chore(data): Switch to -1/0/1 coding for impact direction

So we can also express simply no relation between the variables in
any direction.

02-data/processed/relevant/Clark2019.yml

@@ -32,7 +32,7 @@ observation:
     measures: employment probability; hours worked
     findings: subsidy increased employment probability (8.5ppts) for poor married mothers
     channels: increased ability to work through lower childcare burden
-    direction: 1 # 0 neg / 1 pos
+    direction: 1 
     significance: 2 # 0 nsg / 1 msg / 2 sg
   - intervention: childcare subsidy
     institutional: 0
@@ -44,7 +44,7 @@ observation:
     measures: hours worked
     findings: subsidy decreased hours worked without decreasing income for single mothers
     channels: allows shifting to jobs with more regular hours
-    direction: 0 # 0 neg / 1 pos
+    direction: -1 # -1 neg
     significance: 2 # 0 nsg / 1 msg / 2 sg
 
 notes:

02-data/supplementary/extraction_template.yml

@@ -32,7 +32,7 @@ observation:
     measures:
     findings:
     channels:
-    direction: # 0 neg / 1 pos
+    direction: # -1 neg / 0 none / 1 pos
     significance: # 0 nsg / 1 msg / 2 sg
 
 notes: