Analysis of King County bike-related citation records¶

kc_citations = pd.read_excel(data_dir + '2021-10-18 - compiled King County bike citation records.xlsx').drop(columns=['Unnamed: 0'])
kc_citations.head(3)

display(cities)

# recent average annual citations by city
city_contributions = cities['Average annual citations (2015-2019 only)'].sort_values().copy()
city_contributions = city_contributions[city_contributions > 0]
labels = [name if 100 * city_contributions[name] / city_contributions.sum() > 0.8 else '' for name in city_contributions.index]
def autopct(percent): return ('%.1f%%' % percent) if percent > 0.8 else ''

plt.figure(figsize=(7,5),facecolor='w')
with plt.style.context({"axes.prop_cycle":plt.cycler('color',roll(plt.cm.tab20c.colors[2:],4,axis=0)**0.8)}):
    wedges, labels, autopct = plt.pie(city_contributions,explode=tile(0.02,len(city_contributions)),
                                      labels=labels,labeldistance=1.10,autopct=autopct,pctdistance=0.75)
for lab in labels: lab.set_fontsize(9)
for pct in autopct: pct.set_fontsize(7)
plt.title('Contributions to total King County helmet citations\nby jurisdiction (2015-2019)');
plt.savefig(current_results_dir + 'king_co_helmet_citations_by_jurisdiction.pdf')

# helmet vs. non-helmet citations in King County outside of Seattle
kc_subset = kc_citations[logical_and(kc_citations['City'] != 'Seattle',kc_citations['Violation Date'].dt.year >= 2003)]
helmet_mask = kc_subset['Law Description'].str.contains('Bicycle Helmet Required')
helmet_percent = 100 * sum(helmet_mask) / len(helmet_mask)
nonhelmet_percent = 100 - helmet_percent
def autopct(percent): return ('%.1f%%' % percent)

plt.figure(figsize=(7,5),facecolor='w')
wedges, labels, autopct = plt.pie([helmet_percent,nonhelmet_percent],explode=tile(0.01,2),startangle=45,
                                  labels=['Helmet\ncitations','Non-helmet\ncitations'],
                                  colors=['steelblue','navy'],
                                  labeldistance=1.10,autopct=autopct,pctdistance=0.75)
autopct[1].set_color('w')
for lab in labels: lab.set_fontsize(11)
for pct in autopct: pct.set_fontsize(10)
plt.title('Citations issued to bicyclists by law enforcement\nin King County excluding Seattle (2003-present)');
plt.text(0.5,0.84,'$n$ = {0} citations'.format(len(helmet_mask)),fontsize=10,
         horizontalalignment='center',transform=plt.gcf().transFigure);
plt.savefig(current_results_dir + 'king_co_bike_citations_type.pdf')

# years of available data
plt.figure(figsize=(7,5),facecolor='w')
plt.hist(kc_citations['Violation Date'].dt.year,bins=arange(1999.5,2022.5),color='k',rwidth=0.85,zorder=1,
         label='All bike-related citations')
plt.hist(kc_citations.loc[kc_citations['Law Description'].str.contains('Bicycle Helmet Required'),'Violation Date'].dt.year,
         bins=arange(1999.5,2022.5),color='0.7',rwidth=0.65,zorder=2,label='Helmet citations only')
plt.legend(loc='upper left',frameon=False)
plt.text(0.5,0.9,'Available bike citation records for King County (including Seattle)',size=14,
         horizontalalignment='center',transform=plt.gcf().transFigure)
plt.title('(Note: data availability is affected by both citation rates and data retention practices)',
          style='italic',size=8,y=1.025)
plt.ylabel('Count');
plt.tight_layout(rect=[0,0.03,1,0.90])
plt.savefig(current_results_dir + 'king_co_bike_citations_by_year.pdf')

# per-capita helmet citation rates by jurisdiction
cities_subset = cities[['Total helmet citations in records',
                        'Annual citation rate per 100k people (all years)',
                        'Annual citation rate per 100k people (2015-2019 only)']].copy()
cities_without_enforcement = cities_subset[cities_subset['Total helmet citations in records'] == 0].index.values
cities_subset = cities_subset[cities_subset['Total helmet citations in records'] != 0]
cities_subset.sort_values(by='Annual citation rate per 100k people (all years)',inplace=True)

plt.figure(figsize=(11,7.5),facecolor='w')
plt.scatter(cities_subset['Annual citation rate per 100k people (all years)'],range(len(cities_subset)),
            marker='o',s=75,c='k',zorder=3,alpha=0.8,label='All years with citation records')
plt.scatter(cities_subset['Annual citation rate per 100k people (2015-2019 only)'],range(len(cities_subset)),
            marker='x',s=50,c='k',zorder=3,alpha=0.6,label='2015-2019 only')
plt.legend(bbox_to_anchor=(1.0,0.88),loc='upper right',frameon=True,labelspacing=1.0)
plt.yticks(ticks=range(len(cities_subset)),
           labels=[city + ' ($n$ = {0})'.format(cities_subset.loc[city,'Total helmet citations in records']) \
                   for city in cities_subset.index])
plt.xlim([-5,None])
plt.ylim([-0.5,len(cities_subset)-0.5])
plt.grid(alpha=0.5,zorder=1)
plt.gca().spines['left'].set_position(('outward',5))
plt.gca().spines['top'].set_position(('outward',5))
plt.gca().spines['left'].set_visible(False)
plt.gca().spines['bottom'].set_visible(False)
plt.gca().spines['right'].set_visible(False)
plt.gca().tick_params(axis='y',length=0)
plt.gca().xaxis.set_ticks_position("top")
plt.gca().xaxis.set_label_position("top")
plt.xticks(ticks=arange(0,100,10))
plt.xlabel('Annual citation rate per 100k people (2021 WA OFM population estimates)',labelpad=10)
plt.title('Per capita rates of helmet tickets issued in King County jurisdictions             ',size=14,y=1.10)
cities_without_enforcement[2] = '\n' + cities_without_enforcement[2]
cities_without_enforcement[10] = '\n' + cities_without_enforcement[10]
plt.text(0.4,0.085,'$^{*}$ ' + 'Cities/towns with no records of bicycle helmet enforcement: {0}.'\
         .format(', '.join(cities_without_enforcement).rstrip(',')),
         size=9,transform=plt.gcf().transFigure,style='italic')
plt.tight_layout()
plt.savefig(current_results_dir + 'king_co_helmet_citation_rates_by_jurisdiction.pdf')

# helmet citation disparities for Black defendants
cities_subset = cities[['Total helmet citations in records',
                        'Citations - White (%)','Census (2015-2019) - White (%)',
                        'Citations - Black (%)','Census (2015-2019) - Black (%)']].copy().drop('Unincorporated King County')
cities_subset = cities_subset[cities_subset['Total helmet citations in records'] != 0]
cities_subset['Disparity in citations issued to Black cyclists'] \
    = (cities_subset['Citations - Black (%)'] / cities_subset['Census (2015-2019) - Black (%)'])
disparities = cities_subset['Disparity in citations issued to Black cyclists']\
    [cities_subset['Total helmet citations in records'] > 10]
disparities.sort_values(inplace=True)

plt.figure(figsize=(7.5,6.5),facecolor='w')
plt.scatter(cities_subset['Census (2015-2019) - Black (%)'][disparities.index],range(len(disparities)),
            marker='o',s=75,c='k',zorder=3,label='Percent of local population\n(U.S. Census ACS, 2015-2019)')
plt.scatter(cities_subset['Citations - Black (%)'][disparities.index],range(len(disparities)),
            marker='d',edgecolors='k',s=75,zorder=3,
            c=disparities,cmap='RdBu_r',norm=mcolors.LogNorm(vmin=1.35e-1,vmax=0.7e1))
plt.scatter(NaN,NaN,marker='d',c='k',edgecolors='k',s=75,label='Percent of helmet citations')
for c_idx, city in enumerate(disparities.index):
    if cities_subset.loc[city,'Census (2015-2019) - Black (%)'] > cities_subset.loc[city,'Citations - Black (%)']:
        d = 0.1
    else:
        d = -0.1
    plt.annotate(text=None,xy=(cities_subset.loc[city,'Census (2015-2019) - Black (%)'],c_idx),
                 xytext=(cities_subset.loc[city,'Citations - Black (%)']+d,c_idx),
                 arrowprops=dict(arrowstyle='<-, head_width=0.3',linewidth=1.5))
    plt.text(0.5*(cities_subset.loc[city,'Census (2015-2019) - Black (%)'] + 
                  cities_subset.loc[city,'Citations - Black (%)']),c_idx + 0.23,
             '{0:.1f}x '.format(disparities.loc[city]),weight='bold',horizontalalignment='center')
plt.legend(bbox_to_anchor=(0.05,1.20),loc='upper left',ncol=2,handletextpad=0.4,frameon=False)
plt.yticks(ticks=range(len(disparities)),
           labels=[city + '\n($n$ = {0})'.format(cities_subset.loc[city,'Total helmet citations in records']) \
                   for city in disparities.index])
plt.xlim([0,None])
plt.ylim([-0.5,len(disparities)-0.25])
plt.grid(alpha=0.5,zorder=1)
plt.gca().spines['left'].set_position(('outward',5))
plt.gca().spines['top'].set_position(('outward',5))
plt.gca().spines['left'].set_visible(False)
plt.gca().spines['bottom'].set_visible(False)
plt.gca().spines['right'].set_visible(False)
plt.gca().tick_params(axis='y',length=0)
plt.gca().xaxis.set_ticks_position("top")
plt.gca().xaxis.set_label_position("top")
plt.xticks(ticks=arange(0,25,5))
plt.gca().xaxis.set_major_formatter(mtick.PercentFormatter(decimals=0))
# plt.colorbar()
plt.title('Disparities in helmet citations issued to Black bicyclists         ',size=14,y=1.2)
plt.text(0.09,0.055,'$^{*}$ Note: only cities/towns with at least 10 helmet citation records are shown.',
         size=9,transform=plt.gcf().transFigure,style='italic');
plt.tight_layout()
plt.savefig(current_results_dir + 'king_co_helmet_citation_disparities.pdf')

# fines ordered, assessed, and paid for citations
fig = plt.figure(figsize=(7.5,6.5),facecolor='w')
cities_with_data = (cities['Fines charged - number of data points'] >= 5)
cities_with_data = list(cities_with_data[cities_with_data].index)
helmet_mask = kc_citations['Law Description'].str.contains('Bicycle Helmet Required')
non_dup_rows = ~kc_citations['Case Number'].duplicated(keep=False)
helmet_non_dup_mask = logical_and(helmet_mask,non_dup_rows)
helmet_ordered_grouped = kc_citations.loc[helmet_non_dup_mask].groupby('City')['AR Ordered Amount']
helmet_adjusted_grouped = kc_citations.loc[helmet_non_dup_mask].groupby('City')['AR Adjusted Amount']

gs = fig.add_gridspec(3,1)
ax1 = fig.add_subplot(gs[0:2])
x_off = 0.15
labels = []
def add_label(violin,label):
    color = violin['bodies'][0].get_facecolor().flatten()
    labels.append((mpatches.Patch(color=color),label))
for c_idx, city in enumerate(cities_with_data):
    med = cities.loc[city,'Fines charged - median ($)']
    vp1 = plt.gca().violinplot(helmet_ordered_grouped.get_group(city).dropna(),positions=[c_idx-x_off],
                               vert=True,widths=0.25,showmeans=True,showextrema=False,
                               showmedians=False,quantiles=None)
    for vp_element in vp1['bodies']:
        vp_element.set_color('darkolivegreen')
        vp_element.set_alpha(0.5)
        vp_element.set_zorder(2)
    vp1['cmeans'].set_color('darkolivegreen')
    vp1['cmeans'].set_zorder(3)
    if c_idx == 0: add_label(vp1,'Fines charged')
    charged_mean = cities.loc[city,'Fines charged - average ($)']
    plt.text(c_idx-x_off*1.5,charged_mean,f'${int(round(charged_mean,0))}',fontsize=7,
             horizontalalignment='right',verticalalignment='center')
        
    vp2 = plt.gca().violinplot(helmet_adjusted_grouped.get_group(city).dropna(),positions=[c_idx+x_off],
                               vert=True,widths=0.25,showmeans=True,showextrema=False,
                               showmedians=False,quantiles=None)
    for vp_element in vp2['bodies']:
        vp_element.set_color('steelblue')
        vp_element.set_alpha(0.5)
        vp_element.set_zorder(2)
    vp2['cmeans'].set_color('steelblue')
    vp2['cmeans'].set_zorder(3)
    if c_idx == 0: add_label(vp2,'Fines assessed after adjustments')
    assessed_mean = cities.loc[city,'Fines assessed after adjustments - average ($)']
    plt.text(c_idx+x_off*1.5,assessed_mean,f'${int(round(assessed_mean,0))}',fontsize=7,
             horizontalalignment='left',verticalalignment='center',zorder=3)
xticklabels = [' '] + ['\n{0}\n($n$ = {1})'.format(city,int(cities.loc[city,'Fines charged - number of data points'])) \
                       for city in cities_with_data]
for l_idx, label in enumerate(xticklabels):
    if 'Seattle' in label:
        xticklabels[l_idx] = xticklabels[l_idx].replace('Seattle','Seattle**')
    elif 'Lake Forest Park' in label:
        xticklabels[l_idx] = xticklabels[l_idx].replace('Lake Forest Park\n(','Lake Forest\nPark\n(')
plt.gca().set_xticklabels(xticklabels)

plt.xticks(fontsize=8)
plt.yticks(fontsize=8)
labels.append((plt.plot(NaN,NaN,c='k')[0],'Average value'))
plt.legend(*zip(*labels),ncol=3,frameon=False,fontsize=8,
           bbox_to_anchor=(0,1.1),loc='upper left')
plt.ylim([0,260])
plt.xlim([-0.6,len(cities_with_data)-1+0.6])
plt.grid(alpha=0.5,lw=0.5,zorder=1)
plt.gca().yaxis.set_major_formatter('${x:1.0f}')
plt.ylabel('Helmet violation fine amount')
plt.text(0.53,0.97,'Helmet law fines charged/assessed and fraction unpaid',size=14,
         horizontalalignment='center',transform=plt.gcf().transFigure)
plt.title('(Note: fines charged include base violation and court fees, including late penalties; adjustments\n' \
          + 'may include court-issued reductions or waivers, such as for proof of helmet purchase)',
          style='italic',size=8,y=1.10)

ax2 = fig.add_subplot(gs[2])
plt.bar(x=range(len(cities_with_data)),
        height=cities['Fraction of citations unpaid'][cities_with_data],
        width=0.25,color='0.5',edgecolor='0.3',alpha=0.6,zorder=3,
        label='Percent of helmet citations unpaid')
for c_idx, city in enumerate(cities_with_data):
    frac_unpaid = cities['Fraction of citations unpaid'][city]
    plt.text(c_idx,frac_unpaid,f'{int(round(frac_unpaid,0))}%',fontsize=7,
             horizontalalignment='center',verticalalignment='bottom',zorder=3)
plt.xlim([-0.6,len(cities_with_data)-1+0.6])
plt.ylim([0,100])
plt.gca().yaxis.set_major_formatter(mtick.PercentFormatter())
plt.tick_params(axis='x',which='both',top=True,bottom=False,labelbottom=False)
plt.tick_params(axis='both',labelsize=8)
plt.grid(alpha=0.5,lw=0.5,zorder=1)
plt.legend(fontsize=8,frameon=False,bbox_to_anchor=(0,-0.02),loc='upper left')
plt.ylabel('Citations unpaid')

note_text = '$^{*}$ Note: only cities/towns with at least 5 helmet citation financial records are shown, and only helmet citations issued as the sole\n            violation are included due to the inability to disaggregate multiple charges issued together. Instances of \$0 in fines\n            assessed after adjustments are considered paid citations in the bottom panel.\n$^{**}$ Fine records for Seattle shown here do not include late/default penalties, commonly \$52 when a defendant fails to respond\n            within 19 days of receiving a citation.'
plt.text(0.10,0.004,note_text,
         size=7,transform=plt.gcf().transFigure,style='italic');

plt.tight_layout()
plt.savefig(current_results_dir + 'king_co_helmet_citation_fines.pdf')

# helmet citation records per year in Seattle
seattle_helmet_mask = logical_and(kc_citations['City'] == 'Seattle', kc_citations['Law Description'] == 'Bicycle Helmet Required')
seattle_helmet_citations = kc_citations[seattle_helmet_mask]
seattle_helmet_citations_per_year = seattle_helmet_citations.groupby(seattle_helmet_citations['Violation Date'].dt.year).count().max(axis=1)

# Seattle cyclist demographics, from notebook 'csg_bike_infraction_analysis.ipynb'
#     stats copied from displayed chart: 'Seattle cyclist demographics (frequency-weighted using SDOT/EMC results) inferred from PRR survey'
percent_white_nonhispanic_cyclists = 70.726539
percent_white_hispanic_cyclists = 5.166718
percent_white_cyclists = percent_white_nonhispanic_cyclists + percent_white_hispanic_cyclists
percent_black_cyclists = 4.714178

# helmet citation racial disparities over time in Seattle
def disparity_function(input_dataframe):
    citation_count_total = input_dataframe['Defendant Race'].value_counts().sum()
    percent_white_citations = 100 * sum(input_dataframe['Defendant Race'] == 'White') / citation_count_total
    percent_black_citations = 100 * sum(input_dataframe['Defendant Race'] == 'Black') / citation_count_total
    # percent_white_census = cities.loc['Seattle']['Census (2015-2019) - White (%)']
    # percent_black_census = cities.loc['Seattle']['Census (2015-2019) - Black (%)']
    black_white_disparity = (percent_black_citations / percent_black_cyclists) / (percent_white_citations / percent_white_cyclists)
    return black_white_disparity
bw_disparity_by_year = seattle_helmet_citations.groupby(seattle_helmet_citations['Violation Date'].dt.year).apply(disparity_function)
average_bw_disparity = disparity_function(seattle_helmet_citations)

# plot
plt.figure(figsize=(7.5,3.5),facecolor='w')
plt.plot(bw_disparity_by_year,c='k',zorder=2)
plt.scatter(bw_disparity_by_year.index,bw_disparity_by_year.values,s=seattle_helmet_citations_per_year.values,c='k',zorder=2)
leg1 = plt.scatter(NaN,NaN,s=25,c='k',label='10 tickets')
leg2 = plt.scatter(NaN,NaN,s=50,c='k',label='50 tickets')
leg3 = plt.scatter(NaN,NaN,s=100,c='k',label='200 tickets per year')
plt.plot([2002,2021],[average_bw_disparity,average_bw_disparity],c='b',ls='--',zorder=1)
plt.text(2002.5,2.75,'Average ({0:.1f}x disparity)'.format(average_bw_disparity),color='b')
plt.xticks(arange(2003,2021,2))
plt.yticks(arange(0,16,3))
plt.xlim([2002,2021])
plt.xlabel('Year')
plt.ylabel('Rate of citations to Black cyclists\ncompared to white cyclists')
plt.title('Black-white disparities in Seattle helmet citation records')
plt.grid(lw=0.5,alpha=0.5,zorder=1)
plt.legend(loc='upper right',ncol=3,frameon=False)
plt.tight_layout()
plt.savefig(current_results_dir + 'seattle_helmet_citation_disparities_by_year.pdf')

# compare 2008-2010 with 2018-2020 average disparities (weighted by # tickets in records)
average_0 = (bw_disparity_by_year[2008] * seattle_helmet_citations_per_year[2008] + \
             bw_disparity_by_year[2009] * seattle_helmet_citations_per_year[2009] + \
             bw_disparity_by_year[2010] * seattle_helmet_citations_per_year[2010]) / \
            (seattle_helmet_citations_per_year.loc[2008:2010].sum())
print('2008-2010 average disparity, weighted by # annual recorded tickets: {0:.2f}x'.format(average_0))
average_1 = (bw_disparity_by_year[2018] * seattle_helmet_citations_per_year[2018] + \
             bw_disparity_by_year[2019] * seattle_helmet_citations_per_year[2019] + \
             bw_disparity_by_year[2020] * seattle_helmet_citations_per_year[2020]) / \
            (seattle_helmet_citations_per_year.loc[2018:2020].sum())
print('2018-2020 average disparity, weighted by # annual recorded tickets: {0:.2f}x'.format(average_1))