• No results found

III. STOCK MARKET REACTION TO FIRM WITHDRAWAL FROM STATE SPONSOR

4. SHORT-TERM EFFECT OF THE TOOL’S PUBLICATION

4.3. Short-term effect: robustness checks

4.3.2. Additional robustness checks

In addition to the portfolio level analysis presented in the preceding chapter, we conduct numerous additional robustness checks. First, we check our results for robustness on the US equity market. For reasons declared in Section 3, we believe that respective home markets are better suited for our investigation. Nevertheless, not least for reasons of arbitrage, similar findings should be observable on the US equity market. US results reveal no notable differ- ences to the results obtained from the home exchange analysis. Corresponding results are presented in Table IV-10, in Table IV-11, and in Table IV-12. Results for US firms on the US market differ from results for US firms on the (US) home market, because the factors used for the US market analysis were downloaded from the data library on Kenneth French’s web- site (http://mba.tuck.dartmouth.edu/pages/faculty/ken.french/data_library.html), while factors for the home market were taken from DataStream.

Table IV-10

Pre Event Period Short Event Period Long Event Period Post Event Period Panel A H0: γi = 0 0.100 1) 0.122 1) 0.087 1) −0.282 1) (Wald test) (0.46) (0.09) (0.35) (5.83)** Panel B H0: Median γi = 0 0.001 2) 0.002 2) 0.001 2) −0.003 2) Coef. positive/negative 56/25 46/35 49/32 15/66

(Wilcoxon signed-rank test) (3.11)*** (1.02) (0.89) (−5.64)***

Observations 363 363 363 363

Hypothesis testing: full sample stock returns (US market)

This table displays hypotheses about γi, the event-dummy coefficient as estimated in regression equation (5) adapted for the US market. The sample encompasses a total of 81 firms. Panel A provides results for the hypothesis that the sum over all γiis equal to zero (hypothesis A). Panel B reports results for the hypothesis that the median γiis equal to zero (hypothesis B). The estimation window ranges from 07/2006 to 12/2007. Chi2 values for Wald tests and z-values for Wilcoxon signed-rank tests are in parentheses. 1) Displayed is the sum of coefficient estimates, 2) Displayed is the median coefficient.

* p<0.10, ** p<0.05, *** p<0.01

N 1 i

The SEC’s Publication of an Online Tool for Detecting Firms Doing Business in Countries Designat- ed as State Sponsors of Terrorism: Short-Term Investor Reaction and Long-Term Consequences

Table IV-11

Table IV-12

Second, we check robustness to potentially erroneous allegations of the tool. When the tool went online, some firms mentioned were demonstrably inactive in each of the designated countries. Other firms’ activities were humanitarian in nature and could therefore not be re- garded as sinful. Therefore, we also build subsamples according to the firms’ involvement in

Pre Event Period Short Event Period Long Event Period Post Event Period Panel A H0: γi = 0 0.059 1) −0.027 1) 0.066 1) −0.115 1) (Wald test) (0.24) ( 0.01) (0.30) (1.45) Panel B H0: Median γi = 0 0.002 1) −0.005 2) 0.000 2) −0.003 2) Coef. positive/negative 22/7 11/18 16/13 8/21

(Wilcoxon signed-rank test) (2.20) (−1.14) (0.64) (−2.89)***

Observations 363 363 363 363

This table displays hypotheses about γi, the event-dummy coefficient as estimated in regression equation (5) adapted for the US market. The sample encompasses a total of 29 firms. Panel A provides results for the hypothesis that the sum over all γiis equal to zero (hypothesis A). Panel B reports results for the hypothesis that the median γiis equal to zero (hypothesis B). The estimation window ranges from 07/2006 to 12/2007. Chi2 values for Wald tests and z-values for Wilcoxon signed-rank tests are in parentheses. 1) Displayed is the sum of coefficient estimates, 2) Displayed is the median coefficient.

* p<0.10, ** p<0.05, *** p<0.01

Hypothesis testing: US sample stock returns (US market)

N 1 i

Pre Event Period Short Event Period Long Event Period Post Event Period Panel A H0: γi = 0 −0.015 1) 0.064 1) 0.109 1) −0.162 1) (Wald test) (0.02) (0.07) (1.65) (5.60)** Panel B H0: Median γi = 0 0.001 2) 0.004 2) 0.001 2) −0.003 2) Coef. positive/negative 35/15 31/19 34/16 9/41

(Wilcoxon signed-rank test) (2.34)** (2.16)** (2.50)** (−4.56)***

Observations 363 363 363 363

This table displays hypotheses about γi, the event-dummy coefficient as estimated in regression equation (5) adapted for the US market. The sample encompasses a total of 50 firms. Panel A provides results for the hypothesis that the sum over all γiis equal to zero (hypothesis A). Panel B reports results for the hypothesis that the median γiis equal to zero (hypothesis B). The estimation window ranges from 07/2006 to 12/2007. Chi2 values for Wald tests and z-values for Wilcoxon signed-rank tests are in parentheses. 1) Displayed is the sum of coefficient estimates, 2) Displayed is the median coefficient.

* p<0.10, ** p<0.05, *** p<0.01

Hypothesis testing: Iran operations sample stock returns (US market)

N 1 i

The SEC’s Publication of an Online Tool for Detecting Firms Doing Business in Countries Designat- ed as State Sponsors of Terrorism: Short-Term Investor Reaction and Long-Term Consequences

the designated countries. Results displayed in Table IV-13 and Table IV-14 reveal no strong- er negative effect than for the full sample of firms. After all, investors do not seem to have based their decisions on the firms’ respective involvement.

Table IV-13

Table IV-14

Pre Event Period Short Event Period Long Event Period Post Event Period Panel A H0: γi = 0 0.049 1) 0.082 1) 0.087 1) −0.194 1) (Wald test) (0.18) (0.04) (0.50) (3.61)* Panel B H0: Median γi = 0 0.001 2) 0.000 2) 0.000 2) −0.001 2) Coef. positive/negative 43/33 39/37 44/32 20/56

(Wilcoxon signed-rank test) (1.97)** (0.31) (0.99) (−4.37)***

Observations 363 363 363 363

Hypothesis testing: involvement sample stock returns

This table displays hypotheses about γi, the event-dummy coefficient as estimated in regression equation (5). The sample encompasses a total of 76 firms. Firms that had demonstrably no business in the designated countries by the time the tool went online are excluded from the sample. Panel A provides results for the hypothesis that the sum over all γiis equal to zero (hypothesis A). Panel B reports results for the hypothesis that the median γi is equal to zero (hypothesis B). The estimation window ranges from 07/2006 to 12/2007. Chi2 values for Wald tests and z-values for Wilcoxon signed-rank test are in parentheses. 1) Displayed is the sum of coefficient estimates, 2) Displayed is the median coefficient.

* p<0.10, ** p<0.05, *** p<0.01

N 1 i

Pre Event Period Short Event Period Long Event Period Post Event Period Panel A H0: γi = 0 0.075 1) 0.148 1) 0.078 1) −0.194 1) (Wald test) (0.47) (0.18) (0.48) (4.24)** Panel B H0: Median γi = 0 0.001 2) 0.002 2) 0.000 2) −0.002 2) Coef. positive/negative 40/28 39/29 42/26 16/52

(Wilcoxon signed-rank test) (2.60)*** (0.97) (1.57) (−4.46)***

Observations 363 363 363 363

Hypothesis testing: material involvement sample stock returns

This table displays hypotheses about γi, the event-dummy coefficient as estimated in regression equation (5). The sample encompasses a total of 68 firms. Excluded are the following types of firms: (i) Firms that had demonstrably no business in the designated countries by the time the tool went online. (ii) Firms that had commited not to engage in any new business or were solely engaged in humanitarian activities by the time the tool went online. Panel A provides results for the hypothesis that the sum over all γiis equal to zero (hypothesis A). Panel B reports results for the hypothesis that the median γiis equal to zero (hypothesis B). The estimation window ranges from 07/2006 to 12/2007. Chi2values for Wald tests and z-values for Wilcoxon signed-rank test are in parentheses. 1) Displayed is the sum of coefficient estimates, 2) Displayed is the median coefficient. * p<0.10, ** p<0.05, *** p<0.01

N 1 i

The SEC’s Publication of an Online Tool for Detecting Firms Doing Business in Countries Designat- ed as State Sponsors of Terrorism: Short-Term Investor Reaction and Long-Term Consequences

A final robustness check concerns the analysis of control groups. To this end, we match two different control samples. One is a size, country and industry matched global sample of firms that were not active in a country designated as a State Sponsor of Terrorism by the time the SEC’s tool went online. The other control group consists of firms that did operate in one or more of the designated countries but were not mentioned in the SEC tool. Data for the lat- ter control sample is provided by IW Financial, which distributes the so-called Global Securi- ty Risk Monitor (GSRM) databases maintained by the Conflict Security Advisory Group (CSAG). As displayed in Table IV-15 and in Table IV-16, none of the analyses of short-term effects yield statistically significant results. Hence, the negative post event period effect is unique to firms active in countries designated as State Sponsors of Terrorism. Furthermore, it is stigmatisation by means of the SEC tool that induced the observed negative announcement effects. If investors’ attention to all firms active in the designated countries was aroused by the tool’s publication, then the GSRM sample should have displayed a negative reaction, as well.

Table IV-15

Pre Event Period Short Event Period Long Event Period Post Event Period Panel A H0: γi = 0 −0.100 1) −0.049 1) 0.062 1) −0.060 1) (Wald test) (0.89) (0.03) (0.37) (0.53) Panel B H0: Median γi = 0 −0.001 2) −0.000 2) 0.001 2) −0.000 2) Coef. positive/negative 37/44 35/46 43/38 35/46

(Wilcoxon signed-rank test) (−1.69) (−0.45) (0.21) (−1.46)

Observations 363 363 363 363

This table displays hypotheses about γi, the event-dummy coefficient as estimated in regression equation (5). The sample encompasses a total of 81 firms. Panel A provides results for the hypothesis that the sum over all γiis equal to zero (hypothesis A). Panel B reports results for the hypothesis that the median γi is equal to zero (hypothesis B). The estimation window ranges from 07/2006 to 12/2007. Chi2 values for Wald tests and z-values for Wilcoxon signed-rank test are in parentheses.

1) Displayed is the sum of coefficient estimates, 2) Displayed is the median coefficient. * p<0.10, ** p<0.05, *** p<0.01

Hypothesis testing: terror-free control sample

N 1 i

The SEC’s Publication of an Online Tool for Detecting Firms Doing Business in Countries Designat- ed as State Sponsors of Terrorism: Short-Term Investor Reaction and Long-Term Consequences

Table IV-16

5. LONG-TERM EFFECT OF THE TOOL’S PUBLICATION