002) (0.002) (0.002) (0.002) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y (1) (2) (3) (4) Panel (a) Dependent variable: lnzdMsid/zfMsif fatalities 1.263*** 1.263*** 1.279*** 1.290*** (0.247) (0.247) (0.247) (0.246) closuredays × dtpassage 0.010** 0.010** 0.010** 0.010** (0.004) (0.004) (0.004) (0.004) Panel (b) Dependent variable: lnRKsi/zfMsif fatalities 0.547*** 0.545*** 0.562*** 0.575*** (0.115) (0.114) (0.112) (0.120) closuredays × dtpassage 0.005** 0.005** 0.005** 0.005** (0.002) (0.002) (0.002) (0.002) Panel (c) Dependent variable: lnwLsi/zfMsif fatalities 0.499*** 0.494*** 0.492*** 0.515*** (0.116) (0.119) (0.119) (0.127) closuredays × dtpassage 0.006** 0.006** 0.006** 0.006*** (0.002) (0.002) (0.002) (0.002) Panel (d) Dependent variable: lnRKsi/zdMsid fatalities −0.713*** −0.715*** −0.715*** −0.713*** (0.157) (0.157) (0.157) (0.151) closuredays × dtpassage −0.005* −0.005* −0.005* −0.005* (0.002) (0.002) (0.002) (0.002) Panel (e) Dependent variable: lnwLsi/zdMsid fatalities −0.694*** −0.698*** −0.694*** −0.690*** (0.169) (0.166) (0.166) (0.166) closuredays × dtpassage −0.006** −0.006** −0.006** −0.006** (0.002) (0.002) (0.002) (0.002) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities variable and the interaction of the yearly number of days of border closure with the road distance of the district capital from the closest entry passage as measured in 10 kilometres units. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The variable closure days captures the yearly number of days of border closure, while dtpassage measures road distance of the district capital from the closest entry passage as measured in 10 kilometre units. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Evidence thus shows that the closure of borders plays a role in inducing changes in input usage. The impact of closures has the same direction as the impact of local conflict intensity. Yet, the two are independent: the estimated coefficient of conflict intensity does not change significantly when we account for border closures, as the two variables are not correlated. This indicates that the relationship we find between conflict intensity and input usage holds independently from border closures and operates through other channels. Unpacking Distortions Results from the previous Section show that firms located in districts that are differentially more exposed to conflict substitute domestically produced materials for imported ones in production. We attribute these changes to conflict‐induced distortions in the functioning and accessibility of markets for imported materials, which lead firms to change their relative demand for inputs. In the following, we explore the nature of these distortions in detail. We rely on additional evidence from the World Bank Enterprise Survey (World Bank, 2006a). This survey is particularly rich in information on firms’ operations and obstacles they face. Data are available for 401 enterprises located in the West Bank and the Gaza Strip. Surveyed enterprises belong to the manufacturing, construction, transport, and service sectors with more than five employees. For those located in the West Bank, information on the village of location is also available. We merge the World Bank Enterprise Survey data with the B’Tselem data on Palestinians fatalities and derive the number of fatalities in 2006 in the village where the firm is located. Despite the fact that we are using data from a cross‐section of enterprises and only for the year 2006, we are confident that the extensive information in the survey coupled with the finer spatial variation in fatalities we can use in this analysis can provide insights into the sources of distortions affecting imported input usage during the Second Intifada. Given our previous results on foreign input usage, here we ask whether conflict affects firm operations in a differential way according to their importing status. We thus implement the following regression specification: yig=α+βmig+γfatalitiesg+δfatalitiesg×mig+Xig′θ+Zg′ψ+vig, (14) where yig measures the relevance of an obstacle for the operations of firm i located in village g, on a scale from 1 (little) to 4 (very). mig is a dummy equal to one if the firm imports any good or service and fatalitiesg is the number of Palestinians fatalities in village g. Xig is a vector of pre‐determined firm‐level variables, while Zg is a vector of village‐level controls. The coefficient δ captures whether the relationship between conflict intensity and firm‐level outcomes is systematically different according to the firm importing status. As obstacles to firms operations we consider: (a) custom and trade regulation, (b) transportation, and (c) percentage of inputs paid before delivery. Table 4 shows the corresponding coefficient estimates from regression 14. In columns (1)–(5), we progressively add sector and district fixed effects, together with a number of control variables. We therefore rely on variation in conflict intensity across localities within sectors and districts, and look at its differential relationship with operation obstacles depending on the firm's importing status. Panel (a) shows how the score attached to custom and trade regulation as an obstacle is differentially and systematically higher for importing firms operating in high conflict localities. Panel (b) shows that the same pattern holds when we consider the score attached to transportation as an obstacle to firm operations. Evidence in panel (c) is particularly interesting. It shows that the percentage of inputs paid before delivery is 4–6 percentage points higher for importing firms in high conflict localities. This suggests that the uncertainty related to conflict is disproportionally more salient for transactions on foreign markets with respect to local markets. This is consistent with the hypothesis that exposure to conflict shapes the terms of the contract between the firm and foreign suppliers, decreasing the bargaining power of the former. This increases the operating costs that the firm faces when accessing the market for foreign produced inputs, making imported inputs relatively more costly. Table 4 Conflict and Obstacles to Firms’ Operations (1) (2) (3) (4) (5) (6) Panel (a) Dependent variable: customs and trade regulations as main obstacle fatalities −0.227*** −0.247*** −0.102 −0.031 −0.057 −0.226** (0.05) (0.05) (0.10) (0.10) (0.10) (0.11) Importer 0.287 0.355 0.362 0.398 0.312 0.089 (0.34) (0.34) (0.31) (0.30) (0.30) (0.43) fatalities × importer 0.249*** 0.237*** 0.256*** 0.259*** 0.312*** 0.470*** (0.06) (0.06) (0.05) (0.06) (0.06) (0.08) fatalities × age 0.013*** (0.00) fatalities × age × importer −0.013*** (0.00) Panel (b) Dependent variable: trasportation as main obstacle fatalities −0.254*** −0.257*** −0.136* −0.045 −0.063 −0.104 (0.07) (0.07) (0.08) (0.07) (0.07) (0.07) Importer 0.255 0.305 0.367 0.462 0.445 0.456 (0.34) (0.34) (0.31) (0.30) (0.27) (0.38) fatalities × importer 0.296*** 0.288*** 0.288*** 0.243*** 0.287*** 0.339*** (0.07) (0.07) (0.07) (0.07) (0.06) (0.07) fatalities × age 0.004 (0.00) fatalities × age × importer −0.004 (0.00) Panel (c) Dependent variable: percentage of inputs paid before delivery fatalities −0.013 −0.003 −0.008 −0.013 −0.018 −0.036 (0.02) (0.01) (0.02) (0.03) (0.03) (0.03) Importer 0.110 0.100 0.123* 0.095 0.089 −0.001 (0.07) (0.07) (0.07) (0.07) (0.07) (0.12) fatalities × importer 0.039** 0.041*** 0.039*** 0.055*** 0.067*** 0.125*** (0.02) (0.01) (0.01) (0.02) (0.02) (0.02) fatalities × age 0.001 (0.00) fatalities × age × importer −0.003** (0.00) Population 1997 N Y Y Y Y Y Age N N N Y Y Y Sales in 2003 N N N Y Y Y Employment in 2003 N N N Y Y Y Other controls N N N N Y Y Age × importer N N N N N Y Sector FE N N Y Y Y Y District FE N N Y Y Y Y (1) (2) (3) (4) (5) (6) Panel (a) Dependent variable: customs and trade regulations as main obstacle fatalities −0.227*** −0.247*** −0.102 −0.031 −0.057 −0.226** (0.05) (0.05) (0.10) (0.10) (0.10) (0.11) Importer 0.287 0.355 0.362 0.398 0.312 0.089 (0.34) (0.34) (0.31) (0.30) (0.30) (0.43) fatalities × importer 0.249*** 0.237*** 0.256*** 0.259*** 0.312*** 0.470*** (0.06) (0.06) (0.05) (0.06) (0.06) (0.08) fatalities × age 0.013*** (0.00) fatalities × age × importer −0.013*** (0.00) Panel (b) Dependent variable: trasportation as main obstacle fatalities −0.254*** −0.257*** −0.136* −0.045 −0.063 −0.104 (0.07) (0.07) (0.08) (0.07) (0.07) (0.07) Importer 0.255 0.305 0.367 0.462 0.445 0.456 (0.34) (0.34) (0.31) (0.30) (0.27) (0.38) fatalities × importer 0.296*** 0.288*** 0.288*** 0.243*** 0.287*** 0.339*** (0.07) (0.07) (0.07) (0.07) (0.06) (0.07) fatalities × age 0.004 (0.00) fatalities × age × importer −0.004 (0.00) Panel (c) Dependent variable: percentage of inputs paid before delivery fatalities −0.013 −0.003 −0.008 −0.013 −0.018 −0.036 (0.02) (0.01) (0.02) (0.03) (0.03) (0.03) Importer 0.110 0.100 0.123* 0.095 0.089 −0.001 (0.07) (0.07) (0.07) (0.07) (0.07) (0.12) fatalities × importer 0.039** 0.041*** 0.039*** 0.055*** 0.067*** 0.125*** (0.02) (0.01) (0.01) (0.02) (0.02) (0.02) fatalities × age 0.001 (0.00) fatalities × age × importer −0.003** (0.00) Population 1997 N Y Y Y Y Y Age N N N Y Y Y Sales in 2003 N N N Y Y Y Employment in 2003 N N N Y Y Y Other controls N N N N Y Y Age × importer N N N N N Y Sector FE N N Y Y Y Y District FE N N Y Y Y Y Notes *p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable in panel (a) is whether customs and trade regulations are reported as obstacles to the operations of the firm on a 1 to 4 scale. Dependent variable in panel (b) is whether transportations are reported as obstacles to the operations of the firm on a 1 to 4 scale. Dependent variable in panel (c) is the share of inputs and services that the firm reports to pay before delivery. The main regressors are as follows: the number of Palestinians killed by IDF in the year and locality where surveyed establishment is located (measured in standard deviation units), dummy for whether the firm reports a positive share of imported inputs in production, and the interaction between the two. We also include as other controls a dummy capturing whether the firm is female‐owned, and its legal status. Depending on the dependent variable used and the implemented specification, the number of observations is between 192 and 222. Sources. World Bank Enterprise Survey 2006, B’Tselem. View Large Table 4 Conflict and Obstacles to Firms’ Operations (1) (2) (3) (4) (5) (6) Panel (a) Dependent variable: customs and trade regulations as main obstacle fatalities −0.227*** −0.247*** −0.102 −0.031 −0.057 −0.226** (0.05) (0.05) (0.10) (0.10) (0.10) (0.11) Importer 0.287 0.355 0.362 0.398 0.312 0.089 (0.34) (0.34) (0.31) (0.30) (0.30) (0.43) fatalities × importer 0.249*** 0.237*** 0.256*** 0.259*** 0.312*** 0.470*** (0.06) (0.06) (0.05) (0.06) (0.06) (0.08) fatalities × age 0.013*** (0.00) fatalities × age × importer −0.013*** (0.00) Panel (b) Dependent variable: trasportation as main obstacle fatalities −0.254*** −0.257*** −0.136* −0.045 −0.063 −0.104 (0.07) (0.07) (0.08) (0.07) (0.07) (0.07) Importer 0.255 0.305 0.367 0.462 0.445 0.456 (0.34) (0.34) (0.31) (0.30) (0.27) (0.38) fatalities × importer 0.296*** 0.288*** 0.288*** 0.243*** 0.287*** 0.339*** (0.07) (0.07) (0.07) (0.07) (0.06) (0.07) fatalities × age 0.004 (0.00) fatalities × age × importer −0.004 (0.00) Panel (c) Dependent variable: percentage of inputs paid before delivery fatalities −0.013 −0.003 −0.008 −0.013 −0.018 −0.036 (0.02) (0.01) (0.02) (0.03) (0.03) (0.03) Importer 0.110 0.100 0.123* 0.095 0.089 −0.001 (0.07) (0.07) (0.07) (0.07) (0.07) (0.12) fatalities × importer 0.039** 0.041*** 0.039*** 0.055*** 0.067*** 0.125*** (0.02) (0.01) (0.01) (0.02) (0.02) (0.02) fatalities × age 0.001 (0.00) fatalities × age × importer −0.003** (0.00) Population 1997 N Y Y Y Y Y Age N N N Y Y Y Sales in 2003 N N N Y Y Y Employment in 2003 N N N Y Y Y Other controls N N N N Y Y Age × importer N N N N N Y Sector FE N N Y Y Y Y District FE N N Y Y Y Y (1) (2) (3) (4) (5) (6) Panel (a) Dependent variable: customs and trade regulations as main obstacle fatalities −0.227*** −0.247*** −0.102 −0.031 −0.057 −0.226** (0.05) (0.05) (0.10) (0.10) (0.10) (0.11) Importer 0.287 0.355 0.362 0.398 0.312 0.089 (0.34) (0.34) (0.31) (0.30) (0.30) (0.43) fatalities × importer 0.249*** 0.237*** 0.256*** 0.259*** 0.312*** 0.470*** (0.06) (0.06) (0.05) (0.06) (0.06) (0.08) fatalities × age 0.013*** (0.00) fatalities × age × importer −0.013*** (0.00) Panel (b) Dependent variable: trasportation as main obstacle fatalities −0.254*** −0.257*** −0.136* −0.045 −0.063 −0.104 (0.07) (0.07) (0.08) (0.07) (0.07) (0.07) Importer 0.255 0.305 0.367 0.462 0.445 0.456 (0.34) (0.34) (0.31) (0.30) (0.27) (0.38) fatalities × importer 0.296*** 0.288*** 0.288*** 0.243*** 0.287*** 0.339*** (0.07) (0.07) (0.07) (0.07) (0.06) (0.07) fatalities × age 0.004 (0.00) fatalities × age × importer −0.004 (0.00) Panel (c) Dependent variable: percentage of inputs paid before delivery fatalities −0.013 −0.003 −0.008 −0.013 −0.018 −0.036 (0.02) (0.01) (0.02) (0.03) (0.03) (0.03) Importer 0.110 0.100 0.123* 0.095 0.089 −0.001 (0.07) (0.07) (0.07) (0.07) (0.07) (0.12) fatalities × importer 0.039** 0.041*** 0.039*** 0.055*** 0.067*** 0.125*** (0.02) (0.01) (0.01) (0.02) (0.02) (0.02) fatalities × age 0.001 (0.00) fatalities × age × importer −0.003** (0.00) Population 1997 N Y Y Y Y Y Age N N N Y Y Y Sales in 2003 N N N Y Y Y Employment in 2003 N N N Y Y Y Other controls N N N N Y Y Age × importer N N N N N Y Sector FE N N Y Y Y Y District FE N N Y Y Y Y Notes *p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable in panel (a) is whether customs and trade regulations are reported as obstacles to the operations of the firm on a 1 to 4 scale. Dependent variable in panel (b) is whether transportations are reported as obstacles to the operations of the firm on a 1 to 4 scale. Dependent variable in panel (c) is the share of inputs and services that the firm reports to pay before delivery. The main regressors are as follows: the number of Palestinians killed by IDF in the year and locality where surveyed establishment is located (measured in standard deviation units), dummy for whether the firm reports a positive share of imported inputs in production, and the interaction between the two. We also include as other controls a dummy capturing whether the firm is female‐owned, and its legal status. Depending on the dependent variable used and the implemented specification, the number of observations is between 192 and 222. Sources. World Bank Enterprise Survey 2006, B’Tselem. View Large Several pieces of qualitative evidence further support this hypothesis. Service Group (2002) reports that the outbreak of the conflict led many Israeli suppliers to demand advance payment instead of payment‐upon‐receipt to protect themselves from collection difficulties. This has created serious problem for Palestinian firms, which face financing constraints. According to UNCTAD (2004), the outbreak of hostilities has made banks adopt more conservative lending policies and restrict access to trade financing services. This has increased the liquidity problems of Palestinian companies, making it extremely difficult to pay for inputs in advance. Indeed, in 2002 – when conflict intensity was the highest – more than 80% of the Palestinian SMEs relied on internal savings for covering operating costs. Moreover, using data from an original survey of traders, UNWFP (2009) shows how West Bank traders localise their activities within their districts when Israeli suppliers reduce their credit lines. Taken together, this evidence suggests that a link between conflict‐induced uncertainty and credit tightening from Israeli suppliers can possibly explain the substitution of foreign with domestically produced materials that we documented in our analysis. The estimates in column (6) further validate this interpretation of results. We implement a triple difference specification where we look at whether the differential relationship between conflict and operation obstacles for importing firms further depends on the age of the firm. Older firms are likely to have engaged in more transactions with Israeli suppliers before the conflict. Through these past interactions, they have revealed information on their reliability. Therefore, if the uncertainty related to conflict affects the terms of the contract between Palestinian firms and Israeli suppliers, we expect these changes to affect older firms disproportionally less. Consistent with this hypothesis, the evidence in column (6) of panel (c) shows that the negative relationship between conflict and the percentage of inputs paid before delivery by importing firms is systematically lower for older firms. Conflict, Misallocation and Output Value Reduced‐form Evidence Aggregate figures indicate the existence of a negative relationship between conflict intensity and output. The real GDP of the OPT falls by 20% between 2000 and 2002, mirroring the steep increase in the number of Palestinian fatalities over the period. By the same token, a downward trend in the number of fatalities in the period thereafter is associated with an increase in GDP, with the latter reaching its 2000 values in 2004. Establishment‐level data allow us to investigate this negative relationship between conflict intensity and output in a more systematic way. We implement the same specification as in (12) but replacing as outcome the (log) value of output produced by the firm. Table 5 shows the corresponding coefficient estimates. Column (1) shows the estimate for the coefficient of the fatalities variable from a simple regression specification in which it is the only included regressor. A one standard deviation increase in the number of fatalities in the district is associated with a 12.6% decrease in establishment's output value, significant at the 5% level. In columns (2)–(4), we progressively include year and district fixed effects, sector fixed effects, and firm‐level controls. In column (5), we include the full set of sector‐year fixed effects allowing for sector‐specific trends. We find that one standard deviation increase in the number of fatalities in the district is associated with an 8.6% drop in output value, significant at the 1% level. Table 5 Conflict and Output Value Log of output value, ln(PY) (1) (2) (3) (4) (5) (6) (7) fatalities −0.126** −0.073*** −0.063* −0.089*** −0.086*** −0.094*** −0.091*** (0.049) (0.024) (0.036) (0.033) (0.033) (0.035) (0.035) fatalities×zfMsfzdMsd −0.178** −0.436*** (0.072) (0.111) Family workersTotal −1.522*** −1.533*** −1.526*** −1.537*** (0.100) (0.097) (0.099) (0.096) ProprietorsTotal −2.713*** −2.717*** −2.700*** −2.703*** (0.112) (0.112) (0.115) (0.116) District FE N Y Y Y Y Y Y Year FE N Y Y Y n.a. Y n.a. Sector FE N N Y Y n.a. Y n.a. Sector × Year FE N N N N Y N Y Observations 10,042 10,042 10,042 10,039 10,039 10,006 10,006 R2 0.007 0.035 0.156 0.434 0.443 0.329 0.340 Log of output value, ln(PY) (1) (2) (3) (4) (5) (6) (7) fatalities −0.126** −0.073*** −0.063* −0.089*** −0.086*** −0.094*** −0.091*** (0.049) (0.024) (0.036) (0.033) (0.033) (0.035) (0.035) fatalities×zfMsfzdMsd −0.178** −0.436*** (0.072) (0.111) Family workersTotal −1.522*** −1.533*** −1.526*** −1.537*** (0.100) (0.097) (0.099) (0.096) ProprietorsTotal −2.713*** −2.717*** −2.700*** −2.703*** (0.112) (0.112) (0.115) (0.116) District FE N Y Y Y Y Y Y Year FE N Y Y Y n.a. Y n.a. Sector FE N N Y Y n.a. Y n.a. Sector × Year FE N N N N Y N Y Observations 10,042 10,042 10,042 10,039 10,039 10,006 10,006 R2 0.007 0.035 0.156 0.434 0.443 0.329 0.340 Notes *p‐value < 0.1; **p‐value < 0.05; ***p‐value < 0.01. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of output value in Israeli New Sheqel (NIS). The main independent variables are the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units), the sector‐level input value ratio between foreign and domestically produced materials as measured in 1999, zfMsf/zdMsd ⁠, and the interaction between the two. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table 5 Conflict and Output Value Log of output value, ln(PY) (1) (2) (3) (4) (5) (6) (7) fatalities −0.126** −0.073*** −0.063* −0.089*** −0.086*** −0.094*** −0.091*** (0.049) (0.024) (0.036) (0.033) (0.033) (0.035) (0.035) fatalities×zfMsfzdMsd −0.178** −0.436*** (0.072) (0.111) Family workersTotal −1.522*** −1.533*** −1.526*** −1.537*** (0.100) (0.097) (0.099) (0.096) ProprietorsTotal −2.713*** −2.717*** −2.700*** −2.703*** (0.112) (0.112) (0.115) (0.116) District FE N Y Y Y Y Y Y Year FE N Y Y Y n.a. Y n.a. Sector FE N N Y Y n.a. Y n.a. Sector × Year FE N N N N Y N Y Observations 10,042 10,042 10,042 10,039 10,039 10,006 10,006 R2 0.007 0.035 0.156 0.434 0.443 0.329 0.340 Log of output value, ln(PY) (1) (2) (3) (4) (5) (6) (7) fatalities −0.126** −0.073*** −0.063* −0.089*** −0.086*** −0.094*** −0.091*** (0.049) (0.024) (0.036) (0.033) (0.033) (0.035) (0.035) fatalities×zfMsfzdMsd −0.178** −0.436*** (0.072) (0.111) Family workersTotal −1.522*** −1.533*** −1.526*** −1.537*** (0.100) (0.097) (0.099) (0.096) ProprietorsTotal −2.713*** −2.717*** −2.700*** −2.703*** (0.112) (0.112) (0.115) (0.116) District FE N Y Y Y Y Y Y Year FE N Y Y Y n.a. Y n.a. Sector FE N N Y Y n.a. Y n.a. Sector × Year FE N N N N Y N Y Observations 10,042 10,042 10,042 10,039 10,039 10,006 10,006 R2 0.007 0.035 0.156 0.434 0.443 0.329 0.340 Notes *p‐value < 0.1; **p‐value < 0.05; ***p‐value < 0.01. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of output value in Israeli New Sheqel (NIS). The main independent variables are the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units), the sector‐level input value ratio between foreign and domestically produced materials as measured in 1999, zfMsf/zdMsd ⁠, and the interaction between the two. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large If the proposed mechanism of conflict‐induced distortions in input usage is partly responsible for the fall in output value, the latter should be more pronounced among firms operating in sectors that rely more on imported inputs. To explore this possibility, we calculate a sector‐level measure of intensity in imported inputs using firm‐level data from 1999, prior to the beginning of the Second Intifada. Consistently with our framework, we calculate the ratio between the total value of foreign vs. domestically produced materials used in production in each sector, and include it in our main regression specification, together with its interaction with the fatalities variable. Columns (6) and (7) of Table 5 show the corresponding coefficient estimates. The relationship between conflict and output value is systematically more negative for firms operating in sectors with higher baseline intensity in imported inputs. We interpret this as reduced‐form evidence that conflict‐induced distortions in foreign vs. domestically produced input usage is a relevant mechanism for the negative relationship between conflict and output value that we observe in the data. These results show that conflict intensity is negatively associated with firm output value. While our data do not allow us to look at establishment‐level output quantity and prices separately, a number of considerations lead us to conclude that these results indicate a negative relationship between conflict and output level. First, the conceptual framework suggests that this is the case. Equation (9) in Section 2 shows that, when firms enjoy a certain degree of market power, any increase in output or input distortion will result in higher firm‐level output prices. It follows that, if conflict increases distortions, only a more than proportional decrease in output quantity would generate the negative result we find on output value. Perhaps more importantly, there is no empirical evidence that conflict reduces output prices. To test for this, we identify those sectors which are most geographically localised, and check for whether the producer price index (PPI) of these sectors tracks the evolution of Palestinian fatalities in the same district. Evidence does not suggest the existence of a negative relationship between prices and conflict intensity over time in any of the sectors we investigate. Structural Estimates and Counterfactual Analysis Our goal in this subsection is to determine the extent to which increased distortions in the functioning and accessibility of markets for production inputs account for the observed fall in output value in high conflict districts. Once again, the proposed conceptual framework guides us in the empirical analysis. We start by recalling 10 for the output value of firm i operating in sector s. In the absence of any distortion, after taking logs, 10 reduces to: lnPsiYsi=lnσσ−1+ln1αsαs1βsβs11−αs−βs1−αs−βs+αslnRKsi+βslnwLsi+(1−αs−βs)lnzMsi. (15) We thus restrict our sample to those firm‐level observations belonging to the year 1999, prior to the beginning of the Second Intifada. Given the above equation, we structurally estimate the factor share parameters of the production function by implementing the following regression specification: lnPsiYsi=φs+αslnRKsi+βslnwLsi+(1−αs−βs)lnzMsi+uis, (16) where RKsi, wLsi and zMsi are the value used in production of capital, labour and materials respectively, and φs accounts for the sector‐specific intercept in the previous equation. As before, we separate foreign from domestically produced materials. We also estimate factor share parameters for each sector by interacting all regressors with 2‐digit sector dummies, and under the constraint that the sum of the corresponding coefficients is equal to one. We use these estimates to predict the value of output we would observe in the absence of conflict‐induced input substitution. To this end, we build upon the results in Table 1 from our most demanding regression specification where the dependent variable is the ratio between domestically and foreign produced materials used in production (row (d), column (4)). We set the level of conflict intensity to zero and predict the counterfactual input value ratio in the absence of conflict for each establishment in the period. Holding the total value of materials constant, we can then calculate the counterfactual value of domestically and foreign produced materials respectively. Using the estimated factor share parameters of the production function, we then calculate the value of output that we would have observed in the absence of conflict. Our calculations indicate that, in the absence of conflict, the value of output would have been 6.4% higher for the average firm in our sample. Such gains are lower than those suggested by the reduced‐form estimate in column (5) of Table 5. Taking the ratio of the two, we can infer that conflict‐induced distortions in the functioning and accessibility of markets for foreign inputs and the consequent changes in input usage can account for more than 70% of the drop in the output value of firms operating in high conflict areas. According to the weighted sums of establishments’ actual and counterfactual output value figures, aggregate output value could have been as much as 14.5% higher if conflict‐induced distortions in the accessibility of markets had not materialised. In the period 2000–6, the manufacturing sector accounts for 10–13% of Palestinian GDP. We therefore infer that, without taking into account activity in other industries and inter‐industry linkages, the Palestinian GDP could have been around 1.5% higher in the absence of conflict‐induced distortions. The proposed methodology builds upon the structural estimation of the factor share parameters of the production function and, therefore, relies on the validity of the estimation approach. This is determined by the extent to which the theoretical expression for output value in (15) matches our data. First, we do not need to assume that no market distortions were present in 1999 but rather we consider that as our benchmark. This means that we interpret the results from our counterfactual exercise in relation to the level of distortions already present in 1999, and focus on the distortions generated and/or exacerbated during conflict times. Second, the market structure assumption of monopolistic competition within sectors is relevant: if all firms were price takers, the derived estimates of the factor share parameters using 1999 data would be biased. This is because the choice of inputs would be endogenous to total factor productivity, which would be captured by the residual in the regression specification. Still, our exercise would remain valid as long as the extent of the bias is constant throughout the period under consideration. Third, our specification assumes a Cobb–Douglas production function. This restricts the magnitude of the elasticity of substitution between foreign and domestically produced materials, possibly leading to an overestimation of the impact of material input substitution on output value. In subsection A.2 of Appendix A, we structurally estimate the elasticity of substitution between foreign and domestically produced materials in each sector and find it remarkably close to one. Evidence is therefore supportive of our approach. Finally, our procedure implicitly assumes that a negative shock to demand would decrease the value of all inputs used in production, but not their relative proportions. The evidence presented when questioning the homotheticity assumption in subsection 5.2.2 is supportive of this hypothesis. Conclusions Understanding the microeconomic mechanisms behind the relationship between conflict and aggregate economic outcomes is crucial for the design and implementation of successful conflict recovery policies. In this article, we have provided direct evidence on one of such mechanisms. Using firm‐level data from the OPT during the Second Intifada, we have shown that conflict disrupts the functioning and accessibility of markets for foreign inputs, leading firms to substitute locally produced materials for imported ones in production. Evidence suggests that this is linked to a worsening of the bargaining position of Palestinian firms in their relationship with foreign suppliers. Our counterfactual policy analysis shows that conflict‐induced distortions in the accessibility of foreign market can account for more than 70% of the drop in output value of firms in the OPT during the Second Intifada. While the relative importance of different types of input distortions and the magnitude of the estimated impact on output value may be context‐dependent, our methodology to infer the relative extent of conflict‐induced distortions has general validity. We have shown that the analysis of input usage can provide direct micro‐founded evidence of the negative impact of conflict on aggregate economic outcomes. Our results indicate that conflict recovery policies that target the supply side of the economy and restore the functioning of markets for inputs can be particularly effective. As conflict introduces distortions in the accessibility of foreign markets for inputs, policies aimed to restore trade and its financing are the most suited to mitigate the negative impact of warfare. For example, if conflict‐related uncertainty is the source of the loss in bargaining power that firms experience in their contractual relationship with foreign suppliers, a policy that insures the latter against insolvency risk would target this friction directly. What other actual forms policy intervention should take and how to evaluate its impact in a rigorous way are questions that we leave for future research. Appendix A. Additional Results Table A1. Summary Statistics Observations Mean SD Minimum Maximum Log of output value 11,397 11.741 1.511 0 19.656 Log of value of capital 14,221 10.138 1.942 0.693 18.531 Log of value of labour 10,243 10.492 1.24 5.994 16.746 Log of value of materials 14,160 11.308 2.045 3.932 18.769 Log of value of domestic materials 14,160 8.826 3.138 0 18.785 Log of value of imported materials 14,160 6.456 4.801 0 18.688 Fraction of family workers 14,284 0.167 0.247 0 1 Fraction of proprietors 14,284 0.444 0.324 0 1 Log of value of capital/materials 14,100 −0.553 1.816 −13.169 6.828 Log of value of labour/materials 10,183 −0.856 1.361 −8.593 4.185 Log of value of capital/labour 10,197 0.223 1.67 −10.786 6.161 Log of value of domestic/imported materials 14,160 2.37 6.345 −18.405 18.112 Log of value of capital/imported materials 14,100 3.687 4.645 −12.855 17.751 Log of value of capital/domestic materials 14,100 1.322 3.198 −13.155 17.231 Log of value of labour/imported materials 10,183 3.117 4.69 −6.367 16.544 Log of value of labour/domestic materials 10,183 1.046 2.96 −8.699 15.451 Log of value of initial capital 14,222 10.606 2.601 0 18.550 Log of value of initial inventory of output 11,397 4.143 4.945 0 16.317 Log of value of initial inventory of materials 14,161 6.317 4.327 0 16.710 Palestinians killed by IDF (District × Year) 112 35.044 42.010 0 210 Observations Mean SD Minimum Maximum Log of output value 11,397 11.741 1.511 0 19.656 Log of value of capital 14,221 10.138 1.942 0.693 18.531 Log of value of labour 10,243 10.492 1.24 5.994 16.746 Log of value of materials 14,160 11.308 2.045 3.932 18.769 Log of value of domestic materials 14,160 8.826 3.138 0 18.785 Log of value of imported materials 14,160 6.456 4.801 0 18.688 Fraction of family workers 14,284 0.167 0.247 0 1 Fraction of proprietors 14,284 0.444 0.324 0 1 Log of value of capital/materials 14,100 −0.553 1.816 −13.169 6.828 Log of value of labour/materials 10,183 −0.856 1.361 −8.593 4.185 Log of value of capital/labour 10,197 0.223 1.67 −10.786 6.161 Log of value of domestic/imported materials 14,160 2.37 6.345 −18.405 18.112 Log of value of capital/imported materials 14,100 3.687 4.645 −12.855 17.751 Log of value of capital/domestic materials 14,100 1.322 3.198 −13.155 17.231 Log of value of labour/imported materials 10,183 3.117 4.69 −6.367 16.544 Log of value of labour/domestic materials 10,183 1.046 2.96 −8.699 15.451 Log of value of initial capital 14,222 10.606 2.601 0 18.550 Log of value of initial inventory of output 11,397 4.143 4.945 0 16.317 Log of value of initial inventory of materials 14,161 6.317 4.327 0 16.710 Palestinians killed by IDF (District × Year) 112 35.044 42.010 0 210 Notes The Table shows summary statistics for the variables used in the empirical analysis. Establishment‐level value variables are in Israeli New Sheqel (NIS). Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A1. Summary Statistics Observations Mean SD Minimum Maximum Log of output value 11,397 11.741 1.511 0 19.656 Log of value of capital 14,221 10.138 1.942 0.693 18.531 Log of value of labour 10,243 10.492 1.24 5.994 16.746 Log of value of materials 14,160 11.308 2.045 3.932 18.769 Log of value of domestic materials 14,160 8.826 3.138 0 18.785 Log of value of imported materials 14,160 6.456 4.801 0 18.688 Fraction of family workers 14,284 0.167 0.247 0 1 Fraction of proprietors 14,284 0.444 0.324 0 1 Log of value of capital/materials 14,100 −0.553 1.816 −13.169 6.828 Log of value of labour/materials 10,183 −0.856 1.361 −8.593 4.185 Log of value of capital/labour 10,197 0.223 1.67 −10.786 6.161 Log of value of domestic/imported materials 14,160 2.37 6.345 −18.405 18.112 Log of value of capital/imported materials 14,100 3.687 4.645 −12.855 17.751 Log of value of capital/domestic materials 14,100 1.322 3.198 −13.155 17.231 Log of value of labour/imported materials 10,183 3.117 4.69 −6.367 16.544 Log of value of labour/domestic materials 10,183 1.046 2.96 −8.699 15.451 Log of value of initial capital 14,222 10.606 2.601 0 18.550 Log of value of initial inventory of output 11,397 4.143 4.945 0 16.317 Log of value of initial inventory of materials 14,161 6.317 4.327 0 16.710 Palestinians killed by IDF (District × Year) 112 35.044 42.010 0 210 Observations Mean SD Minimum Maximum Log of output value 11,397 11.741 1.511 0 19.656 Log of value of capital 14,221 10.138 1.942 0.693 18.531 Log of value of labour 10,243 10.492 1.24 5.994 16.746 Log of value of materials 14,160 11.308 2.045 3.932 18.769 Log of value of domestic materials 14,160 8.826 3.138 0 18.785 Log of value of imported materials 14,160 6.456 4.801 0 18.688 Fraction of family workers 14,284 0.167 0.247 0 1 Fraction of proprietors 14,284 0.444 0.324 0 1 Log of value of capital/materials 14,100 −0.553 1.816 −13.169 6.828 Log of value of labour/materials 10,183 −0.856 1.361 −8.593 4.185 Log of value of capital/labour 10,197 0.223 1.67 −10.786 6.161 Log of value of domestic/imported materials 14,160 2.37 6.345 −18.405 18.112 Log of value of capital/imported materials 14,100 3.687 4.645 −12.855 17.751 Log of value of capital/domestic materials 14,100 1.322 3.198 −13.155 17.231 Log of value of labour/imported materials 10,183 3.117 4.69 −6.367 16.544 Log of value of labour/domestic materials 10,183 1.046 2.96 −8.699 15.451 Log of value of initial capital 14,222 10.606 2.601 0 18.550 Log of value of initial inventory of output 11,397 4.143 4.945 0 16.317 Log of value of initial inventory of materials 14,161 6.317 4.327 0 16.710 Palestinians killed by IDF (District × Year) 112 35.044 42.010 0 210 Notes The Table shows summary statistics for the variables used in the empirical analysis. Establishment‐level value variables are in Israeli New Sheqel (NIS). Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A2. Input Distortions – Implied Relative Values Implied relative distortion values (1) (2) (3) (4) (a) (1 + τM)/(1 + τK) 1.005 1.003 1.008 1.006 (0.919; 1.090) (0.918; 1.088) (0.920; 1.095) (0.916; 1.096) (b) (1 + τM)/(1 + τL) 1.025 1.022 1.024 1.010 (0.948; 1.103) (0.947; 1.097) (0.950; 1.098) (0.931; 1.089) (c) (1 + τL)/(1 + τK) 0.982 0.986 0.985 1.000 (0.905; 1.059) (0.911; 1.061) (0.910; 1.060) (0.919; 1.080) (d) (1+τMf)/(1+τMd) 3.375 3.373 3.434 3.465 (1.578; 5.172) (1.579; 5.168) (1.616; 5.252) (1.634; 5.295) (e) (1+τMf)/(1+τK) 1.687 1.684 1.713 1.736 (1.283; 2.090) (1.283; 2.084) (1.314; 2.112) (1.302; 2.169) (f) (1+τMf)/(1+τL) 1.602 1.596 1.593 1.623 (1.168; 2.036) (1.157; 2.034) (1.156; 2.030) (1.147; 2.099) (g) (1+τMd)/(1+τK) 0.501 0.500 0.501 0.502 (0.334; 0.669) (0.333; 0.668) (0.333; 0.668) (0.340; 0.663) (h) (1+τMd)/(1+τL) 0.513 0.511 0.513 0.516 (0.328; 0.698) (0.329; 0.693) (0.330; 0.696) (0.332; 0.700) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Implied relative distortion values (1) (2) (3) (4) (a) (1 + τM)/(1 + τK) 1.005 1.003 1.008 1.006 (0.919; 1.090) (0.918; 1.088) (0.920; 1.095) (0.916; 1.096) (b) (1 + τM)/(1 + τL) 1.025 1.022 1.024 1.010 (0.948; 1.103) (0.947; 1.097) (0.950; 1.098) (0.931; 1.089) (c) (1 + τL)/(1 + τK) 0.982 0.986 0.985 1.000 (0.905; 1.059) (0.911; 1.061) (0.910; 1.060) (0.919; 1.080) (d) (1+τMf)/(1+τMd) 3.375 3.373 3.434 3.465 (1.578; 5.172) (1.579; 5.168) (1.616; 5.252) (1.634; 5.295) (e) (1+τMf)/(1+τK) 1.687 1.684 1.713 1.736 (1.283; 2.090) (1.283; 2.084) (1.314; 2.112) (1.302; 2.169) (f) (1+τMf)/(1+τL) 1.602 1.596 1.593 1.623 (1.168; 2.036) (1.157; 2.034) (1.156; 2.030) (1.147; 2.099) (g) (1+τMd)/(1+τK) 0.501 0.500 0.501 0.502 (0.334; 0.669) (0.333; 0.668) (0.333; 0.668) (0.340; 0.663) (h) (1+τMd)/(1+τL) 0.513 0.511 0.513 0.516 (0.328; 0.698) (0.329; 0.693) (0.330; 0.696) (0.332; 0.700) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes The Table reports implied relative distortion values as derived using coefficient estimates from Table 4, together with 95% confidence intervals. Standard errors are clustered along both sector‐year and district‐year categories. τK is average distortion level for capital; τM is average distortion level for materials; τL is average distortion value for labour; τMf is average distortion value for imported materials; τMd is average distortion value for domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A2. Input Distortions – Implied Relative Values Implied relative distortion values (1) (2) (3) (4) (a) (1 + τM)/(1 + τK) 1.005 1.003 1.008 1.006 (0.919; 1.090) (0.918; 1.088) (0.920; 1.095) (0.916; 1.096) (b) (1 + τM)/(1 + τL) 1.025 1.022 1.024 1.010 (0.948; 1.103) (0.947; 1.097) (0.950; 1.098) (0.931; 1.089) (c) (1 + τL)/(1 + τK) 0.982 0.986 0.985 1.000 (0.905; 1.059) (0.911; 1.061) (0.910; 1.060) (0.919; 1.080) (d) (1+τMf)/(1+τMd) 3.375 3.373 3.434 3.465 (1.578; 5.172) (1.579; 5.168) (1.616; 5.252) (1.634; 5.295) (e) (1+τMf)/(1+τK) 1.687 1.684 1.713 1.736 (1.283; 2.090) (1.283; 2.084) (1.314; 2.112) (1.302; 2.169) (f) (1+τMf)/(1+τL) 1.602 1.596 1.593 1.623 (1.168; 2.036) (1.157; 2.034) (1.156; 2.030) (1.147; 2.099) (g) (1+τMd)/(1+τK) 0.501 0.500 0.501 0.502 (0.334; 0.669) (0.333; 0.668) (0.333; 0.668) (0.340; 0.663) (h) (1+τMd)/(1+τL) 0.513 0.511 0.513 0.516 (0.328; 0.698) (0.329; 0.693) (0.330; 0.696) (0.332; 0.700) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Implied relative distortion values (1) (2) (3) (4) (a) (1 + τM)/(1 + τK) 1.005 1.003 1.008 1.006 (0.919; 1.090) (0.918; 1.088) (0.920; 1.095) (0.916; 1.096) (b) (1 + τM)/(1 + τL) 1.025 1.022 1.024 1.010 (0.948; 1.103) (0.947; 1.097) (0.950; 1.098) (0.931; 1.089) (c) (1 + τL)/(1 + τK) 0.982 0.986 0.985 1.000 (0.905; 1.059) (0.911; 1.061) (0.910; 1.060) (0.919; 1.080) (d) (1+τMf)/(1+τMd) 3.375 3.373 3.434 3.465 (1.578; 5.172) (1.579; 5.168) (1.616; 5.252) (1.634; 5.295) (e) (1+τMf)/(1+τK) 1.687 1.684 1.713 1.736 (1.283; 2.090) (1.283; 2.084) (1.314; 2.112) (1.302; 2.169) (f) (1+τMf)/(1+τL) 1.602 1.596 1.593 1.623 (1.168; 2.036) (1.157; 2.034) (1.156; 2.030) (1.147; 2.099) (g) (1+τMd)/(1+τK) 0.501 0.500 0.501 0.502 (0.334; 0.669) (0.333; 0.668) (0.333; 0.668) (0.340; 0.663) (h) (1+τMd)/(1+τL) 0.513 0.511 0.513 0.516 (0.328; 0.698) (0.329; 0.693) (0.330; 0.696) (0.332; 0.700) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes The Table reports implied relative distortion values as derived using coefficient estimates from Table 4, together with 95% confidence intervals. Standard errors are clustered along both sector‐year and district‐year categories. τK is average distortion level for capital; τM is average distortion level for materials; τL is average distortion value for labour; τMf is average distortion value for imported materials; τMd is average distortion value for domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A3. Input Distortions – Regression Coefficients Robustness: Wild Bootstrap‐clustered Standard Errors Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.005 0.003 0.008 0.006 (0.90) (0.90) (0.90) (0.92) (b) ln wLsi/zMsi 0.025 0.022 0.024 0.010 (0.54) (0.62) (0.58) (0.80) (c) ln RKsi/wLsi −0.018 −0.014 −0.015 −0.000 (0.80) (0.88) (0.84) (1.00) (d) lnzdMsid/zfMsif 1.216*** 1.216*** 1.234*** 1.243*** (0.00) (0.00) (0.00) (0.00) (e) lnRKsi/zfMsif 0.523*** 0.521*** 0.538*** 0.551*** (0.00) (0.00) (0.00) (0.00) (f) lnwLsi/zfMsif 0.471*** 0.467*** 0.466*** 0.484*** (0.00) (0.00) (0.00) (0.00) (g) lnRKsi/zdMsid −0.690** −0.692** −0.692** −0.690** (0.02) (0.02) (0.02) (0.02) (h) lnwLsi/zdMsid −0.668** −0.671** −0.668** −0.662** (0.02) (0.02) (0.02) (0.02) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.005 0.003 0.008 0.006 (0.90) (0.90) (0.90) (0.92) (b) ln wLsi/zMsi 0.025 0.022 0.024 0.010 (0.54) (0.62) (0.58) (0.80) (c) ln RKsi/wLsi −0.018 −0.014 −0.015 −0.000 (0.80) (0.88) (0.84) (1.00) (d) lnzdMsid/zfMsif 1.216*** 1.216*** 1.234*** 1.243*** (0.00) (0.00) (0.00) (0.00) (e) lnRKsi/zfMsif 0.523*** 0.521*** 0.538*** 0.551*** (0.00) (0.00) (0.00) (0.00) (f) lnwLsi/zfMsif 0.471*** 0.467*** 0.466*** 0.484*** (0.00) (0.00) (0.00) (0.00) (g) lnRKsi/zdMsid −0.690** −0.692** −0.692** −0.690** (0.02) (0.02) (0.02) (0.02) (h) lnwLsi/zdMsid −0.668** −0.671** −0.668** −0.662** (0.02) (0.02) (0.02) (0.02) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes Wild bootstrap p‐values in parenthesis. The Table reports estimates of the coefficient of the fatalities variable. Standard errors are clustered at the district level, and calculating using wild bootstrapping after 100 repetitions (Cameron et al., 2008). Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A3. Input Distortions – Regression Coefficients Robustness: Wild Bootstrap‐clustered Standard Errors Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.005 0.003 0.008 0.006 (0.90) (0.90) (0.90) (0.92) (b) ln wLsi/zMsi 0.025 0.022 0.024 0.010 (0.54) (0.62) (0.58) (0.80) (c) ln RKsi/wLsi −0.018 −0.014 −0.015 −0.000 (0.80) (0.88) (0.84) (1.00) (d) lnzdMsid/zfMsif 1.216*** 1.216*** 1.234*** 1.243*** (0.00) (0.00) (0.00) (0.00) (e) lnRKsi/zfMsif 0.523*** 0.521*** 0.538*** 0.551*** (0.00) (0.00) (0.00) (0.00) (f) lnwLsi/zfMsif 0.471*** 0.467*** 0.466*** 0.484*** (0.00) (0.00) (0.00) (0.00) (g) lnRKsi/zdMsid −0.690** −0.692** −0.692** −0.690** (0.02) (0.02) (0.02) (0.02) (h) lnwLsi/zdMsid −0.668** −0.671** −0.668** −0.662** (0.02) (0.02) (0.02) (0.02) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.005 0.003 0.008 0.006 (0.90) (0.90) (0.90) (0.92) (b) ln wLsi/zMsi 0.025 0.022 0.024 0.010 (0.54) (0.62) (0.58) (0.80) (c) ln RKsi/wLsi −0.018 −0.014 −0.015 −0.000 (0.80) (0.88) (0.84) (1.00) (d) lnzdMsid/zfMsif 1.216*** 1.216*** 1.234*** 1.243*** (0.00) (0.00) (0.00) (0.00) (e) lnRKsi/zfMsif 0.523*** 0.521*** 0.538*** 0.551*** (0.00) (0.00) (0.00) (0.00) (f) lnwLsi/zfMsif 0.471*** 0.467*** 0.466*** 0.484*** (0.00) (0.00) (0.00) (0.00) (g) lnRKsi/zdMsid −0.690** −0.692** −0.692** −0.690** (0.02) (0.02) (0.02) (0.02) (h) lnwLsi/zdMsid −0.668** −0.671** −0.668** −0.662** (0.02) (0.02) (0.02) (0.02) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes Wild bootstrap p‐values in parenthesis. The Table reports estimates of the coefficient of the fatalities variable. Standard errors are clustered at the district level, and calculating using wild bootstrapping after 100 repetitions (Cameron et al., 2008). Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A4. Input Distortions – Regression Coefficients Robustness: Actual Number of Fatalities as Regressor Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.000 0.000 0.000 0.000 (0.001) (0.001) (0.001) (0.001) (b) ln wLsi/zMsi 0.001 0.001 0.001 0.000 (0.001) (0.001) (0.001) (0.001) (c) ln RKsi/wLsi −0.000 −0.000 −0.000 −0.000 (0.001) (0.001) (0.001) (0.001) (d) lnzdMsid/zfMsif 0.029*** 0.029*** 0.029*** 0.030*** (0.006) (0.006) (0.006) (0.006) (e) lnRKsi/zfMsif 0.012*** 0.012*** 0.013*** 0.013*** (0.003) (0.003) (0.003) (0.003) (f) lnwLsi/zfMsif 0.011*** 0.011*** 0.011*** 0.012*** (0.003) (0.003) (0.003) (0.004) (g) lnRKsi/zdMsid −0.016*** −0.016*** −0.016*** −0.016*** (0.004) (0.004) (0.004) (0.004) (h) lnwLsi/zdMsid −0.016*** −0.016*** −0.016*** −0.016*** (0.004) (0.004) (0.004) (0.004) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.000 0.000 0.000 0.000 (0.001) (0.001) (0.001) (0.001) (b) ln wLsi/zMsi 0.001 0.001 0.001 0.000 (0.001) (0.001) (0.001) (0.001) (c) ln RKsi/wLsi −0.000 −0.000 −0.000 −0.000 (0.001) (0.001) (0.001) (0.001) (d) lnzdMsid/zfMsif 0.029*** 0.029*** 0.029*** 0.030*** (0.006) (0.006) (0.006) (0.006) (e) lnRKsi/zfMsif 0.012*** 0.012*** 0.013*** 0.013*** (0.003) (0.003) (0.003) (0.003) (f) lnwLsi/zfMsif 0.011*** 0.011*** 0.011*** 0.012*** (0.003) (0.003) (0.003) (0.004) (g) lnRKsi/zdMsid −0.016*** −0.016*** −0.016*** −0.016*** (0.004) (0.004) (0.004) (0.004) (h) lnwLsi/zdMsid −0.016*** −0.016*** −0.016*** −0.016*** (0.004) (0.004) (0.004) (0.004) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities variable, where the latter capture the actual number of fatalities (without normalising by its standard deviation). Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A4. Input Distortions – Regression Coefficients Robustness: Actual Number of Fatalities as Regressor Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.000 0.000 0.000 0.000 (0.001) (0.001) (0.001) (0.001) (b) ln wLsi/zMsi 0.001 0.001 0.001 0.000 (0.001) (0.001) (0.001) (0.001) (c) ln RKsi/wLsi −0.000 −0.000 −0.000 −0.000 (0.001) (0.001) (0.001) (0.001) (d) lnzdMsid/zfMsif 0.029*** 0.029*** 0.029*** 0.030*** (0.006) (0.006) (0.006) (0.006) (e) lnRKsi/zfMsif 0.012*** 0.012*** 0.013*** 0.013*** (0.003) (0.003) (0.003) (0.003) (f) lnwLsi/zfMsif 0.011*** 0.011*** 0.011*** 0.012*** (0.003) (0.003) (0.003) (0.004) (g) lnRKsi/zdMsid −0.016*** −0.016*** −0.016*** −0.016*** (0.004) (0.004) (0.004) (0.004) (h) lnwLsi/zdMsid −0.016*** −0.016*** −0.016*** −0.016*** (0.004) (0.004) (0.004) (0.004) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.000 0.000 0.000 0.000 (0.001) (0.001) (0.001) (0.001) (b) ln wLsi/zMsi 0.001 0.001 0.001 0.000 (0.001) (0.001) (0.001) (0.001) (c) ln RKsi/wLsi −0.000 −0.000 −0.000 −0.000 (0.001) (0.001) (0.001) (0.001) (d) lnzdMsid/zfMsif 0.029*** 0.029*** 0.029*** 0.030*** (0.006) (0.006) (0.006) (0.006) (e) lnRKsi/zfMsif 0.012*** 0.012*** 0.013*** 0.013*** (0.003) (0.003) (0.003) (0.003) (f) lnwLsi/zfMsif 0.011*** 0.011*** 0.011*** 0.012*** (0.003) (0.003) (0.003) (0.004) (g) lnRKsi/zdMsid −0.016*** −0.016*** −0.016*** −0.016*** (0.004) (0.004) (0.004) (0.004) (h) lnwLsi/zdMsid −0.016*** −0.016*** −0.016*** −0.016*** (0.004) (0.004) (0.004) (0.004) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities variable, where the latter capture the actual number of fatalities (without normalising by its standard deviation). Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A5. Input Distortions – Regression Coefficients Robustness: Per Capita Measure of Conflict Coefficient of fatalities/population variable (1) (2) (3) (4) (a) ln RKsi/zMsi −45.525 −63.317 −35.049 −31.554 (263.724) (264.143) (269.527) (267.587) (b) ln wLsi/zMsi 88.810 49.512 43.214 −23.117 (251.826) (251.367) (254.335) (262.242) (c) ln RKsi/wLsi −104.539 −47.727 −38.363 47.251 (266.669) (260.723) (262.884) (268.779) (d) lnzdMsid/zfMsif 5,586.620*** 5,582.498*** 5,700.515*** 5,653.444*** (1,913.741) (1,912.605) (1,908.057) (1,893.607) (e) lnRKsi/zfMsif 2,476.361*** 2,457.017*** 2,574.175*** 2,569.458*** (912.206) (908.337) (902.523) (935.686) (f) lnwLsi/zfMsif 2,117.569** 2,065.229** 2,070.722** 2,073.873** (993.970) (1,005.264) (1,003.561) (1,037.184) (g) lnRKsi/zdMsid −3,089.911*** −3,106.254*** −3,106.256*** −3,076.875*** (1,125.136) (1,126.431) (1,125.300) (1,082.118) (h) lnwLsi/zdMsid −2,959.658** −3,000.269** −3,010.472** −2,912.969** (1,218.936) (1,210.869) (1,216.564) (1,204.352) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities/population variable (1) (2) (3) (4) (a) ln RKsi/zMsi −45.525 −63.317 −35.049 −31.554 (263.724) (264.143) (269.527) (267.587) (b) ln wLsi/zMsi 88.810 49.512 43.214 −23.117 (251.826) (251.367) (254.335) (262.242) (c) ln RKsi/wLsi −104.539 −47.727 −38.363 47.251 (266.669) (260.723) (262.884) (268.779) (d) lnzdMsid/zfMsif 5,586.620*** 5,582.498*** 5,700.515*** 5,653.444*** (1,913.741) (1,912.605) (1,908.057) (1,893.607) (e) lnRKsi/zfMsif 2,476.361*** 2,457.017*** 2,574.175*** 2,569.458*** (912.206) (908.337) (902.523) (935.686) (f) lnwLsi/zfMsif 2,117.569** 2,065.229** 2,070.722** 2,073.873** (993.970) (1,005.264) (1,003.561) (1,037.184) (g) lnRKsi/zdMsid −3,089.911*** −3,106.254*** −3,106.256*** −3,076.875*** (1,125.136) (1,126.431) (1,125.300) (1,082.118) (h) lnwLsi/zdMsid −2,959.658** −3,000.269** −3,010.472** −2,912.969** (1,218.936) (1,210.869) (1,216.564) (1,204.352) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities/population variable, calculated by dividing the number of Palestinian fatalities by the population of each district in the corresponding year. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A5. Input Distortions – Regression Coefficients Robustness: Per Capita Measure of Conflict Coefficient of fatalities/population variable (1) (2) (3) (4) (a) ln RKsi/zMsi −45.525 −63.317 −35.049 −31.554 (263.724) (264.143) (269.527) (267.587) (b) ln wLsi/zMsi 88.810 49.512 43.214 −23.117 (251.826) (251.367) (254.335) (262.242) (c) ln RKsi/wLsi −104.539 −47.727 −38.363 47.251 (266.669) (260.723) (262.884) (268.779) (d) lnzdMsid/zfMsif 5,586.620*** 5,582.498*** 5,700.515*** 5,653.444*** (1,913.741) (1,912.605) (1,908.057) (1,893.607) (e) lnRKsi/zfMsif 2,476.361*** 2,457.017*** 2,574.175*** 2,569.458*** (912.206) (908.337) (902.523) (935.686) (f) lnwLsi/zfMsif 2,117.569** 2,065.229** 2,070.722** 2,073.873** (993.970) (1,005.264) (1,003.561) (1,037.184) (g) lnRKsi/zdMsid −3,089.911*** −3,106.254*** −3,106.256*** −3,076.875*** (1,125.136) (1,126.431) (1,125.300) (1,082.118) (h) lnwLsi/zdMsid −2,959.658** −3,000.269** −3,010.472** −2,912.969** (1,218.936) (1,210.869) (1,216.564) (1,204.352) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities/population variable (1) (2) (3) (4) (a) ln RKsi/zMsi −45.525 −63.317 −35.049 −31.554 (263.724) (264.143) (269.527) (267.587) (b) ln wLsi/zMsi 88.810 49.512 43.214 −23.117 (251.826) (251.367) (254.335) (262.242) (c) ln RKsi/wLsi −104.539 −47.727 −38.363 47.251 (266.669) (260.723) (262.884) (268.779) (d) lnzdMsid/zfMsif 5,586.620*** 5,582.498*** 5,700.515*** 5,653.444*** (1,913.741) (1,912.605) (1,908.057) (1,893.607) (e) lnRKsi/zfMsif 2,476.361*** 2,457.017*** 2,574.175*** 2,569.458*** (912.206) (908.337) (902.523) (935.686) (f) lnwLsi/zfMsif 2,117.569** 2,065.229** 2,070.722** 2,073.873** (993.970) (1,005.264) (1,003.561) (1,037.184) (g) lnRKsi/zdMsid −3,089.911*** −3,106.254*** −3,106.256*** −3,076.875*** (1,125.136) (1,126.431) (1,125.300) (1,082.118) (h) lnwLsi/zdMsid −2,959.658** −3,000.269** −3,010.472** −2,912.969** (1,218.936) (1,210.869) (1,216.564) (1,204.352) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities/population variable, calculated by dividing the number of Palestinian fatalities by the population of each district in the corresponding year. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A6. Input Distortions – Regression Coefficients Robustness: ICEWS Conflict Data Coefficient of hostile ICEWS events count (1) (2) (3) (4) (a) ln RKsi/zMsi −0.008 −0.006 −0.004 −0.002 (0.015) (0.015) (0.016) (0.018) (b) ln wLsi/zMsi 0.029 0.026 0.027 0.014 (0.025) (0.025) (0.024) (0.026) (c) ln RKsi/wLsi −0.033 −0.029 −0.029 −0.010 (0.022) (0.021) (0.020) (0.030) (d) lnzdMsid/zfMsif 0.533* 0.535* 0.545* 0.560* (0.322) (0.321) (0.320) (0.319) (e) lnRKsi/zfMsif 0.224 0.227 0.237 0.262* (0.148) (0.146) (0.144) (0.152) (f) lnwLsi/zfMsif 0.265** 0.262** 0.261** 0.275** (0.129) (0.130) (0.130) (0.137) (g) lnRKsi/zdMsid −0.306* −0.305* −0.305* −0.296* (0.182) (0.183) (0.183) (0.179) (h) lnwLsi/zdMsid −0.256 −0.259 −0.258 −0.255 (0.177) (0.175) (0.175) (0.173) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of hostile ICEWS events count (1) (2) (3) (4) (a) ln RKsi/zMsi −0.008 −0.006 −0.004 −0.002 (0.015) (0.015) (0.016) (0.018) (b) ln wLsi/zMsi 0.029 0.026 0.027 0.014 (0.025) (0.025) (0.024) (0.026) (c) ln RKsi/wLsi −0.033 −0.029 −0.029 −0.010 (0.022) (0.021) (0.020) (0.030) (d) lnzdMsid/zfMsif 0.533* 0.535* 0.545* 0.560* (0.322) (0.321) (0.320) (0.319) (e) lnRKsi/zfMsif 0.224 0.227 0.237 0.262* (0.148) (0.146) (0.144) (0.152) (f) lnwLsi/zfMsif 0.265** 0.262** 0.261** 0.275** (0.129) (0.130) (0.130) (0.137) (g) lnRKsi/zdMsid −0.306* −0.305* −0.305* −0.296* (0.182) (0.183) (0.183) (0.179) (h) lnwLsi/zdMsid −0.256 −0.259 −0.258 −0.255 (0.177) (0.175) (0.175) (0.173) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; **p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the number of hostile ICEWS events in the district, normalised by its sample standard deviation. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, ICEWS. View Large Table A6. Input Distortions – Regression Coefficients Robustness: ICEWS Conflict Data Coefficient of hostile ICEWS events count (1) (2) (3) (4) (a) ln RKsi/zMsi −0.008 −0.006 −0.004 −0.002 (0.015) (0.015) (0.016) (0.018) (b) ln wLsi/zMsi 0.029 0.026 0.027 0.014 (0.025) (0.025) (0.024) (0.026) (c) ln RKsi/wLsi −0.033 −0.029 −0.029 −0.010 (0.022) (0.021) (0.020) (0.030) (d) lnzdMsid/zfMsif 0.533* 0.535* 0.545* 0.560* (0.322) (0.321) (0.320) (0.319) (e) lnRKsi/zfMsif 0.224 0.227 0.237 0.262* (0.148) (0.146) (0.144) (0.152) (f) lnwLsi/zfMsif 0.265** 0.262** 0.261** 0.275** (0.129) (0.130) (0.130) (0.137) (g) lnRKsi/zdMsid −0.306* −0.305* −0.305* −0.296* (0.182) (0.183) (0.183) (0.179) (h) lnwLsi/zdMsid −0.256 −0.259 −0.258 −0.255 (0.177) (0.175) (0.175) (0.173) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of hostile ICEWS events count (1) (2) (3) (4) (a) ln RKsi/zMsi −0.008 −0.006 −0.004 −0.002 (0.015) (0.015) (0.016) (0.018) (b) ln wLsi/zMsi 0.029 0.026 0.027 0.014 (0.025) (0.025) (0.024) (0.026) (c) ln RKsi/wLsi −0.033 −0.029 −0.029 −0.010 (0.022) (0.021) (0.020) (0.030) (d) lnzdMsid/zfMsif 0.533* 0.535* 0.545* 0.560* (0.322) (0.321) (0.320) (0.319) (e) lnRKsi/zfMsif 0.224 0.227 0.237 0.262* (0.148) (0.146) (0.144) (0.152) (f) lnwLsi/zfMsif 0.265** 0.262** 0.261** 0.275** (0.129) (0.130) (0.130) (0.137) (g) lnRKsi/zdMsid −0.306* −0.305* −0.305* −0.296* (0.182) (0.183) (0.183) (0.179) (h) lnwLsi/zdMsid −0.256 −0.259 −0.258 −0.255 (0.177) (0.175) (0.175) (0.173) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; **p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the number of hostile ICEWS events in the district, normalised by its sample standard deviation. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, ICEWS. View Large Table A7. Input Distortions – Regression Coefficients Robustness: ICEWS Conflict Data – Per Capita Measure Coefficient of hostile ICEWS events/population (1) (2) (3) (4) (a) ln RKsi/zMsi −2.491 −0.445 1.816 0.581 (19.792) (20.311) (20.481) (22.404) (b) ln wLsi/zMsi 24.301 20.076 19.989 7.920 (25.335) (25.018) (24.126) (26.054) (c) ln RKsi/wLsi −30.349 −24.473 −24.968 −8.536 (21.935) (21.061) (20.316) (28.817) (d) lnzdMsid/zfMsif 500.402* 502.441* 512.438* 530.730* (292.202) (290.696) (291.054) (290.729) (e) lnRKsi/zfMsif 219.874 223.131* 232.890* 253.719* (134.968) (133.181) (132.817) (139.994) (f) lnwLsi/zfMsif 249.515** 244.651** 244.727** 262.666** (117.401) (118.857) (118.962) (126.438) (g) lnRKsi/zdMsid −279.165* −277.912* −277.927* −276.457* (165.706) (166.167) (166.227) (162.546) (h) lnwLsi/zdMsid −242.422 −246.943 −247.083 −247.963 (158.871) (157.313) (157.112) (155.468) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of hostile ICEWS events/population (1) (2) (3) (4) (a) ln RKsi/zMsi −2.491 −0.445 1.816 0.581 (19.792) (20.311) (20.481) (22.404) (b) ln wLsi/zMsi 24.301 20.076 19.989 7.920 (25.335) (25.018) (24.126) (26.054) (c) ln RKsi/wLsi −30.349 −24.473 −24.968 −8.536 (21.935) (21.061) (20.316) (28.817) (d) lnzdMsid/zfMsif 500.402* 502.441* 512.438* 530.730* (292.202) (290.696) (291.054) (290.729) (e) lnRKsi/zfMsif 219.874 223.131* 232.890* 253.719* (134.968) (133.181) (132.817) (139.994) (f) lnwLsi/zfMsif 249.515** 244.651** 244.727** 262.666** (117.401) (118.857) (118.962) (126.438) (g) lnRKsi/zdMsid −279.165* −277.912* −277.927* −276.457* (165.706) (166.167) (166.227) (162.546) (h) lnwLsi/zdMsid −242.422 −246.943 −247.083 −247.963 (158.871) (157.313) (157.112) (155.468) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the number of hostile ICEWS events in the district divided by the population in the district. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, ICEWS. View Large Table A7. Input Distortions – Regression Coefficients Robustness: ICEWS Conflict Data – Per Capita Measure Coefficient of hostile ICEWS events/population (1) (2) (3) (4) (a) ln RKsi/zMsi −2.491 −0.445 1.816 0.581 (19.792) (20.311) (20.481) (22.404) (b) ln wLsi/zMsi 24.301 20.076 19.989 7.920 (25.335) (25.018) (24.126) (26.054) (c) ln RKsi/wLsi −30.349 −24.473 −24.968 −8.536 (21.935) (21.061) (20.316) (28.817) (d) lnzdMsid/zfMsif 500.402* 502.441* 512.438* 530.730* (292.202) (290.696) (291.054) (290.729) (e) lnRKsi/zfMsif 219.874 223.131* 232.890* 253.719* (134.968) (133.181) (132.817) (139.994) (f) lnwLsi/zfMsif 249.515** 244.651** 244.727** 262.666** (117.401) (118.857) (118.962) (126.438) (g) lnRKsi/zdMsid −279.165* −277.912* −277.927* −276.457* (165.706) (166.167) (166.227) (162.546) (h) lnwLsi/zdMsid −242.422 −246.943 −247.083 −247.963 (158.871) (157.313) (157.112) (155.468) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of hostile ICEWS events/population (1) (2) (3) (4) (a) ln RKsi/zMsi −2.491 −0.445 1.816 0.581 (19.792) (20.311) (20.481) (22.404) (b) ln wLsi/zMsi 24.301 20.076 19.989 7.920 (25.335) (25.018) (24.126) (26.054) (c) ln RKsi/wLsi −30.349 −24.473 −24.968 −8.536 (21.935) (21.061) (20.316) (28.817) (d) lnzdMsid/zfMsif 500.402* 502.441* 512.438* 530.730* (292.202) (290.696) (291.054) (290.729) (e) lnRKsi/zfMsif 219.874 223.131* 232.890* 253.719* (134.968) (133.181) (132.817) (139.994) (f) lnwLsi/zfMsif 249.515** 244.651** 244.727** 262.666** (117.401) (118.857) (118.962) (126.438) (g) lnRKsi/zdMsid −279.165* −277.912* −277.927* −276.457* (165.706) (166.167) (166.227) (162.546) (h) lnwLsi/zdMsid −242.422 −246.943 −247.083 −247.963 (158.871) (157.313) (157.112) (155.468) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the number of hostile ICEWS events in the district divided by the population in the district. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, ICEWS. View Large Table A8. Input Distortions – Regression Coefficients Robustness: Leads of fatalities Variable (1) (2) (3) (4) Panel (a) Dependent variable: lnzdMsid/zfMsif fatalitiest 0.773*** 0.773*** 0.806*** 0.815*** (0.297) (0.297) (0.298) (0.291) fatalitiest+1 −0.331 −0.331 −0.336 −0.329 (0.408) (0.408) (0.409) (0.404) Panel (b) Dependent variable: lnRKsi/zfMsif fatalitiest 0.300** 0.300** 0.331*** 0.320*** (0.121) (0.121) (0.122) (0.121) fatalitiest+1 −0.139 −0.136 −0.141 −0.133 (0.171) (0.172) (0.171) (0.169) Panel (c) Dependent variable: lnwLsi/zfMsif fatalitiest 0.275 0.261 0.261 0.266 (0.175) (0.176) (0.175) (0.184) fatalitiest+1 −0.071 −0.079 −0.086 −0.121 (0.193) (0.194) (0.191) (0.193) Panel (d) Dependent variable: lnRKsi/zdMsid fatalitiest −0.469** −0.469** −0.471** −0.493** (0.203) (0.203) (0.203) (0.194) fatalitiest+1 0.195 0.198 0.199 0.196 (0.262) (0.261) (0.261) (0.257) Panel (e) Dependent variable: lnwLsi/zdMsid fatalitiest −0.415** −0.423** −0.423** −0.411** (0.194) (0.194) (0.197) (0.195) fatalitiest+1 0.251 0.246 0.256 0.292 (0.254) (0.254) (0.254) (0.249) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y (1) (2) (3) (4) Panel (a) Dependent variable: lnzdMsid/zfMsif fatalitiest 0.773*** 0.773*** 0.806*** 0.815*** (0.297) (0.297) (0.298) (0.291) fatalitiest+1 −0.331 −0.331 −0.336 −0.329 (0.408) (0.408) (0.409) (0.404) Panel (b) Dependent variable: lnRKsi/zfMsif fatalitiest 0.300** 0.300** 0.331*** 0.320*** (0.121) (0.121) (0.122) (0.121) fatalitiest+1 −0.139 −0.136 −0.141 −0.133 (0.171) (0.172) (0.171) (0.169) Panel (c) Dependent variable: lnwLsi/zfMsif fatalitiest 0.275 0.261 0.261 0.266 (0.175) (0.176) (0.175) (0.184) fatalitiest+1 −0.071 −0.079 −0.086 −0.121 (0.193) (0.194) (0.191) (0.193) Panel (d) Dependent variable: lnRKsi/zdMsid fatalitiest −0.469** −0.469** −0.471** −0.493** (0.203) (0.203) (0.203) (0.194) fatalitiest+1 0.195 0.198 0.199 0.196 (0.262) (0.261) (0.261) (0.257) Panel (e) Dependent variable: lnwLsi/zdMsid fatalitiest −0.415** −0.423** −0.423** −0.411** (0.194) (0.194) (0.197) (0.195) fatalitiest+1 0.251 0.246 0.256 0.292 (0.254) (0.254) (0.254) (0.249) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities variable and its lead. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The variable closure days captures the yearly number of days of border closure, while dtpassage measures road distance of the district capital from the closest entry passage as measured in 10 kilometre units. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A8. Input Distortions – Regression Coefficients Robustness: Leads of fatalities Variable (1) (2) (3) (4) Panel (a) Dependent variable: lnzdMsid/zfMsif fatalitiest 0.773*** 0.773*** 0.806*** 0.815*** (0.297) (0.297) (0.298) (0.291) fatalitiest+1 −0.331 −0.331 −0.336 −0.329 (0.408) (0.408) (0.409) (0.404) Panel (b) Dependent variable: lnRKsi/zfMsif fatalitiest 0.300** 0.300** 0.331*** 0.320*** (0.121) (0.121) (0.122) (0.121) fatalitiest+1 −0.139 −0.136 −0.141 −0.133 (0.171) (0.172) (0.171) (0.169) Panel (c) Dependent variable: lnwLsi/zfMsif fatalitiest 0.275 0.261 0.261 0.266 (0.175) (0.176) (0.175) (0.184) fatalitiest+1 −0.071 −0.079 −0.086 −0.121 (0.193) (0.194) (0.191) (0.193) Panel (d) Dependent variable: lnRKsi/zdMsid fatalitiest −0.469** −0.469** −0.471** −0.493** (0.203) (0.203) (0.203) (0.194) fatalitiest+1 0.195 0.198 0.199 0.196 (0.262) (0.261) (0.261) (0.257) Panel (e) Dependent variable: lnwLsi/zdMsid fatalitiest −0.415** −0.423** −0.423** −0.411** (0.194) (0.194) (0.197) (0.195) fatalitiest+1 0.251 0.246 0.256 0.292 (0.254) (0.254) (0.254) (0.249) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y (1) (2) (3) (4) Panel (a) Dependent variable: lnzdMsid/zfMsif fatalitiest 0.773*** 0.773*** 0.806*** 0.815*** (0.297) (0.297) (0.298) (0.291) fatalitiest+1 −0.331 −0.331 −0.336 −0.329 (0.408) (0.408) (0.409) (0.404) Panel (b) Dependent variable: lnRKsi/zfMsif fatalitiest 0.300** 0.300** 0.331*** 0.320*** (0.121) (0.121) (0.122) (0.121) fatalitiest+1 −0.139 −0.136 −0.141 −0.133 (0.171) (0.172) (0.171) (0.169) Panel (c) Dependent variable: lnwLsi/zfMsif fatalitiest 0.275 0.261 0.261 0.266 (0.175) (0.176) (0.175) (0.184) fatalitiest+1 −0.071 −0.079 −0.086 −0.121 (0.193) (0.194) (0.191) (0.193) Panel (d) Dependent variable: lnRKsi/zdMsid fatalitiest −0.469** −0.469** −0.471** −0.493** (0.203) (0.203) (0.203) (0.194) fatalitiest+1 0.195 0.198 0.199 0.196 (0.262) (0.261) (0.261) (0.257) Panel (e) Dependent variable: lnwLsi/zdMsid fatalitiest −0.415** −0.423** −0.423** −0.411** (0.194) (0.194) (0.197) (0.195) fatalitiest+1 0.251 0.246 0.256 0.292 (0.254) (0.254) (0.254) (0.249) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports estimates of the coefficient of the fatalities variable and its lead. Standard errors are clustered along both sector‐year and district‐year categories. Dependent variable is log of ratio of input values in Israeli New Sheqel (NIS). The main independent variable is the number of Palestinians killed by IDF in the year and district where surveyed establishment is located (measured in standard deviation units). RKsi is value of capital; zMsi is value of materials; wLsi is value of labour; zfMsif is value of imported materials; zdMsid is value of domestically produced materials. The variable closure days captures the yearly number of days of border closure, while dtpassage measures road distance of the district capital from the closest entry passage as measured in 10 kilometre units. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A9. Input Distortions – Regression Coefficients Robustness: Road Distance of District Capital from Closest Entry Passage Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi −0.004 −0.006 −0.001 −0.001 (0.039) (0.040) (0.041) (0.043) (b) ln wLsi/zMsi 0.007 0.003 0.004 −0.009 (0.045) (0.044) (0.043) (0.046) (c) ln RKsi/wLsi 0.001 0.007 0.007 0.020 (0.036) (0.035) (0.035) (0.037) (d) lnzdMsid/zfMsif 1.174*** 1.173*** 1.194*** 1.204*** (0.260) (0.260) (0.260) (0.259) (e) lnRKsi/zfMsif 0.504*** 0.502*** 0.522*** 0.536*** (0.121) (0.121) (0.119) (0.126) (f) lnwLsi/zfMsif 0.452*** 0.446*** 0.445*** 0.473*** (0.122) (0.125) (0.124) (0.133) (g) lnRKsi/zdMsid −0.667*** −0.668*** −0.668*** −0.668*** (0.162) (0.162) (0.162) (0.156) (h) lnwLsi/zdMsid −0.664*** −0.669*** −0.667*** −0.668*** (0.178) (0.175) (0.175) (0.177) dtpassage × Year FE Y Y Y Y Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi −0.004 −0.006 −0.001 −0.001 (0.039) (0.040) (0.041) (0.043) (b) ln wLsi/zMsi 0.007 0.003 0.004 −0.009 (0.045) (0.044) (0.043) (0.046) (c) ln RKsi/wLsi 0.001 0.007 0.007 0.020 (0.036) (0.035) (0.035) (0.037) (d) lnzdMsid/zfMsif 1.174*** 1.173*** 1.194*** 1.204*** (0.260) (0.260) (0.260) (0.259) (e) lnRKsi/zfMsif 0.504*** 0.502*** 0.522*** 0.536*** (0.121) (0.121) (0.119) (0.126) (f) lnwLsi/zfMsif 0.452*** 0.446*** 0.445*** 0.473*** (0.122) (0.125) (0.124) (0.133) (g) lnRKsi/zdMsid −0.667*** −0.668*** −0.668*** −0.668*** (0.162) (0.162) (0.162) (0.156) (h) lnwLsi/zdMsid −0.664*** −0.669*** −0.667*** −0.668*** (0.178) (0.175) (0.175) (0.177) dtpassage × Year FE Y Y Y Y Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes The Table reports implied relative distortion values (together with 95% confidence intervals) as derived from estimating the input value ratio regression, including as regressors the full set of year dummies interacted with the road distance of the district capital from the closest entry passage as measured in 10 kilometre units. Standard errors are clustered along both sector‐year and district‐year categories. τK is average distortion level for capital; τM is average distortion level for materials; τL is average distortion value for labour; τMf is average distortion value for imported materials; τMd is average distortion value for domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set (for all other cases). Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A9. Input Distortions – Regression Coefficients Robustness: Road Distance of District Capital from Closest Entry Passage Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi −0.004 −0.006 −0.001 −0.001 (0.039) (0.040) (0.041) (0.043) (b) ln wLsi/zMsi 0.007 0.003 0.004 −0.009 (0.045) (0.044) (0.043) (0.046) (c) ln RKsi/wLsi 0.001 0.007 0.007 0.020 (0.036) (0.035) (0.035) (0.037) (d) lnzdMsid/zfMsif 1.174*** 1.173*** 1.194*** 1.204*** (0.260) (0.260) (0.260) (0.259) (e) lnRKsi/zfMsif 0.504*** 0.502*** 0.522*** 0.536*** (0.121) (0.121) (0.119) (0.126) (f) lnwLsi/zfMsif 0.452*** 0.446*** 0.445*** 0.473*** (0.122) (0.125) (0.124) (0.133) (g) lnRKsi/zdMsid −0.667*** −0.668*** −0.668*** −0.668*** (0.162) (0.162) (0.162) (0.156) (h) lnwLsi/zdMsid −0.664*** −0.669*** −0.667*** −0.668*** (0.178) (0.175) (0.175) (0.177) dtpassage × Year FE Y Y Y Y Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi −0.004 −0.006 −0.001 −0.001 (0.039) (0.040) (0.041) (0.043) (b) ln wLsi/zMsi 0.007 0.003 0.004 −0.009 (0.045) (0.044) (0.043) (0.046) (c) ln RKsi/wLsi 0.001 0.007 0.007 0.020 (0.036) (0.035) (0.035) (0.037) (d) lnzdMsid/zfMsif 1.174*** 1.173*** 1.194*** 1.204*** (0.260) (0.260) (0.260) (0.259) (e) lnRKsi/zfMsif 0.504*** 0.502*** 0.522*** 0.536*** (0.121) (0.121) (0.119) (0.126) (f) lnwLsi/zfMsif 0.452*** 0.446*** 0.445*** 0.473*** (0.122) (0.125) (0.124) (0.133) (g) lnRKsi/zdMsid −0.667*** −0.668*** −0.668*** −0.668*** (0.162) (0.162) (0.162) (0.156) (h) lnwLsi/zdMsid −0.664*** −0.669*** −0.667*** −0.668*** (0.178) (0.175) (0.175) (0.177) dtpassage × Year FE Y Y Y Y Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes The Table reports implied relative distortion values (together with 95% confidence intervals) as derived from estimating the input value ratio regression, including as regressors the full set of year dummies interacted with the road distance of the district capital from the closest entry passage as measured in 10 kilometre units. Standard errors are clustered along both sector‐year and district‐year categories. τK is average distortion level for capital; τM is average distortion level for materials; τL is average distortion value for labour; τMf is average distortion value for imported materials; τMd is average distortion value for domestically produced materials. The number of observations is 12,410. Observations are weighted using the original weight in the Industry Survey data set (for all other cases). Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A10. Input Distortions and Homothetic Production Functions Regression Coefficients: Restricted Sample Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.027 0.023 0.029 0.022 (0.048) (0.048) (0.049) (0.052) (b) ln wLsi/zMsi 0.058 0.059 0.057 0.045 (0.046) (0.044) (0.045) (0.049) (c) ln RKsi/wLsi −0.012 −0.013 −0.010 −0.005 (0.052) (0.050) (0.049) (0.050) (d) lnzdMsid/zfMsif 1.247*** 1.246*** 1.265*** 1.263*** (0.300) (0.299) (0.296) (0.294) (e) lnRKsi/zfMsif 0.542*** 0.539*** 0.559*** 0.556*** (0.137) (0.136) (0.131) (0.140) (f) lnwLsi/zfMsif 0.498*** 0.498*** 0.500*** 0.506*** (0.176) (0.178) (0.177) (0.184) (g) lnRKsi/zdMsid −0.707*** −0.709*** −0.708*** −0.709*** (0.185) (0.185) (0.185) (0.178) (h) lnwLsi/zdMsid −0.679*** −0.679*** −0.681*** −0.668*** (0.211) (0.207) (0.206) (0.205) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.027 0.023 0.029 0.022 (0.048) (0.048) (0.049) (0.052) (b) ln wLsi/zMsi 0.058 0.059 0.057 0.045 (0.046) (0.044) (0.045) (0.049) (c) ln RKsi/wLsi −0.012 −0.013 −0.010 −0.005 (0.052) (0.050) (0.049) (0.050) (d) lnzdMsid/zfMsif 1.247*** 1.246*** 1.265*** 1.263*** (0.300) (0.299) (0.296) (0.294) (e) lnRKsi/zfMsif 0.542*** 0.539*** 0.559*** 0.556*** (0.137) (0.136) (0.131) (0.140) (f) lnwLsi/zfMsif 0.498*** 0.498*** 0.500*** 0.506*** (0.176) (0.178) (0.177) (0.184) (g) lnRKsi/zdMsid −0.707*** −0.709*** −0.708*** −0.709*** (0.185) (0.185) (0.185) (0.178) (h) lnwLsi/zdMsid −0.679*** −0.679*** −0.681*** −0.668*** (0.211) (0.207) (0.206) (0.205) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes The Table reports implied relative distortion values (together with 95% confidence intervals) as derived from estimating the input value ratio regression over the restricted sample of observations belonging to sectors where no significant relationship between material value ratio and output value is found in 1999. Standard errors are clustered along both sector‐year and district‐year categories. τK is average distortion level for capital; τM is average distortion level for materials; τL is average distortion value for labour; τMf is average distortion value for imported materials; τMd is average distortion value for domestically produced materials. The number of observations is 7,957. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A10. Input Distortions and Homothetic Production Functions Regression Coefficients: Restricted Sample Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.027 0.023 0.029 0.022 (0.048) (0.048) (0.049) (0.052) (b) ln wLsi/zMsi 0.058 0.059 0.057 0.045 (0.046) (0.044) (0.045) (0.049) (c) ln RKsi/wLsi −0.012 −0.013 −0.010 −0.005 (0.052) (0.050) (0.049) (0.050) (d) lnzdMsid/zfMsif 1.247*** 1.246*** 1.265*** 1.263*** (0.300) (0.299) (0.296) (0.294) (e) lnRKsi/zfMsif 0.542*** 0.539*** 0.559*** 0.556*** (0.137) (0.136) (0.131) (0.140) (f) lnwLsi/zfMsif 0.498*** 0.498*** 0.500*** 0.506*** (0.176) (0.178) (0.177) (0.184) (g) lnRKsi/zdMsid −0.707*** −0.709*** −0.708*** −0.709*** (0.185) (0.185) (0.185) (0.178) (h) lnwLsi/zdMsid −0.679*** −0.679*** −0.681*** −0.668*** (0.211) (0.207) (0.206) (0.205) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Coefficient of fatalities variable (1) (2) (3) (4) (a) ln RKsi/zMsi 0.027 0.023 0.029 0.022 (0.048) (0.048) (0.049) (0.052) (b) ln wLsi/zMsi 0.058 0.059 0.057 0.045 (0.046) (0.044) (0.045) (0.049) (c) ln RKsi/wLsi −0.012 −0.013 −0.010 −0.005 (0.052) (0.050) (0.049) (0.050) (d) lnzdMsid/zfMsif 1.247*** 1.246*** 1.265*** 1.263*** (0.300) (0.299) (0.296) (0.294) (e) lnRKsi/zfMsif 0.542*** 0.539*** 0.559*** 0.556*** (0.137) (0.136) (0.131) (0.140) (f) lnwLsi/zfMsif 0.498*** 0.498*** 0.500*** 0.506*** (0.176) (0.178) (0.177) (0.184) (g) lnRKsi/zdMsid −0.707*** −0.709*** −0.708*** −0.709*** (0.185) (0.185) (0.185) (0.178) (h) lnwLsi/zdMsid −0.679*** −0.679*** −0.681*** −0.668*** (0.211) (0.207) (0.206) (0.205) Family workersTotal N Y Y Y ProprietorsTotal N N Y Y Sector FE Y Y Y n.a. Year FE Y Y Y n.a. District FE Y Y Y Y Sector × Year FE N N N Y Notes The Table reports implied relative distortion values (together with 95% confidence intervals) as derived from estimating the input value ratio regression over the restricted sample of observations belonging to sectors where no significant relationship between material value ratio and output value is found in 1999. Standard errors are clustered along both sector‐year and district‐year categories. τK is average distortion level for capital; τM is average distortion level for materials; τL is average distortion value for labour; τMf is average distortion value for imported materials; τMd is average distortion value for domestically produced materials. The number of observations is 7,957. Observations are weighted using the original weight in the Industry Survey data set. Sources. Industry Survey, Palestinian Bureau of Statistics, B’Tselem. View Large Table A11. Factor Share Parameter Estimates Sector (ISIC) K L Mf Md Other mining and quarrying 0.288 0.022 0.040 0.649*** (0.204) (0.054) (0.035) (0.231) Manufacture of food products and beverages 0.176*** 0.035** 0.075*** 0.715*** (0.058) (0.017) (0.017) (0.063) Manufacture of tobacco products 0.622*** 0.236*** 0.100 0.042 (0.067) (0.063) (0.086) (0.034) Manufacture of textiles 0.287** 0.054* 0.136*** 0.524*** (0.136) (0.028) (0.034) (0.116) Manufacture of wearing apparel 0.132 0.034 0.135** 0.699*** (0.127) (0.049) (0.058) (0.118) Tanning and dressing of leather 0.215*** 0.014 0.088*** 0.683*** (0.059) (0.021) (0.020) (0.057) Manufacture of wood and of wooden products 0.170*** 0.020 0.079*** 0.732*** (0.057) (0.023) (0.027) (0.070) Manufacture of paper and paper products 0.296*** −0.091 0.136*** 0.659*** (0.083) (0.072) (0.030) (0.097) Publishing, printing, etc. 0.098* 0.036* 0.115*** 0.751*** (0.059) (0.021) (0.030) (0.080) Manufacture of chemicals and chem. products 0.344*** −0.003 0.199*** 0.460*** (0.111) (0.031) (0.026) (0.108) Manufacture of rubber and plastics products 0.472*** 0.059 0.087*** 0.382*** (0.066) (0.040) (0.025) (0.063) Manufacture of other non‐metallic mineral prod. 0.154*** 0.042** 0.036* 0.768*** (0.058) (0.020) (0.020) (0.058) Manufacture of basic metals 0.073 0.083 0.108*** 0.736*** (0.140) (0.080) (0.024) (0.095) Manufacture of fabricated metal products 0.083 0.022 0.047*** 0.848*** (0.056) (0.016) (0.017) (0.051) Manufacture of machinery and equipment n.e.c. 0.289*** 0.008 0.113*** 0.590*** (0.071) (0.023) (0.025) (0.084) Manufacture of medical, precision and optical instruments 0.661*** 0.051 0.099*** 0.189* (0.068) (0.053) (0.038) (0.108) Manufacture of motor vehicles, trailers and semi‐trailers 0.005 0.068** 0.114*** 0.814*** (0.106) (0.033) (0.037) (0.099) Manufacture of other transport equipment −0.018 −0.215*** 0.004 1.229*** (0.090) (0.030) (0.029) (0.109) Manufacture of furniture; manufacturing n.e.c. 0.219*** 0.032* 0.056** 0.694*** (0.066) (0.019) (0.026) (0.072) Electricity, gas, steam and hot water supply 0.295** 0.174* 0.075** 0.456*** (0.144) (0.094) (0.037) (0.087) Collection, purification and distribution of water 0.431*** 0.038 0.107*** 0.423*** (0.135) (0.059) (0.032) (0.157) Sector (ISIC) K L Mf Md Other mining and quarrying 0.288 0.022 0.040 0.649*** (0.204) (0.054) (0.035) (0.231) Manufacture of food products and beverages 0.176*** 0.035** 0.075*** 0.715*** (0.058) (0.017) (0.017) (0.063) Manufacture of tobacco products 0.622*** 0.236*** 0.100 0.042 (0.067) (0.063) (0.086) (0.034) Manufacture of textiles 0.287** 0.054* 0.136*** 0.524*** (0.136) (0.028) (0.034) (0.116) Manufacture of wearing apparel 0.132 0.034 0.135** 0.699*** (0.127) (0.049) (0.058) (0.118) Tanning and dressing of leather 0.215*** 0.014 0.088*** 0.683*** (0.059) (0.021) (0.020) (0.057) Manufacture of wood and of wooden products 0.170*** 0.020 0.079*** 0.732*** (0.057) (0.023) (0.027) (0.070) Manufacture of paper and paper products 0.296*** −0.091 0.136*** 0.659*** (0.083) (0.072) (0.030) (0.097) Publishing, printing, etc. 0.098* 0.036* 0.115*** 0.751*** (0.059) (0.021) (0.030) (0.080) Manufacture of chemicals and chem. products 0.344*** −0.003 0.199*** 0.460*** (0.111) (0.031) (0.026) (0.108) Manufacture of rubber and plastics products 0.472*** 0.059 0.087*** 0.382*** (0.066) (0.040) (0.025) (0.063) Manufacture of other non‐metallic mineral prod. 0.154*** 0.042** 0.036* 0.768*** (0.058) (0.020) (0.020) (0.058) Manufacture of basic metals 0.073 0.083 0.108*** 0.736*** (0.140) (0.080) (0.024) (0.095) Manufacture of fabricated metal products 0.083 0.022 0.047*** 0.848*** (0.056) (0.016) (0.017) (0.051) Manufacture of machinery and equipment n.e.c. 0.289*** 0.008 0.113*** 0.590*** (0.071) (0.023) (0.025) (0.084) Manufacture of medical, precision and optical instruments 0.661*** 0.051 0.099*** 0.189* (0.068) (0.053) (0.038) (0.108) Manufacture of motor vehicles, trailers and semi‐trailers 0.005 0.068** 0.114*** 0.814*** (0.106) (0.033) (0.037) (0.099) Manufacture of other transport equipment −0.018 −0.215*** 0.004 1.229*** (0.090) (0.030) (0.029) (0.109) Manufacture of furniture; manufacturing n.e.c. 0.219*** 0.032* 0.056** 0.694*** (0.066) (0.019) (0.026) (0.072) Electricity, gas, steam and hot water supply 0.295** 0.174* 0.075** 0.456*** (0.144) (0.094) (0.037) (0.087) Collection, purification and distribution of water 0.431*** 0.038 0.107*** 0.423*** (0.135) (0.059) (0.032) (0.157) Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports OLS factor share parameter estimates for each of the inputs as derived following the procedure explained in Section 7. We estimate the parameters from (16) by restricting the sample to those observations belonging to the year 1999, prior to the beginning of the conflict. In order to identify the parameters of interest, we exclude those four sectors for which we have 5 or less establishment‐level observations. The number of observations is 1,322. Observations are weighted using the original weight in the Industry Survey data set. Sources: Industry Survey, Palestinian Bureau of Statistics. View Large Table A11. Factor Share Parameter Estimates Sector (ISIC) K L Mf Md Other mining and quarrying 0.288 0.022 0.040 0.649*** (0.204) (0.054) (0.035) (0.231) Manufacture of food products and beverages 0.176*** 0.035** 0.075*** 0.715*** (0.058) (0.017) (0.017) (0.063) Manufacture of tobacco products 0.622*** 0.236*** 0.100 0.042 (0.067) (0.063) (0.086) (0.034) Manufacture of textiles 0.287** 0.054* 0.136*** 0.524*** (0.136) (0.028) (0.034) (0.116) Manufacture of wearing apparel 0.132 0.034 0.135** 0.699*** (0.127) (0.049) (0.058) (0.118) Tanning and dressing of leather 0.215*** 0.014 0.088*** 0.683*** (0.059) (0.021) (0.020) (0.057) Manufacture of wood and of wooden products 0.170*** 0.020 0.079*** 0.732*** (0.057) (0.023) (0.027) (0.070) Manufacture of paper and paper products 0.296*** −0.091 0.136*** 0.659*** (0.083) (0.072) (0.030) (0.097) Publishing, printing, etc. 0.098* 0.036* 0.115*** 0.751*** (0.059) (0.021) (0.030) (0.080) Manufacture of chemicals and chem. products 0.344*** −0.003 0.199*** 0.460*** (0.111) (0.031) (0.026) (0.108) Manufacture of rubber and plastics products 0.472*** 0.059 0.087*** 0.382*** (0.066) (0.040) (0.025) (0.063) Manufacture of other non‐metallic mineral prod. 0.154*** 0.042** 0.036* 0.768*** (0.058) (0.020) (0.020) (0.058) Manufacture of basic metals 0.073 0.083 0.108*** 0.736*** (0.140) (0.080) (0.024) (0.095) Manufacture of fabricated metal products 0.083 0.022 0.047*** 0.848*** (0.056) (0.016) (0.017) (0.051) Manufacture of machinery and equipment n.e.c. 0.289*** 0.008 0.113*** 0.590*** (0.071) (0.023) (0.025) (0.084) Manufacture of medical, precision and optical instruments 0.661*** 0.051 0.099*** 0.189* (0.068) (0.053) (0.038) (0.108) Manufacture of motor vehicles, trailers and semi‐trailers 0.005 0.068** 0.114*** 0.814*** (0.106) (0.033) (0.037) (0.099) Manufacture of other transport equipment −0.018 −0.215*** 0.004 1.229*** (0.090) (0.030) (0.029) (0.109) Manufacture of furniture; manufacturing n.e.c. 0.219*** 0.032* 0.056** 0.694*** (0.066) (0.019) (0.026) (0.072) Electricity, gas, steam and hot water supply 0.295** 0.174* 0.075** 0.456*** (0.144) (0.094) (0.037) (0.087) Collection, purification and distribution of water 0.431*** 0.038 0.107*** 0.423*** (0.135) (0.059) (0.032) (0.157) Sector (ISIC) K L Mf Md Other mining and quarrying 0.288 0.022 0.040 0.649*** (0.204) (0.054) (0.035) (0.231) Manufacture of food products and beverages 0.176*** 0.035** 0.075*** 0.715*** (0.058) (0.017) (0.017) (0.063) Manufacture of tobacco products 0.622*** 0.236*** 0.100 0.042 (0.067) (0.063) (0.086) (0.034) Manufacture of textiles 0.287** 0.054* 0.136*** 0.524*** (0.136) (0.028) (0.034) (0.116) Manufacture of wearing apparel 0.132 0.034 0.135** 0.699*** (0.127) (0.049) (0.058) (0.118) Tanning and dressing of leather 0.215*** 0.014 0.088*** 0.683*** (0.059) (0.021) (0.020) (0.057) Manufacture of wood and of wooden products 0.170*** 0.020 0.079*** 0.732*** (0.057) (0.023) (0.027) (0.070) Manufacture of paper and paper products 0.296*** −0.091 0.136*** 0.659*** (0.083) (0.072) (0.030) (0.097) Publishing, printing, etc. 0.098* 0.036* 0.115*** 0.751*** (0.059) (0.021) (0.030) (0.080) Manufacture of chemicals and chem. products 0.344*** −0.003 0.199*** 0.460*** (0.111) (0.031) (0.026) (0.108) Manufacture of rubber and plastics products 0.472*** 0.059 0.087*** 0.382*** (0.066) (0.040) (0.025) (0.063) Manufacture of other non‐metallic mineral prod. 0.154*** 0.042** 0.036* 0.768*** (0.058) (0.020) (0.020) (0.058) Manufacture of basic metals 0.073 0.083 0.108*** 0.736*** (0.140) (0.080) (0.024) (0.095) Manufacture of fabricated metal products 0.083 0.022 0.047*** 0.848*** (0.056) (0.016) (0.017) (0.051) Manufacture of machinery and equipment n.e.c. 0.289*** 0.008 0.113*** 0.590*** (0.071) (0.023) (0.025) (0.084) Manufacture of medical, precision and optical instruments 0.661*** 0.051 0.099*** 0.189* (0.068) (0.053) (0.038) (0.108) Manufacture of motor vehicles, trailers and semi‐trailers 0.005 0.068** 0.114*** 0.814*** (0.106) (0.033) (0.037) (0.099) Manufacture of other transport equipment −0.018 −0.215*** 0.004 1.229*** (0.090) (0.030) (0.029) (0.109) Manufacture of furniture; manufacturing n.e.c. 0.219*** 0.032* 0.056** 0.694*** (0.066) (0.019) (0.026) (0.072) Electricity, gas, steam and hot water supply 0.295** 0.174* 0.075** 0.456*** (0.144) (0.094) (0.037) (0.087) Collection, purification and distribution of water 0.431*** 0.038 0.107*** 0.423*** (0.135) (0.059) (0.032) (0.157) Notes * p‐value < 0.1; ** p‐value < 0.05; *** p‐value < 0.01. The Table reports OLS factor share parameter estimates for each of the inputs as derived following the procedure explained in Section 7. We estimate the parameters from (16) by restricting the sample to those observations belonging to the year 1999, prior to the beginning of the conflict. In order to identify the parameters of interest, we exclude those four sectors for which we have 5 or less establishment‐level observations. The number of observations is 1,322. Observations are weighted using the original weight in the Industry Survey data set. Sources: Industry Survey, Palestinian Bureau of Statistics. View Large A.1. Additional Tables and Figures Figure A1. View largeDownload slide Labour and Output Value. (a) Distribution of Employment and (b) Distribution of Output Value Notes. The Figures show the distribution of number of workers and value of output for Palestinian firms. Sources. Palestinian Central Bureau of Statistics. B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A1. View largeDownload slide Labour and Output Value. (a) Distribution of Employment and (b) Distribution of Output Value Notes. The Figures show the distribution of number of workers and value of output for Palestinian firms. Sources. Palestinian Central Bureau of Statistics. B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A2. View largeDownload slide Cross‐district and Time Conflict Variability – Maps. (a) Cross‐district Variability. (b) Within‐district Variability Notes. The maps show the distribution of the number of Palestinians killed by IDF across districts in given years and its changes over given time spans. In each map, districts are coloured according to the quintiles they belong to in the distribution of levels and changes respectively. Looking at the top maps, we see that there is large cross‐district variation in the number of fatalities. The three bottom maps show that there is also meaningful variation in the number of fatalities within each district over time. In particular, differential changes in conflict intensity across districts constitute a source of variability which does not seem to overlap with the cross‐sectional one. Sources. B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A2. View largeDownload slide Cross‐district and Time Conflict Variability – Maps. (a) Cross‐district Variability. (b) Within‐district Variability Notes. The maps show the distribution of the number of Palestinians killed by IDF across districts in given years and its changes over given time spans. In each map, districts are coloured according to the quintiles they belong to in the distribution of levels and changes respectively. Looking at the top maps, we see that there is large cross‐district variation in the number of fatalities. The three bottom maps show that there is also meaningful variation in the number of fatalities within each district over time. In particular, differential changes in conflict intensity across districts constitute a source of variability which does not seem to overlap with the cross‐sectional one. Sources. B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A3. View largeDownload slide Conflict and Value of Foreign Trade Notes. The Figure plots the evolution of the total real value of the net balance of trade over time, together with the evolution of total number of Palestinians killed by IDF. Sources. Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A3. View largeDownload slide Conflict and Value of Foreign Trade Notes. The Figure plots the evolution of the total real value of the net balance of trade over time, together with the evolution of total number of Palestinians killed by IDF. Sources. Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A4. View largeDownload slide Trade Composition: Imports Notes. The Figures plot import composition (sector share over total import) in 1999 and 2002. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A4. View largeDownload slide Trade Composition: Imports Notes. The Figures plot import composition (sector share over total import) in 1999 and 2002. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A5. View largeDownload slide Trade Composition: Exports Notes. The Figures plot the export composition (sector share over total export) in 1999 and 2002. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A5. View largeDownload slide Trade Composition: Exports Notes. The Figures plot the export composition (sector share over total export) in 1999 and 2002. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A6. View largeDownload slide Trade Composition: Imports By Origin Notes. The Figures plot import composition (sector share over total import) in 1999 and 2002, together with the Israeli share of each import category. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A6. View largeDownload slide Trade Composition: Imports By Origin Notes. The Figures plot import composition (sector share over total import) in 1999 and 2002, together with the Israeli share of each import category. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A7 shows import value shares for 1999 and 2002 at a more disaggregated level. Those ISIC 2‐digit sectors for which import shares decrease the most are: textile yarn, fabrics, and related products; non‐metallic mineral manufactures; manufactures of metals; machinery specialised for particular industries; industrial machinery and equipment, and parts; telecommunications and sound‐recording; electrical machinery apparatus and parts thereof; road vehicles (including air‐cushion vehicles); furniture and parts thereof; miscellaneous manufactured articles. Most of these sectors are producers of manufacturing inputs. Although we cannot directly disentangle inputs from finished products, we interpret this aggregate evidence on trade flows as being consistent with our findings at the firm level. Figure A7. View largeDownload slide Disaggregated Import Composition Notes. The Figures plot import composition (sector share over total import) in 1999 and 2002 at the ISIC 2‐digit sector level. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A7. View largeDownload slide Disaggregated Import Composition Notes. The Figures plot import composition (sector share over total import) in 1999 and 2002 at the ISIC 2‐digit sector level. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A8. View largeDownload slide Conflict and Import and Domestic Wholesale Prices Notes. The Figure shows the evolution of the ratio between imported vs. domestic wholesale prices, together with the evolution of the total number of Palestinians killed by IDF. Sources. Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A8. View largeDownload slide Conflict and Import and Domestic Wholesale Prices Notes. The Figure shows the evolution of the ratio between imported vs. domestic wholesale prices, together with the evolution of the total number of Palestinians killed by IDF. Sources. Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A9. View largeDownload slide Conflict and Palestinian GDP Notes. The Figure shows the evolution of real Palestine GDP (Million US$) over time, together with the evolution of the total number of Palestinians killed by IDF. Sources. Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A9. View largeDownload slide Conflict and Palestinian GDP Notes. The Figure shows the evolution of real Palestine GDP (Million US$) over time, together with the evolution of the total number of Palestinians killed by IDF. Sources. Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A10. View largeDownload slide Conflict and Aggregate Output Value Notes. The Figure shows the evolution of the total real value of production in Israeli New Sheqel (NIS) over time, as derived from the Industry Survey. It plots the weighted sum of establishments’ output value over time together with Palestinians fatalities in the same period. Establishment‐level output values are aggregated after adjusting its value using yearly 2‐digit sector‐level deflators. The Figure also plots the evolution of the total number of Palestinians killed by IDF in the same years. Sources. Industry Survey, Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A10. View largeDownload slide Conflict and Aggregate Output Value Notes. The Figure shows the evolution of the total real value of production in Israeli New Sheqel (NIS) over time, as derived from the Industry Survey. It plots the weighted sum of establishments’ output value over time together with Palestinians fatalities in the same period. Establishment‐level output values are aggregated after adjusting its value using yearly 2‐digit sector‐level deflators. The Figure also plots the evolution of the total number of Palestinians killed by IDF in the same years. Sources. Industry Survey, Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A11. View largeDownload slide Producer Price Index and Conflict Notes. The Figures plot the evolution of Producer Price Indexes for selected 2‐digit sectors clustered in one particular district over time, together with the total number of Palestinians killed by IDF in the same district. Sources. Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A11. View largeDownload slide Producer Price Index and Conflict Notes. The Figures plot the evolution of Producer Price Indexes for selected 2‐digit sectors clustered in one particular district over time, together with the total number of Palestinians killed by IDF in the same district. Sources. Palestinian Central Bureau of Statistics; B’Tselem. Colour figure can be viewed at wileyonlinelibrary.com. Figure A12. View largeDownload slide Estimated Elasticities of Substitution Within Materials Notes. The Figure plots the estimated elasticity of substitution between foreign and domestically produced materials in each ISIC 2‐digit sector following the procedure described above. Estimated elasticities are remarkably close to one for all sectors, thus supporting the counterfactual analysis in subsection 7.2. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. Figure A12. View largeDownload slide Estimated Elasticities of Substitution Within Materials Notes. The Figure plots the estimated elasticity of substitution between foreign and domestically produced materials in each ISIC 2‐digit sector following the procedure described above. Estimated elasticities are remarkably close to one for all sectors, thus supporting the counterfactual analysis in subsection 7.2. Sources. Palestinian Central Bureau of Statistics. Colour figure can be viewed at wileyonlinelibrary.com. A.2. Elasticity of Substitution between Foreign and Domestically Produced Materials The validity of our counterfactual analysis in subsection 7.2 rests on the assumptions we make on the shape of the production function. That is, we assume a Cobb–Douglas production function with sector‐specific factor shares and structurally estimate the corresponding parameters from the data. One possible concern with such procedure is that the true production function may have a different shape. In particular, foreign and domestically produced materials may be a very close substitute, with the elasticity of substitution between the two being far greater than one. If this was the case, substitution between the two would have a small impact on output. The procedure we implement by assuming a Cobb–Douglas production function would thus lead to an overestimation of the impact of material input substitution on output value. Notice first that the claim that foreign and domestically produced materials are very close substitute finds little support in the existing literature. A number of empirical papers document productivity increases in domestic firms through access to imported inputs (Schor, 2004; Amiti and Konings, 2007; Kasahara and Rodrigue, 2008; Topalova and Khandelwal, 2011; Boehm et al., 2015). Despite the limitations of our data, we take these concerns seriously and structurally estimate the elasticity of substitution between foreign and domestically produced materials in the data. Consider the expression of output value in (A.1) in the following modified version: PsiYsi=σσ−111−τYi1+τKiαsαs1+τLiβsβs1+τMi1−αs−βs1−αs−βs(RKsi)αs(wLsi)βs(z~M~si)1−αs−βs, A.1 where z~M~si is the adjusted value of materials in production. This value is not observed by the econometrician. It combines both foreign and domestically produced materials according to the following expression: z~M~si=μiszdMisdεs−1εs+(1−μis)zfMisfεs−1εsεsεs−1, A.2 where μis is the relative efficiency of domestically produced materials for firm i operating in sector s. By contrast, εs is the sector‐specific constant elasticity of substitution between foreign and domestically produced materials in production. As εs approaches 1, the above expression reduces to a Cobb–Douglas production function, which maps into the empirical analysis we implement in subsection 7.2. If instead εs is far greater than 1, foreign and domestically produced materials are very close substitute, thus moving away from the Cobb–Douglas framework and questioning the validity of our main exercise. It follows that evidence of εs close to 1 would be supportive of our approach and counterfactual analysis. Our goal is therefore to provide estimates of εs ⁠. We implement a two‐stage estimation procedure. First, notice that: lnz~M~si=εsεs−1lnμiszdMisdεs−1εs+(1−μis)zfMisfεs−1εslnz~M~si=lnzfMisf+εsεs−1lnμiszdMisdzfMisfεs−1εs+1−μis. A.3 After taking logs of (A.1) and substituting the above expression we get: lnPsiYsi=lnσσ−1+ln1αsαs1βsβs11−αs−βs1−αs−βs++αslnRKsi+βslnwLsi+(1−αs−βs)lnzfMisf++(1−αs−βs)εsεs−1lnμiszdMisdzfMisfεs−1εs+1−μis. A.4 Using firm‐level data from 1999, we can therefore run a first regression of ln PsiYsi on ln RKsi, ln wLsi and lnzfMisf ⁠, with sector fixed effects, and sector‐specific factor shares. The residuals from such regression are equal to: uis=(1−αs−βs)εsεs−1lnμiszdMisdzfMisfεs−1εs+1−μis, A.5 so that we have: uis1−αs−βs=εsεs−1lnμiszdMisdzfMisfεs−1εs+1−μis. A.6 With the estimated residuals and parameters from the first regression in hand, we can take logs and run a second regression of the above adjusted residuals over sector fixed effects. We can thus estimate εs from the estimated sector fixed effects ψ^s from this second specification using the delta method, that is: ε^s=exp(ψ^s)exp(ψ^s)−1. A.7 Figure A12 shows the corresponding point estimates of εs ⁠, together with their 95% confidence intervals. Point estimates are remarkably close to 1, with the smallest value being equal to 1.01 and the largest being equal to 1.37. Evidence is therefore supportive of the hypothesis that foreign and domestically produced materials are not close substitutes, and in favour of the adoption of the Cobb–Douglas specification of the production function. We interpret this as validating our approach and the counterfactual analysis in subsection 7.2. Notes 1 The Occupied Palestinian Territories (OPT) are the West Bank (including East Jerusalem) and the Gaza Strip. The Second Intifada is a period of intensified violence which took place between 2000 and 2006. Section 1 provides extensive background information on the Israeli–Palestinian conflict and the Second Intifada in particular. 2 For a thoughtful discussion about the causes of the Second Intifada see Pressman (2003). 3 For a detailed description of the different periods of violence during the Second Intifada see Jaeger and Paserman (2008). 4 According to the Israeli Army, this system has been devised as a security measure to protect its citizens (both in Israel and inside Israeli settlements in the West Bank) from surges, or expected surges (Miaari and Sauer, 2011; IDF Military Advocate General, 2012.) 5 Within this same framework, the presence of distortions brings about dispersion in the marginal revenue product across firms. An alternative way of detecting conflict‐induced distortions would therefore be to test whether the distribution of marginal revenue products within the OPT is more dispersed in high conflict years relatively to low conflict years. This is indeed the approach followed by Hsieh and Klenow (2009) in studying misallocation of inputs in India and China relative to the United States. Notice that our approach of detecting market distortions by comparing input value ratios across firms does not involve any additional assumption. If anything, our reduced‐form approach avoids relying on other countries’ data to benchmark the structural parameters of the model, such as the elasticity of substitution σ, and the sector‐specific factor shares of the production function. 6 For more details on the study sample, variables definition and additional data on the aggregate level please refer to the online Appendix B. 7 These were established after the signing of the Oslo Accords, together with the division of the Israeli‐occupied territories into the West Bank and the Gaza Strip. Districts in the West Bank are: Jenin, Tubas, Tulkarm, Nablus, Qalqilya, Salfit, Ramallah and Al‐Bireh, Jericho, Jerusalem (including Israeli annexed East Jerusalem), Bethlehem and Hebron. Districts in the Gaza Strip are: North Gaza, Gaza, Deir al Balah, Khan Yunis and Rafah. 8 Data issues and sample derivation are described in detail in online Appendix B. 9 As for the sector of activity, 75% of the establishments in the sample operate in the following five sectors: Fabricated metal products, except machinery and equipment (22%); Furniture (15%); Food products and beverages (14%); Other non‐metallic mineral products (14%); wearing apparel and dressing, and dyeing of fur (12%). 10 In practice, we divide the fatalities count variable by the standard deviation of its district‐year distribution. 11 These fixed effects also average out systematic differences in factor prices across establishments in different years, districts, and sectors. This implies that results are robust to deviations from our conceptual framework (in which prices are assumed to be the same for all firms). We discuss the role of prices in more detail in subsection 5.1.1. 12 The number of clusters is above 50 in both dimensions, so that the cluster‐robust estimates of the variance‐covariance matrix of residuals are reliable. Table A3 in Appendix A reports the results when standard errors are clustered at the district level, and calculating using wild bootstrapping after 100 repetitions (Cameron et al., 2008). 13 Table A2 in Appendix A reports the corresponding estimates for all inputs, together with 95% confidence intervals. Consistent estimates of standard errors are derived accordingly from the standard error coefficient estimates in Table 1 using the delta method. 14 In subsection A.2 of Appendix A, we structurally estimate the elasticity of substitution between foreign and domestically produced materials in our data for each sector and find it to be remarkably close to one. See subsection 7.2 for a detailed discussion of this issue. 15 The Palestinian economy is highly dependent upon imported goods and services. During the Second Intifada, the total value of Palestinian imports was 6–8 times the total value of its export, with the negative balance of trade being equal to 40–50% of GDP at its current value. Israel was the main trade partner of the OPT during the period: Palestinian imports from Israel were around 70% of total value of imports while Palestinian exports to Israel were 90% of total value of exports. For Israel, however, trade with the OPT represents only a small share of foreign trade. See the online Appendix B for data sources and methodology. 16 The absence of meaningful changes in export composition also makes the possibility that changes in external demand drive the changes in firms input usage seem very unlikely, as the former does not correlate with conflict intensity. We discuss this issue further in subsection 5.2. 17 Data are available at: http://www.pcbs.gov.ps/Portals/_Rainbow/Documents/e-whpi-serise-2012.htm. 18 Unfortunately, the information on the value of inventory of materials at the beginning of the year is not available separately for foreign and domestically produced materials. 19 Since geographical coordinates of the firm's location are not available, we proxy the distance of the firm from the border gate using the district capital under the assumption that firms are more likely to be located close to the largest urban centre of the district. 20 We compute the per capita measure diving the district‐level umber of Palestinian fatalities by the total population of the district in the year. PCBS provides population data for each district and year on its website. 21 For additional details on this data set and our derived measure of proxy of conflict intensity, please refer to the online Appendix B. 22 Available sources of data on internal mobility restrictions are the Applied Research Institute of Jerusalem (ARIJ) and, since 2003, the United Nations Office for the Coordination of Humanitarian Affairs (UN‐OCHA). Using a combination of these data, Cali and Miaari (2013) find that internal checkpoints in the West Bank have a significant negative effect on employment, wages and days worked per month. Abrahams (2015) finds instead that the effects of mobility restrictions within the West Bank are spatially differentiated: core locations benefit while peripheral locations suffer in terms of employment. 23 B’Tselem also provides the data on closure days of the border between Israel and the OPT. 24 As discussed in subsection 5.1 (see Table A9 in Appendix A), the magnitude of the estimated distortions induced by the conflict does not change if we control for the differential effects that any year‐specific shock may have according to the distance from the border, as captured by the interactions of distance from the border with year fixed effects. 25 As in subsection 5.1, we proxy the distance of the firm from the gate using the district capital under the assumption that firms are more likely to be located close to the largest urban centre of the district. 26 This result is consistent with the findings in Etkes and Zimring (2015), which provide evidence of the negative impact of the Israeli‐imposed 2007 Gaza blockade on the economy of the Gaza Strip. In particular, they find that labour productivity decreased with the enforcement of the blockade and that labour reallocated away from trade‐oriented sectors. 27 Pre‐determined firm‐level variables included in the data set are: sales and employment in 2003, the year in which the firm began operating, whether the firm is female‐owned and its legal status. As for village level controls, we include the population in 1997, which is the last year before the Second Intifada for which data on population are available at the village level. 28 The procedure for clearing Palestinian goods through Israeli ports and controlled border crossings is long and extremely complicated. Israel requires that Palestinian trucks use the ‘back‐to‐back’ system according to which all goods need to be unloaded from and re‐loaded again onto trucks at checkpoints after the security check (World Bank, 2007). 29 Data show that only 18% of small enterprises benefit from bank loans, and the percentage of Gaza‐based medium enterprises accessing bank loans does not exceed 4%. By September 2002, private sector credit had fallen by 24% from its 2000 level. In particular, bank credit by sector shows that industry receiving smaller shares of total credit in 2001–2 than they had received before the beginning of the conflict (UNCTAD, 2004). 30 Figure A9 in Appendix A plots the value of Palestine GDP over time between 2000 and 2006, together with the total number of Palestinians fatalities caused by IDF. The evolution of real GDP is inversely related to conflict intensity as measured by the number of Palestinian fatalities. Figure A10 shows that similar inversely related trends can be observed between conflict intensity real aggregate output value as computed using the data from the Industry Survey. 31 In our data, 70% of establishments in the manufacture of tobacco products sector are located in the district of Jenin; 70% of establishments in the manufacture of leather products are located in Hebron, where 43% of the establishments in the manufacture of basic metals are also located. Figure A11 in Appendix A shows the evolution of PPI in these three sectors over time, together with the evolution of fatalities in the corresponding district. 32 Table A11 in Appendix A reports the corresponding factor share parameter estimates for each 2‐digit sector. The lack of a panel dimension in our firm‐level data set prevents us from using more sophisticated techniques to estimate the factor shares of the production function such as Olley and Pakes (1996). 33 Given the counterfactual input value ratio zdMsid/zfMsif ⁠, and being total value of materials equal to zdMsid+zfMsif ⁠, we can derive the counterfactual value of foreign and domestically produced materials separately by solving the corresponding system of two equations in two unknowns. 34 The results from such counterfactual analysis are essentially unchanged if, instead of using the estimates from the specification in row (d) of column (4) of Table 1, we implement an alternative specification where we interact the fatalities variable with sector fixed effects. Following the same procedure as above, we calculate that the value of output would have been 6.8% higher for the average firm in our sample and aggregate output value could have been as much as 14.2% higher. References Abadie , A. and Gardeazabal , J. ( 2003 ). ‘ The economic costs of conflict: a case study of the Basque Country’, American Economic Review, vol. 93 ( 1 ), pp. 113 – 32 . Google Scholar Crossref Search ADS Abrahams , A. ( 2015 ). Hard traveling: commuting costs and welfare in the Second Palestinian Uprising, mimeo, Harvard Kennedy School. Akkaya , S. , Fiess , N. , Kaminski , B. and Raballand , G. ( 2011 ). ’Fragility and conflict in Palestine’, in ( A.U.S.‐P. Wim Naudé and M. McGillivray , eds.), Fragile States Causes, Costs, and Responses, pp. 111 – 32 , vol. 1 , Oxford : Oxford University Press . Alesina , A. , Özler , S. , Roubini , N. and Swagel , P. ( 1996 ). ‘ Political instability and economic growth’, Journal of Economic Growth, vol. 1 ( 2 ), pp. 189 – 211 . Google Scholar Crossref Search ADS Amiti , M. and Konings , J. ( 2007 ). ‘ Trade liberalization, intermediate inputs, and productivity: evidence from Indonesia’, American Economic Review, vol. 97 ( 5 ), pp. 1611 – 38 . Google Scholar Crossref Search ADS Asker , J. , Collard‐Wexler , A. and Loecker , J.D. ( 2014 ). ‘ Dynamic inputs and resource (mis)allocation’, Journal of Political Economy, vol. 122 ( 5 ), pp. 1013 – 63 . Google Scholar Crossref Search ADS Blattman , C. and Miguel , E. ( 2010 ). ‘ Civil War’, Journal of Economic Literature, vol. 48 ( 1 ), pp. 3 – 57 . Google Scholar Crossref Search ADS Bloom , N. ( 2009 ). ‘ The impact of uncertainty shocks’, Econometrica, vol. 77 ( 3 ), pp. 623 – 85 . Google Scholar Crossref Search ADS Boehm , C. , Flaaen , A. and Pandalai‐Nayar , N. ( 2015 ). ‘ Input linkages and the transmission of shocks: firm‐level evidence from the 2011 Tōhoku Earthquake ’, mimeo, University of Texas at Austin. Brück , T. , Di Maio , M. and Miaari , S.H. ( 2014 ). ‘ Learning the HardWay: the effect of violent conflict on student academic achievement ’, Institute for the Study of Labor (IZA). B’Tselem ( 2007 ). ‘ Annual report: human rights in the occupied territories ’, B’Tselem. Cali , M. and Miaari , S.H. ( 2013 ). ‘ The labor market impact of mobility restrictions: evidence from the West Bank ’, The World Bank. Camacho , A. and Rodriguez , C. ( 2013 ). ‘ Firm exit and armed conflict in Colombia’, Journal of Conflict Resolution, vol. 57 ( 1 ), pp. 89 – 116 . Google Scholar Crossref Search ADS Cameron , A.C. , Gelbach , J.B. and Miller , D.L. ( 2008 ). ‘ Bootstrap‐based improvements for inference with clustered errors’, The Review of Economics and Statistics, vol. 90 ( 3 ), pp. 414 – 27 . Google Scholar Crossref Search ADS Cerra , V. and Saxena , S.C. ( 2008 ). ‘ Growth dynamics: the myth of economic recovery’, American Economic Review, vol. 98 ( 1 ), pp. 439 – 457 . Google Scholar Crossref Search ADS Chen , S. , Loayza , N.V. and Reynal‐Querol , M. ( 2008 ). ‘ The aftermath of civil War’, World Bank Economic Review, vol. 22 ( 1 ), pp. 63 – 85 . Google Scholar Crossref Search ADS Collier , P. and Duponchel , M. ( 2013 ). ‘ The economic legacy of civil war: firm‐level evidence from Sierra Leone’, Journal of Conflict Resolution, vol. 57 ( 1 ), pp. 65 – 88 . Google Scholar Crossref Search ADS Collier , P. , Elliott , V. , Hegre , H. , Hoeffler , A. , Reynal‐Querol , M. and Sambanis , N. ( 2003 ). Breaking the Conflict Trap: Civil War and Development Policy (Policy Research Reports), Washington, DC : World Bank; Oxford University Press . David , J.M. , Hopenhayn , H.A. and Venkateswaran , V. ( 2016 ). ‘ Information, misallocation, and aggregate productivity’, Quarterly Journal of Economics, vol. 131 ( 2 ), pp. 943 – 1005 . Google Scholar Crossref Search ADS Di Maio , M. and Nandi , T.K. ( 2013 ). ‘ The effect of the Israeli‐Palestinian conflict on child labor and school attendance in the West Bank’, Journal of Development Economics, vol. 100 ( 1 ), pp. 107 – 116 . Google Scholar Crossref Search ADS Dube , O. and Vargas , J.F. ( 2013 ). ‘ Commodity price shocks and civil conflict: evidence from Colombia’, Review of Economic Studies, vol. 80 ( 4 ), pp. 1384 – 421 . Google Scholar Crossref Search ADS Eckstein , Z. and Tsiddon , D. ( 2004 ). ‘ Macroeconomic consequences of terror: theory and the case of Israel’, Journal of Monetary Economics, vol. 51 ( 5 ), pp. 971 – 1002 . Google Scholar Crossref Search ADS Ethier , W.J. ( 1982 ). ‘ National and international returns to scale in the modern theory of international trade’, American Economic Review, vol. 72 ( 3 ), pp. 389 – 405 . Etkes , H. and Zimring , A. ( 2015 ). ‘ When trade stops: lessons from the Gaza blockade 2007–2010’, Journal of International Economics, vol. 95 ( 1 ), pp. 16 – 27 . Google Scholar Crossref Search ADS Garicano , L. , Lelarge , C. and Van Reenen , J. ( 2016 ). ‘ Firm size distortions and the productivity distribution: evidence from france’, American Economic Review, vol. 106 ( 11 ), pp. 3439 – 79 . Google Scholar Crossref Search ADS Guidolin , M. and La Ferrara , E. ( 2007 ). ‘ Diamonds are forever, wars are not: is conflict bad for private firms?’, American Economic Review, vol. 97 ( 5 ), pp. 1978 – 93 . Google Scholar Crossref Search ADS Guner , N. , Ventura , G. and Xu , Y. ( 2008 ). ‘ Macroeconomic implications of size‐dependent policies’, Review of Economic Dynamics, vol. 11 ( 4 ), pp. 721 – 44 . Google Scholar Crossref Search ADS Hjort , J. ( 2014 ). ‘ Ethnic divisions and production in firms’, Quarterly Journal of Economics, vol. 129 ( 4 ), pp. 1899 – 946 . Google Scholar Crossref Search ADS Hsieh , C.T. and Klenow , P.J. ( 2009 ). ‘ Misallocation and manufacturing TFP in China and India’, Quarterly Journal of Economics, vol. 124 ( 4 ), pp. 1403 – 48 . Google Scholar Crossref Search ADS IDF Military Advocate General ( 2012 ). ‘ Legal issues in Judea and Samaria: movement and access restrictions ’, Israel Defence Forces. Jaeger , D.A. and Paserman , M.D. ( 2008 ). ‘ The cycle of violence? An empirical analysis of fatalities in the Palestinian‐Israeli conflict’, American Economic Review, vol. 98 ( 4 ), pp. 591 – 1604 . Google Scholar Crossref Search ADS Kasahara , H. and Rodrigue , J. ( 2008 ). ‘ Does the use of imported intermediates increase productivity? Plant‐level evidence’, Journal of Development Economics, vol. 87 ( 1 ), pp. 106 – 18 . Google Scholar Crossref Search ADS Klapper , L. , Richmond , C. and Tran , T. ( 2013 ). ‘ Civil conflict and firm performance. Evidence from Côte d’Ivoire ’, The World Bank. Ksoll , C. , Macchiavello , R. and Morjaria , A. ( 2015 ). ‘ Guns and roses: flower exports and electoral violence in Kenya ’, mimeo, Northwestern University. Kugler , M. and Verhoogen , E. ( 2012 ). ‘ Prices, plant size, and product quality’, Review of Economic Studies, vol. 79 ( 1 ), pp. 307 – 39 . Google Scholar Crossref Search ADS Mansour , H. and Rees , D.I. ( 2012 ). ‘ Armed conflict and birth weight: evidence from the Al‐Aqsa Intifada’, Journal of Development Economics, vol. 99 ( 1 ), pp. 190 – 9 . Google Scholar Crossref Search ADS Miaari , S. and Sauer , R. ( 2011 ). ‘ The labor market costs of conflict: closures, foreign workers, and Palestinian employment and earnings’, Review of Economics of the Household, vol. 9 ( 1 ), pp. 129 – 48 . Google Scholar Crossref Search ADS Midrigan , V. and Xu , D.Y. ( 2014 ). ‘ Finance and misallocation: evidence from plant‐level data’, American Economic Review, vol. 104 ( 2 ), pp. 422 – 58 . Google Scholar Crossref Search ADS Olley , G.S. and Pakes , A. ( 1996 ). ‘ The dynamics of productivity in the telecommunications equipment industry’, Econometrica, vol. 64 ( 6 ), pp. 1263 – 97 . Google Scholar Crossref Search ADS PCBS ( 2007 ). ‘ Industry Survey 1999–2006 ’, Available at www.pcbs.gov.ps (last accessed: 10 October 2013). Pressman , J. ( 2003 ). ‘ The second Intifada: background and causes of the Israeli Palestinian conflict’, Journal of Conflict Studies, vol. 23 ( 2 ), pp. 114 – 48 . Restuccia , D. and Rogerson , R. ( 2008 ). ‘ Policy distortions and aggregate productivity with heterogeneous establishments’, Review of Economic Dynamics, vol. 11 ( 4 ), pp. 707 – 20 . Google Scholar Crossref Search ADS Schor , A. ( 2004 ). ‘ Heterogeneous productivity response to tariff reduction. Evidence from brazilian manufacturing firms’, Journal of Development Economics, vol. 75 ( 2 ), pp. 373 – 96 . Google Scholar Crossref Search ADS Service Group ( 2002 ). ‘ Economic feasibility study for the Gaza Construction Materials Logistics Facility (CMLF) ’, USAID Mission to the West Bank and Gaza. Shilliday , A. and Lautenschlager , J. ( 2012 ). ‘ Data for a global ICEWS and ongoing research ’, 2nd International Conference on Cross‐Cultural Decision Making: Focus 2012. Singh , P. ( 2013 ). ‘ Impact of terrorism on investment decisions of farmers: evidence from the Punjab Insurgency’, Journal of Conflict Resolution, vol. 57 ( 1 ), pp. 143 – 68 . Google Scholar Crossref Search ADS Sletten , P. and Pedersen , J. ( 2003 ). ‘ Copying with conflict. Palestinian communities two years into the Intifada ’, FAFO Report 408, Oslo, Norway. Topalova , P. and Khandelwal , A. ( 2011 ). ‘ Trade liberalization and firm productivity: the case of India’, The Review of Economics and Statistics, vol. 93 ( 3 ), pp. 995 – 1009 . Google Scholar Crossref Search ADS UNCTAD ( 2004 ). ‘Palestinian small and medium‐sized enterprises: dynamics and contribution to development, UNCTAD/GDS/APP/2004/1. UNCTAD ( 2006 ). ‘ The palestinian war‐torn economy: aid, development and state formation ’, UNCTAD/GDS/APP/2006/1, United Nations Publications. UNWFP ( 2009 ). ‘ The impact of closure and high food prices on performance of imported staple foods and vegetable and fruits market in the OPT’, United Nations World Food Programme. Wasmer , E. and Weil , P. ( 2004 ). ‘ The macroeconomics of labor and credit market imperfections’, American Economic Review, vol. 94 ( 4 ), pp. 944 – 63 . Google Scholar Crossref Search ADS World Bank ( 2004 ). Four Years. Intifada, Closures and Palestinian Economic Crisis, Washington, DC : World Bank . World Bank ( 2006a ). ‘ Enterprise Analysis Unit – World Bank Group ’, Available at www.enterprisesurveys.org (last accessed: 22 January 2013). World Bank ( 2006b ). West Bank and Gaza Country. Economic Memorandum Growth in West Bank and Gaza: Opportunities and Constraints, World Bank . World Bank ( 2007 ). ‘ Movement and access restrictions in the west bank: uncertainty and inefficiency in the palestinian economy ’, World Bank Technical Team. World Bank ( 2008 ). ‘ Implementing the palestinian reform and development agenda ’, Economic monitoring report to the Ad Hoc Liaison Committee. World Bank ( 2011 ). World Development Report. Conflict, Security, and Development, Washington, DC : World Bank . World Bank ( 2013 ). ‘ Enterprise Analysis Unit – World Bank Group ’, www.enterprisesurveys.org. (last accessed: 20 June 2015) © 2017 Royal Economic Society
TI - Making Do With What You Have: Conflict, Input Misallocation and Firm Performance
JF - The Economic Journal
DO - 10.1111/ecoj.12518
DA - 2018-11-01
UR - https://www.deepdyve.com/lp/oxford-university-press/making-do-with-what-you-have-conflict-input-misallocation-and-firm-Sen7bN8zy5
SP - 2559
VL - 128
IS - 615
DP - DeepDyve
ER -