diff --git a/pandas_questions.py b/pandas_questions.py
index 262ad29..72dcb58 100644
--- a/pandas_questions.py
+++ b/pandas_questions.py
@@ -14,10 +14,10 @@
 
 
 def load_data():
-    """Load data from the CSV files referundum/regions/departments."""
-    referendum = pd.DataFrame({})
-    regions = pd.DataFrame({})
-    departments = pd.DataFrame({})
+    """Load data from the CSV files referendum/regions/departments."""
+    referendum = pd.read_csv("./data/referendum.csv", sep=';')
+    regions = pd.read_csv("./data/regions.csv")
+    departments = pd.read_csv("./data/departments.csv")
 
     return referendum, regions, departments
 
@@ -28,8 +28,14 @@ def merge_regions_and_departments(regions, departments):
     The columns in the final DataFrame should be:
     ['code_reg', 'name_reg', 'code_dep', 'name_dep']
     """
-
-    return pd.DataFrame({})
+    merged = pd.merge(regions, departments,
+                      left_on='code', right_on='region_code')
+    merged.columns = ['id_x', 'code_reg',
+                      'name_reg', 'slug_x',
+                      'id_y', 'region_code',
+                      'code_dep', 'name_dep',
+                      'slug_y']
+    return merged[['code_reg', 'name_reg', 'code_dep', 'name_dep']]
 
 
 def merge_referendum_and_areas(referendum, regions_and_departments):
@@ -39,7 +45,14 @@ def merge_referendum_and_areas(referendum, regions_and_departments):
     french living abroad.
     """
 
-    return pd.DataFrame({})
+    r_and_d = regions_and_departments
+    dc = "Department code"
+    referendum[dc] = referendum[dc].astype(str).str.zfill(2)
+    r_and_d["code_dep"] = r_and_d["code_dep"].astype(str).str.zfill(2)
+    referendum_and_areas = referendum.merge(r_and_d,
+                                            left_on="Department code",
+                                            right_on="code_dep")
+    return referendum_and_areas
 
 
 def compute_referendum_result_by_regions(referendum_and_areas):
@@ -49,7 +62,12 @@ def compute_referendum_result_by_regions(referendum_and_areas):
     ['name_reg', 'Registered', 'Abstentions', 'Null', 'Choice A', 'Choice B']
     """
 
-    return pd.DataFrame({})
+    grouped = referendum_and_areas.groupby(["code_reg", "name_reg"]).sum()
+    grouped = grouped.reset_index()
+    grouped = grouped.set_index("code_reg")[[
+        "name_reg", "Registered", "Abstentions",
+        "Null", "Choice A", "Choice B"]]
+    return grouped
 
 
 def plot_referendum_map(referendum_result_by_regions):
@@ -62,7 +80,17 @@ def plot_referendum_map(referendum_result_by_regions):
     * Return a gpd.GeoDataFrame with a column 'ratio' containing the results.
     """
 
-    return gpd.GeoDataFrame({})
+    geodatas = gpd.read_file("./data/regions.geojson")
+    merged = pd.merge(geodatas,
+                      referendum_result_by_regions,
+                      left_on="code",
+                      right_on="code_reg")
+    all_expressed = merged["Choice A"] + merged["Choice B"]
+    merged['ratio'] = merged['Choice A'] / all_expressed
+    merged = gpd.GeoDataFrame(merged)
+    merged.plot(column="ratio")
+    plt.show()
+    return merged
 
 
 if __name__ == "__main__":