ACS Japanese Speaker Data

I needed some language data, so I ended up making something similar to this site: https://probablyhelpful.com/languages/Japanese.html. Except with newer data, more visualizations, searchable tables, and some helpful statistics.

Why

There are a number of factors unique to Japanese that create upward wage pressure for bilingual speakers.

From the demand side:

1. Japan is and has been the country with the most direct investment in the US. 1, 2
   
2. Japan is the fourth largest economy (investment target) 3, fifth when adjusting for PPP 4

On the supply side:

1. Japan ranks poorly in the “English Proficiency Index”. 5
   
2. Japanese is among the hardest language to learn for English speakers (along with Arabic, Chinese, Korean). 6
   
3. Japan has a small emigrant population relative to its population size. 7
   
4. Regional concentrations/disparities within the US (the purpose of this analysis). 8

Together, these factors make Japanese a uniquely valuable language in the US. China/Chinese is similar in some respects, but lacks, basically any, direct investments into the US. For reference, in 2025 Austria had more direct investment into the US than China.

As an aside, I wish the BLS would release language specific data on Interpreters and Translators. Rather than the general translator bucket. (How am I supposed to interpret the wage data!)
There is also the complication of specializations that require language ability as an addition, but are not a “translator” role. With the right data I imagine they could create an average % wage premium that a language adds.
An example, wherein the wage premium of a foreign language skill (FLS) was identified (in Poland).

Data

The ACS question:

The data originates from https://data.census.gov/.

U.S. Census Bureau. “Language Spoken at Home by Ability to Speak English for the Population 5 Years and Over.” American Community Survey, ACS 5-Year Estimates Detailed Tables, Table B16001.

full questionnare: https://www.census.gov/programs-surveys/acs/about/forms-and-instructions.html

Potential issues with the data to be concious of:

• The errors are based on a 90% confidence interval.
  
• ACS data is not a census and may suffer from non-sampling error.
  
• Data is for all age groups when one might only be interested in those of working age (18-64) my best estimate puts it at ~60%.
  • This 60% figure can also invite scrutiny. Explained further below.
    
• The bilingual statistic is based on those who responded as speaking english “very well”, excluding those who selected as speaking “well” or below.
  
• Individuals who know the language may interpret “speak language at home” more strictly and not respond with the language spoken. ie. A language learner w/o someone to speak w/ at home.
  
• Speaking at home is used as a proxy to fluency, but such fluency may be considered insufficient by other standards. (尊敬語 or reading/writing skills come to mind)

Observations

Hawaii has the highest proportion of Japanese, but due to size Hawaii actually has fewer bilingual speakers than the NYC metro area (20552 Bilingual speakers in Hawaii; 23268 Bilingual speakers in NY-NJ metro area).

Alaska, Utah, Nevada stood out as locations not traditionally associated with a Japanese diaspora, but had relatively high populations of Japanese speakers.

Florida and Texas have a low proportion of Japanese speakers, but make up for it by sheer size.

Tables & Visuals 表・見える化

Nominal / Proportional

import pandas as pd
import geopandas as gpd
import numpy as np
import folium
import itables
import geoplot
import mapclassify as mc
import geoplot.crs as gcrs
import matplotlib

jpmap = gpd.read_file("../../census/jpmap.gpkg")
itables.options.css = """
.dt-column-header {
  font-size: small;
}
"""

itables.show(pd.DataFrame(jpmap.loc[:,["ST","state","population","epopulation","eng","eeng","jpn","ejpn","bilingual","ebilingual",
                            "jpnloweng","ejpnloweng","pctjpn","pctbilingual","bilingualratio"]]), 
                            maxBytes=200000, style="width:100%", 
                            buttons=["pageLength", {"extend": "colvisGroup","text": "Hide Errors","hide": [1,3,5,7,9,11]},
                                                    {"extend": "colvisGroup","text": "Show all","show": "*"}],
                            columnDefs=[{"className":"dt-left", "targets":[0]}, 
                                        {"targets":[1,3,5,7,9,11], "visible":False}],
                            columnControl=["order"],
                            ordering={"indicators": False, "handler": False})

ⓘST	state	population	epopulation	eng	eeng	jpn	ejpn	bilingual	ebilingual	jpnloweng	ejpnloweng	pctjpn	pctbilingual	bilingualratio
AL	alabama	4793030	601	4503218	5359	2528	541	1228	414	1300	327	0.052743	0.025621	48.575949
AK	alaska	689070	169	585162	2727	1494	390	1033	281	461	254	0.216814	0.149912	69.143240
AZ	arizona	6981320	285	5184054	11544	7952	907	5119	706	2833	511	0.113904	0.073324	64.373742
AR	arkansas	2869111	548	2630451	4441	935	266	495	195	440	218	0.032588	0.017253	52.941176
CA	california	37121454	487	20642638	54450	133419	4594	78900	3190	54519	2222	0.359412	0.212546	59.137004
CO	colorado	5549558	577	4650283	9223	5893	658	4494	563	1399	266	0.106189	0.080979	76.259969
CT	connecticut	3442817	90	2638366	6723	3281	640	2322	504	959	329	0.095300	0.067445	70.771106
DE	delaware	966580	46	820336	3432	805	272	603	229	202	102	0.083283	0.062385	74.906832
DC	district of columbia	640841	66	518478	3204	1054	285	692	244	362	145	0.164471	0.107983	65.654649
FL	florida	21289728	707	14769458	25976	12222	1183	8153	993	4069	647	0.057408	0.038295	66.707577
(41 more rows not shown)

Map

m = folium.Map(location=[35.3, -97.6], zoom_start=4)

for idx, row in jpmap.iterrows():
    popup_html = f"""
    <style>
        .leaflet-popup-content-wrapper,
        .leaflet-popup-content {{
            min-width: 300px !important;
            max-width: 400px !important;
        }}
        table {{
            table-layout: fixed;
            width: 100%;
            border-collapse: collapse;
        }}
        th, td {{
            font-size:12px;
            border: 1px solid black;
            padding: 6px 8px;
            white-space: nowrap;
            overflow: hidden;
            text-overflow: ellipsis;
        }}
        th {{background-color: #f5f5f5; font-weight: bold;}}
    </style>
    <table>
        <colgroup>
            <col style="width: 65%">
            <col style="width: 35%">
        </colgroup>
        <tr>
            <th colspan="2" style="font-weight:bold;">{row['ST']} - {row['state'].title()}</th>
        </tr>
        <tr>
            <th>Population</th>
            <td>{row['population']:,}</td>
        </tr>
        <tr>
            <th>Bilingual Rank (Nominal)</th>
            <td>{row['nombilingualrank']} (x̄ = {row['bilingual']:,})</td>
        </tr>
        <tr>
            <th>Bilingual Rank (Per Capita)</th>
            <td>{row['bilingualrank']} ({row['pctbilingual']:.4f}%)</td>
        </tr>
        <tr>
            <th>JPN Speaker Rank (Nominal)</th>
            <td>{row['nomjpnrank']} (x̄ = {row['jpn']:,})</td>
        </tr>
        <tr>
            <th>Bilingual Ratio</th>
            <td>{row['bilingualratio']:.4f}</td>
        </tr>
    </table>
    """
    bins = mc.FisherJenks(jpmap["bilingual"], k=6).bins.astype(int)
    cmap = matplotlib.colormaps.get_cmap("seismic")
    gradient = np.linspace(0, 1, 6)
    colors = [matplotlib.colors.to_hex(cmap(i)) for i in gradient]
    fcolor = colors[np.digitize(row["bilingual"]*.99999,bins)]
    folium.GeoJson(
        row.geometry,
        style_function=lambda x, fc=fcolor: {"fillColor": fc, "fillOpacity":"65%","color":"white"},
        tooltip=f"{row.state.title()}",
        popup=folium.Popup(popup_html)
    ).add_to(m)

legend_html = """
<div style="position:fixed; bottom:40px; left:50px; width:240px; background:rgba(255,255,255,0.8); 
            padding:12px 14px; border:2px solid #666; border-radius:6px; z-index:999; 
            box-shadow: 2px 2px 6px rgba(0,0,0,0.2); color:black;text-align:center;">
    <div style="font-weight:bold; font-size:15px; margin-bottom:5px;">
        Bilingual Population</div>
    <div style="height:14px; background:linear-gradient(to right,#00004c 0%,#0000db 20%,#9999ff 40%,#ff9999 60%,#e60000 80%,#800000 100%); 
                border-radius:3px; margin-bottom:6px;"></div>
    <div style="font-size:12px; text-align:center;">
        Lower &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; ⇒ &nbsp;&nbsp;&nbsp;&nbsp;&nbsp;&nbsp; Higher</div></div>
"""
m.get_root().html.add_child(folium.Element(legend_html))
m

Make this Notebook Trusted to load map: File -> Trust Notebook

---

I wanted to try a cartogram, and I guess Florida ended up even further underwater.

dc48 = jpmap[~jpmap.state.isin(["alaska", "hawaii"])]
scheme = mc.FisherJenks(dc48["bilingual"], k=6)
nb = mc.FisherJenks(dc48["bilingual"], k=6).bins
scheme.bins = np.round(nb, decimals=3)
ax = geoplot.cartogram(
    dc48, scale="bilingual", projection=gcrs.AlbersEqualArea(),
    legend=True, legend_kwargs={'bbox_to_anchor': (0.22, 0.42)}, legend_var="hue",
    hue="bilingual", scheme=scheme, cmap="seismic"
)
ax.set_title("Japanese Bilinguals (Nominal) of Contiguous States")
geoplot.polyplot(dc48, facecolor="lightgray", edgecolor="white", ax=ax)

scheme = mc.FisherJenks(dc48["pctbilingual"], k=6)
nb = mc.FisherJenks(dc48["pctbilingual"], k=6).bins
scheme.bins = np.round(nb, decimals=3)
ax = geoplot.cartogram(
    dc48, scale="pctbilingual", projection=gcrs.AlbersEqualArea(),
    legend=True, legend_kwargs={'bbox_to_anchor': (0.22, 0.42)}, legend_var="hue",
    hue="pctbilingual", scheme=scheme, cmap="seismic"
)
ax.set_title("Japanese Bilinguals per Capita of Contiguous States (% of Pop.)")
geoplot.polyplot(dc48, facecolor="lightgray", edgecolor="white", ax=ax)

Regional

regionallang = jpmap.loc[:,["population","eng","jpn","bilingual","jpnloweng","region"]].groupby(["region"], as_index=False).sum()
regionallang["pctjpn"] = (regionallang.jpn * 100) / regionallang.population
regionallang["pctbilingual"] = (regionallang.bilingual * 100) / regionallang.population
itables.show(regionallang, maxBytes=200000, style="width:100%",
                            columnDefs=[{"className":"dt-left", "targets":[0]}],
                            columnControl=["order"],
                            ordering={"indicators": False, "handler": False})

ⓘregion	population	eng	jpn	bilingual	jpnloweng	pctjpn	pctbilingual
midwest	65194303	57229624	58473	33954	24519	0.089690	0.052081
northeast	54393479	41199327	68163	41873	26290	0.125315	0.076982
south	121811243	96711452	96036	59928	36108	0.078840	0.049197
west	74743523	50626636	239733	148271	91462	0.320741	0.198373

Metro Areas

jpmetro = pd.read_csv("../../census/jpmetro.csv")

itables.show(pd.DataFrame(jpmetro.loc[:,["metro-area","population","epopulation","eng","eeng","jpn","ejpn","bilingual","ebilingual","pctjpn","pctbilingual","bilingualratio"]]), 
                            maxBytes=200000, style="width:100%", 
                            buttons=["pageLength", {"extend": "colvisGroup","text": "Hide Errors","hide": [2,4,6,8]},
                                                    {"extend": "colvisGroup","text": "Show all","show": "*"}],
                            columnDefs=[{"className":"dt-left", "targets":[0]}, 
                                        {"targets":[2,4,6,8], "visible":False}],
                            columnControl=["order"],
                            ordering={"indicators": False, "handler": False,},
                            classes="display")

ⓘmetro-area	population	epopulation	eng	eeng	jpn	ejpn	bilingual	ebilingual	pctjpn	pctbilingual	bilingualratio
atlanta-sandy springs-roswell, ga metro division; atlanta-sandy springs-roswell, ga metro area	4591766	311	3642234	10169	6239	1064	3574	761	0.135874	0.077835	57.284821
marietta, ga metro division; atlanta-sandy springs-roswell, ga metro area	1301511	115	1076929	5376	1237	306	859	248	0.095043	0.066000	69.442199
boston, ma metro division; boston-cambridge-newton, ma-nh metro area	1947655	150	1421953	6523	3073	660	2165	531	0.157779	0.111159	70.452327
cambridge-newton-framingham, ma metro division; boston-cambridge-newton, ma-nh metro area	2325264	78	1652376	7223	4417	480	3116	405	0.189957	0.134006	70.545619
rockingham county-strafford county, nh metro division; boston-cambridge-newton, ma-nh metro area	431393	101	404899	1812	140	80	135	81	0.032453	0.031294	96.428571
chicago-naperville-schaumburg, il metro division; chicago-naperville-elgin, il-in metro area	6789711	111	4576500	12296	9881	1000	5936	774	0.145529	0.087426	60.074891
elgin, il metro division; chicago-naperville-elgin, il-in metro area	712200	39	517436	3885	375	194	256	158	0.052654	0.035945	68.266667
lake county, il metro division; chicago-naperville-elgin, il-in metro area	676256	4	463738	3120	1269	292	901	243	0.187651	0.133234	71.000788
lake county-porter county-jasper county, in metro division; chicago-naperville-elgin, il-in metro area	682320	96	596637	2864	268	158	191	132	0.039278	0.027993	71.268657
dallas-plano-irving, tx metro division; dallas-fort worth-arlington, tx metro area	5038495	155	3303773	12593	7177	1326	4410	1043	0.142443	0.087526	61.446287
(27 more rows not shown)

Other Stats

The data does not provide information on age (we might be interested in working age pop).
Below is my best estimate at a working age population correction factor.

203,771,037 / 316,142,548 = 64.5%	of the general populace is working age (18 to 64)
330,000 * 67% ~= 220,000	Working age JPN immigrants
1,060,000 * 58% ~= 615,000	Working age native born JPN-Americans
(220,000 + 615,000) / 1,390,000 ~= 60%	Estimate for working age population among JPN in the US.
pew fact sheet using ~2021-2023 ACS data

One issue with this estimate is the use of race/ethnicity instead of language ability. However the alternative, language ability/age dataset, is grouped up into the larger group of API (Asia/Pacific Islander). This is problematic due to the sheer size and diversity of individuals coming from Asia (Japanese immigrants are but a small fraction of that).